aiLL June 11, 2023 - 2 - The Future of AI

Description

Chapters

Transcript

0:11 foreign
0:24 with the AI learning lab
0:27 just had a lovely dinner I'm coming back
0:29 doing a part two of this live
0:32 and uh if you're uh if you're a long
0:37 time fan I've been in lives for about
0:39 two weeks
0:40 so if you're a fan of the last two weeks
0:43 uh welcome if you're new uh this is the
0:46 AI learning lab I'm Kyle Shannon I'm an
0:48 entrepreneur uh my current position I'm
0:51 co-founder and CEO of a company called
0:53 storyvine and uh I've been deep diving
0:57 into generative AI for about the past
0:59 year and there's a lot of parallels to
1:02 what's going on with AI right now to
1:04 what was happening in the early days of
1:06 the World Wide Web where I did a lot of
1:08 work there as well
1:10 and so uh
1:12 so this channel is just about talking
1:15 about what's going on with AI why it's
1:17 interesting demystifying it answering
1:19 any questions I can if you have
1:20 questions about generative AI pop them
1:23 into the
1:24 um
1:26 into the comments and I will answer what
1:29 I can and I won't [ __ ] you if I
1:32 can't and I cuss
1:35 and uh what else uh I make silly voices
1:39 and use a lot of self-deprecating humor
1:42 so if any of those things bother you
1:47 um what else
1:50 um the the the uh the parallels I see to
1:53 the early days of the web so so
1:57 the internet has been around for a bunch
1:59 of decades and it was being used by wait
2:01 what do I think of people
2:04 thinking of the Tom Cruise stunt double
2:07 wait
2:08 what do you think of people thinking the
2:11 Tom Cruise stunt double photos are real
2:14 Oh you mean the the the tick tock videos
2:17 the the one the fake Tom Cruise ones is
2:20 that I don't know if that's what you're
2:21 talking about but um
2:24 well I think they think they're real
2:25 because they look pretty real right the
2:27 Deep fake stuff is getting good and it's
2:29 going to get incredibly good in um in
2:33 the next couple of years so
2:36 um so what do I think of those people I
2:37 think I think that
2:41 they they uh
2:43 they saw something that looked like Tom
2:45 Cruise and the guy the guy's a good
2:46 enough actor and he's you know he's got
2:48 enough of the mannerisms of Tom Cruise
2:49 but but the quite frankly these
2:52 Technologies are getting so good that
2:53 you know I'm just going to be able to
2:55 talk like me and it will pick up you
2:58 know his intonation and his his
3:00 mannerism so I think that's coming
3:04 um I have a sophisticated uh I have a
3:07 sophisticated camera and lighting setup
3:09 I assume that's a reference to the last
3:11 part of the live where I showed you the
3:13 cardboard boxes I have stacked up
3:16 to put my my Lighting on in my uh helmet
3:20 box to put my laptop on uh anyway if
3:22 you're new it looks like there's a bunch
3:23 of people here uh my name is Kyle
3:25 Shannon this is the AI learning lab
3:27 doing what I can to demystify generative
3:30 AI the the URLs right here they're the
3:33 reason they're here is the top one is
3:34 the official chat CPT website if you've
3:37 not played with chat GPT
3:39 please do
3:41 um this stuff is coming do the the whole
3:44 purpose of this channel is to get people
3:46 curious about generative AI get them
3:48 engaged I have no affiliation with any
3:50 of these sites so
3:52 um they're really just there's a public
3:53 service announcement bing.com is
3:55 Microsoft search engine but it now has a
3:57 chat button and that's chatgpt4
3:59 connected to the internet so you can use
4:01 gpt4 for free if you've heard of that
4:05 um and then the bottom on po.com is uh
4:08 one where you can use GPT Claude and uh
4:13 what's the other one called I forgot
4:14 what it's called but anyway there's a
4:15 bunch of models there you can play with
4:17 and then prompts.chat will teach you
4:19 about prompting and uh futurepedia.io
4:23 will is just a giant pile of AI sites
4:27 and apps and services that
4:30 um
4:31 you know there's like like 10 000 rabbit
4:34 holes there for you to run down so so
4:36 that's what's going on if you have
4:37 questions pop them in the uh in the
4:39 comments below I'll do my best to answer
4:41 them I will fall behind on questions so
4:44 it gets frustrating for those that are
4:46 not initiated because I actually try to
4:47 answer all the questions and then I get
4:49 behind and people are like he's not even
4:51 listening to me who I only listen to me
4:53 I'm important too
4:57 so those people usually ask the same
4:59 question 472 times if it's in there I'll
5:03 get to it yay I'm so glad you're here
5:04 thank you appreciate that how will you
5:07 know what's real or fake
5:09 um so
5:12 what you're trying to trying to survive
5:15 awesome awesome name an awesome question
5:19 how will we know what's real or fake I
5:21 don't think we will I think that the
5:24 2024 U.S presidential election is going
5:28 to be an absolute [ __ ] [ __ ] show it's
5:30 going to make 2016 and 2020 look like
5:33 Child's Play in terms of misinformation
5:35 and deep fakes and things like that and
5:38 it you know you know three days before a
5:41 vote I guarantee you there'll be some
5:44 you know some politician French kissing
5:46 a gorilla at the Cincinnati Zoo or
5:49 something like that
5:51 um
5:52 so so I I think that this is something
5:55 that we're gonna have to deal with I
5:56 think the I think the ramifications for
5:59 it will be probably starting in 2025
6:01 there's going to be a real demand and
6:04 probably some government interaction for
6:07 authenticity how do we tell what's
6:08 authentic or not right now right now
6:11 there's there's sort of this weird thing
6:12 going like you know we should Watermark
6:14 it if it's generated by AI
6:16 and there's a couple of things that make
6:20 that just weird you know one is
6:24 like a piece of writing if you write it
6:27 in chat GPT anyone who's used the tool
6:30 you don't just like put in a prompt and
6:31 then you get out a final piece that you
6:33 publish like it tends to be very much a
6:35 back and forth It's you're like
6:36 collaborating with this tool and so
6:39 almost everything that I've put out in
6:41 the world is some combination of some
6:43 chat GPT stuff and my stuff you know
6:47 intermingled and as Microsoft and Google
6:49 roll these tools into Office 365 and the
6:52 Google office suite
6:54 you're not going to be able to tell so I
6:56 think that's at a bit of a McGuffin a
6:59 kind of a false
7:00 um sense of security like you know who
7:03 gives a [ __ ] if someone actually used a
7:05 generative AI tool to produce a piece
7:08 what's what's more relevant is
7:11 is this person of you know is this video
7:14 of of this person saying these things
7:17 you know can we verify that that person
7:20 was recorded saying those things
7:24 um and and that's I I assume that what
7:26 we'll see
7:27 are some sort of tokenization
7:29 Technologies whether it's blockchains or
7:31 nfts or other kind of tokenization are
7:34 going to be layered on top of media
7:37 um so that so that we can know but right
7:40 now I think I think we're about to enter
7:43 a very cloudy you know those magic eight
7:46 balls that you flip over you know answer
7:48 hazy ask again uh I think we're about to
7:51 enter two years of answer hazy ask again
7:53 was that politician really French
7:55 kissing the gorilla in the Cincinnati
7:58 Zoo
8:00 and then there's going to be you know
8:02 because there's going to be that doubt
8:03 on that side there's going to be some
8:05 real footage of a politician French
8:07 kissing a gorilla at the Cincinnati Zoo
8:09 and they're gonna be like well that's
8:10 clearly a deep fake because I couldn't
8:12 be that weird you know so I think we're
8:15 gonna see it on both sides
8:18 um all right if you have questions pop
8:20 them down below I'll do my best to
8:22 answer them
8:23 um by the way so a lot of people who pop
8:25 in here who don't know who I am they're
8:26 like What's your expertise
8:29 um I'm not an ML and AI guy I'm an
8:32 entrepreneur
8:33 um I've I've started I don't know 12 or
8:36 15 companies I I tend to as I discover
8:39 new technology Technologies figure out
8:42 ways to put them into the world and
8:44 and innovate with them so I I would say
8:47 I represent
8:48 try again
8:50 oh Siri Siri
8:53 yeah I know you're Siri can we put you
8:55 away thank you
8:58 Siri just took over my machine anyway
9:02 um
9:03 I I think I represent
9:04 um the the the kind of second wave of of
9:08 AI users so I'm not a machine a data
9:11 scientist or or a machine learning
9:14 specialist but I'm someone who is an
9:16 early adopter and figures out how to use
9:18 these tools in business and in life and
9:20 things like that so that's my expertise
9:22 is just more Hands-On
9:24 than creating this [ __ ]
9:27 um
9:27 I I really like to try to understand how
9:30 this stuff works and what it actually
9:33 makes possible I did this I did a kind
9:35 of Deep dive with blockchain and nfts
9:37 two years ago
9:38 and it was like the deeper I dug into
9:40 that the more those companies seemed
9:42 like [ __ ] Fly by Night grifting
9:44 operations which a lot of them were um
9:47 and the more I dig into the generative
9:48 AI stuff the more profound it seems so
9:51 so anyway that's what's going on
9:54 um in all caps artificial intelligence
9:56 will be your downfall it it may be I
9:59 mean if you if you listen to Hollywood
10:00 that's what Hollywood's told us for the
10:02 last 50 years I think there's
10:06 I think well there are clearly two paths
10:08 the robots will kill us and the robots
10:10 will free us right those are those are
10:12 the two diverging paths the reality will
10:14 probably be somewhere in the middle
10:17 um
10:19 I I am of the I I am of the deep belief
10:22 that these tools are amplifiers of
10:25 humanity and can actually Elevate us to
10:28 be our best selves now Humanity's
10:31 complicated and if you have a Humanity
10:33 amplifier it's going to amplify the bad
10:35 as well as the good
10:36 so it's very possible that AI will be my
10:40 downfall but I don't think so
10:42 what's your experience with AI I just
10:44 said that
10:45 um can you walk through how you made a
10:48 video of yourself yeah I'll do that I'll
10:50 do that in a second let me get get some
10:51 other questions here but I'll show you
10:53 that
10:54 um geez you only have 170 likes get that
10:57 up a ton all right hit that tap that
10:59 like button what do they say you do can
11:01 make money on my chat
11:08 um I don't think I'm gonna get tired of
11:10 that joke but I'm sure most of you are
11:11 all right do you invest in any crypto
11:15 um I've I'm a Neo Tokyo citizen um so I
11:18 invested a bunch of time to get that
11:21 um
11:22 I I have invested in crypto and I have
11:26 successfully parlayed hundreds and
11:28 hundreds of dollars into completely
11:30 worthless pieces of [ __ ] nfts
11:34 uh Neo Tokyo is still a cool project
11:36 though
11:37 um whenever I hear Joe Biden's voice I
11:39 think it's AI at first it's grossly
11:41 under regulated AI is grossly under
11:44 regulated yeah except the challenge here
11:46 is like The Regulators don't even
11:48 [ __ ] understand how Wi-Fi works much
11:51 less something is sophisticated as AI
11:53 they need to very very quickly get like
11:56 Jeffrey Hinton
11:58 who's called The Godfather of IAAI just
12:00 left Google so that he could talk about
12:02 the dangers and perils and you know
12:05 risks of AI
12:06 they should scoop him up immediately or
12:08 someone like him
12:10 um to be someone within government that
12:13 actually understands this [ __ ] and can
12:14 guide the policy because they got to get
12:16 because it's coming fast and I don't
12:18 think they're quite aware of that yet
12:19 they still haven't figured out what a
12:21 tweet is
12:24 Christ we're in trouble
12:26 okay uh
12:30 um what's the best way to test
12:33 um if a third party language mode works
12:37 a third-party language mode I don't
12:39 quite know what that question's about
12:40 can I talk about AI brain hacking yeah
12:44 um I can't talk too deeply about it
12:47 um
12:49 but I think we're about to enter a
12:51 golden age of of AI brain hacking so so
12:54 if you've been paying attention they've
12:56 now successfully
12:58 used fmri scans and large language
13:00 models combined to be able to recreate
13:04 videos that people watched
13:07 um recreate images that they looked at
13:10 or were thinking about
13:12 um there was a a guy I think in France
13:15 or I think it might have been Italy
13:17 um paralyzed you know from mid back down
13:19 from a motorcycle accident they created
13:22 a Bluetooth Bridge
13:24 um around the injured part of his spine
13:26 they used a large language model to scan
13:29 his brain when he was thinking about
13:30 moving his legs and they sent those
13:32 signals to electrodes in his spine and
13:35 now he's walking
13:36 um so I think we're just getting started
13:38 and then on top of that
13:40 Elon musk's neural link was approved by
13:43 the FDA for human trials so they're
13:45 they're doing direct brain you know
13:48 electrode implants that are going to be
13:50 interfaces to
13:52 um all this stuff it was interesting
13:55 when I first heard about neural link I'm
13:57 like okay so it's a faster interface to
13:59 your phone I didn't quite get a AI yet
14:01 like why would you want a faster
14:03 interface to your phone now that I
14:04 understand this AI stuff and and how
14:06 powerful it is and and what it makes
14:08 possible I mean you can imagine all
14:11 sorts of advancements from neural
14:13 degenerative disease to
14:16 um I don't know learning disabilities to
14:19 injuries
14:21 um and and I it's I think I think it's
14:23 going to be a quite remarkable next
14:26 um decade in terms of biology
14:31 um
14:32 so who was it that asked I think it was
14:33 Anne that asked Can I show how I made
14:35 something so if you go to my if you go
14:37 to my Tick Tock Channel
14:39 the I think it's like this
14:41 the second most recent video there's a
14:44 video of me like a like a Pixar looking
14:47 version of me
14:49 um that sort of sounds like me but not
14:52 it's it's a synthesized voice with a
14:54 cartoon head
14:56 talking about Transformers and so
14:59 so how I did that we'll start with
15:03 mid-journey so I went to Mid Journey
15:08 and let me go find the area
15:12 where it was there it is
15:17 so so within mid-journey you can okay
15:20 this could be you know like your
15:22 favorite muscle car from your teenage
15:24 years doesn't really matter you can
15:25 upload a photo and then based on that
15:28 photo you can create a prompt so I did a
15:31 prompt here and those things were okay
15:34 but I looked a little surprised so I
15:36 uploaded a different photo
15:38 and then it made me those things where I
15:40 kind of looked like Keith Keith olberman
15:42 and I was like ah not quite and then I
15:44 did like another batch of them and I
15:46 ended up after some iterations I think
15:50 that was one of them I think that was
15:52 the one I ended up choosing oh like that
15:54 one had like a big fat nose
15:57 and didn't quite look like mine and so I
15:59 skinnied up the nose so I think this was
16:01 the one I went with
16:02 and then I took this and I went over to
16:05 photoshop
16:08 and I wonder if I've got this
16:12 out somewhere let me see if I've got it
16:14 do you do open recent
16:18 uh oh that's one of them
16:22 and you know old
16:26 foreign
16:34 [Music]
16:40 I used a photoshop's generative fill I
16:42 took this image and I said there was a
16:45 parade of furries walking by this guy
16:48 and it made that image uh which was
16:50 which was pretty hilarious anyway I'll
16:52 do with this with this other image hang
16:54 on
16:56 um
16:58 so so with this image what I did was so
17:02 with generative fill
17:03 in Photoshop you can like highlight like
17:06 there was this sort of weird picture
17:08 frame behind my head
17:10 and so the first thing I did was I I
17:14 um highlighted that area and then you
17:16 can say generative fill and I can just
17:18 say remove objects
17:23 and then it will think a little bit and
17:26 it should just
17:29 um
17:29 make that a green background
17:34 um
17:37 and not quite and that one's kind of
17:40 worse and that one's okay
17:43 um and then so I'll just leave I'll
17:45 leave that there for now and then I you
17:47 can get
17:52 so I can highlight the center of this
17:54 thing and then invert my selection
17:59 and now I can say generative fill and I
18:02 can fill in the background here and go
18:03 like
18:05 um okay
18:08 blur home office
18:11 cartoon
18:14 and that should fill in the background
18:16 so anyway so so I basically took this
18:19 image out of mid journey I filled in the
18:21 background let's see if it gives us
18:22 anything interesting here because this
18:24 is this is really remarkable technology
18:26 the Photoshop generative fill so it
18:30 didn't it it still kind of left the that
18:33 green stripe in the in the thing which
18:35 is kind of weird so that's not quite
18:37 right but let me see what other options
18:39 it gives me yeah there's one where it
18:41 extends the the image okay
18:43 and there's one that one's pretty good
18:45 it's like actually blurred background
18:46 it's still got that weird crap behind
18:48 his head but anyway so I started so I
18:50 took that image
18:51 then I created a script using chat GPT
18:56 and
18:57 um for the
19:00 um
19:00 is this 11 laps now let me go to 11 laps
19:04 for the voice I took the script of what
19:06 I wanted to say and I went and went into
19:09 11 labs
19:11 and you can synthesize your own voice so
19:14 I recorded like five minutes of me
19:16 talking
19:17 and then I wrote a script and had it
19:20 generate uh and and a synthesized
19:23 version of my voice and then I went over
19:27 to D Dash ID
19:29 Dash id.com
19:33 oops you gotta actually type it in the
19:37 Dash id.com
19:41 and this is a place where I can take the
19:43 picture and that digitized voice and
19:46 marry them together
19:47 and it will automatically do the
19:49 animation so
19:51 let me log in here
19:59 um
20:05 and so I got a bunch of videos that I
20:06 did in here but here's
20:09 here's one that uses there it's Kyle
20:11 Shannon
20:12 with learning lab and today we're going
20:13 to transform your understanding of AI in
20:16 just 60 seconds you ready let's talk
20:19 Transformers and nope
20:21 and here's the actual one hey I use Kyle
20:23 Shannon with the AI learning lab and
20:25 today we're going to transform your
20:27 understanding of AI in just 60 seconds
20:29 you ready let's talk Transformers and no
20:32 not the robots in Disguise
20:36 that's GPT humor for you and no not the
20:40 robots in Disguise uh and then that was
20:42 the video I uploaded to Tiki Taki
20:44 so with a handful of tools and about I
20:47 don't know half an hour
20:49 um I generated that automated image so
20:53 I hope that helps
20:55 we lost some people there we went down a
20:57 little rabbit hole but that's okay
21:00 um if you're digging what you're seeing
21:01 here feel free to share the live tap the
21:03 like button I think that gets other
21:04 people in if you tap the like button
21:07 um if you're new here my name is Kyle
21:09 Shannon this is the AI learning lab I
21:11 can geek out like I just did there or I
21:13 can answer questions or I can explain
21:15 basic stuff if you've not used chat GPT
21:18 that's what these URLs here are the top
21:21 one is
21:23 um
21:23 uh the official chat GPT site and then
21:27 there's Bing which is also chat GPT if
21:29 you go to bing.com and click on the chat
21:30 button and then pro.com is another
21:32 option where you can use multiple models
21:36 um go start playing with it go start
21:38 playing with it all right
21:41 geez how do I only have a seven 170
21:43 likes I don't know because I'm a 58 year
21:45 old man on Tick Tock and I'm not
21:47 twerking maybe I should start twerking
21:55 come on guys let's get his likes up well
21:58 thank you very much I appreciate that
22:01 yeah but I mean I appreciate you trying
22:04 to get my likes up but you know uh
22:06 people come in and they see me kind of
22:07 rambling on incoherently
22:11 um right it's just text you clean it up
22:13 by pasting it into notepad watermarking
22:15 it doesn't work in Sonoma text okay what
22:18 about not I don't know what that was in
22:20 reference to sorry Tick Tock just
22:22 scrolled me all the way to the bottom so
22:23 I'm going back up trying to find where I
22:25 was
22:27 and let's see
22:31 brain hacking Emilio's wife you're back
22:34 how was dinner it was tasty although it
22:36 was pork loin it was a little overcooked
22:37 and if it's a little overcooked it's
22:39 gonna be a little dry it's going to be a
22:40 little untasty
22:41 what are you gonna do through a little
22:43 sauerkraut on there a little barbecue
22:45 sauce was fine but actually after dinner
22:48 I cut up some strawberries that were
22:50 just on the verge of going bad tossed a
22:53 little sugar on them and that was
22:54 awesome
22:56 All Right theory itself rests
22:59 Theory itself rests at a statue which is
23:03 a border
23:05 if you want what I think
23:08 I don't know that that's
23:10 a sentence
23:12 I don't understand what you're talking
23:13 about can you talk about how to prevent
23:16 AI brain hacking
23:20 I don't know what AI brain hacking means
23:22 like if you mean like
23:26 misinformation is going to manipulate us
23:28 and [ __ ] with our brains that ship is
23:30 already sailed
23:32 I mean I was taught about propaganda in
23:35 grade school but I had no [ __ ] idea
23:37 until the 2016 and 2020 elections it's
23:40 like oh I get it now
23:43 um
23:44 I don't I don't know that like it seems
23:47 there's there's kind of a presumption in
23:49 your statement there that we're just all
23:50 gonna definitely ai's gonna come after
23:53 us and hack our brains I don't that's
23:54 not how it works
23:57 um I think people opt in
24:00 all right
24:04 okay all right cute is one word okay
24:07 what are the practical uses for AI in
24:09 the medical field I well I think there
24:11 are a ton of them I mean there are
24:14 there are clinical things like it's
24:16 going to help us
24:18 if you haven't heard of it's It's Worth
24:20 looking up this thing called Alpha fold
24:22 um so Google the a part of Google called
24:25 deepmind I think they're the ones that
24:26 created Alpha fold but anyway they
24:28 created this thing that predicts the
24:30 shape of complex proteins well all
24:32 proteins are complex but how they fold
24:35 and it takes a tremendous amount of time
24:39 in computing power to to do those things
24:41 traditionally and it created something
24:45 like four hundred thousand
24:47 um
24:47 protein predictions in some ridiculously
24:51 short amount of time two or three days
24:53 or two or three weeks whatever it was
24:55 but so so you're going to start to see
24:57 new drugs
25:00 um you're going to start to see
25:02 um cures to diseases
25:05 um you know potentially even things like
25:06 cancer or neurological or genetic
25:08 diseases
25:10 um I think I think we're gonna see maybe
25:12 in the nearer term
25:14 um augmentations of
25:18 um
25:19 like like engagements with doctors
25:21 basically like like these tools helping
25:24 doctors be better in in one of two ways
25:28 the generative AI is going to make the
25:31 doing of the paperwork much faster and
25:33 much more efficient which will give
25:35 doctors more time to spend with patients
25:37 and then the the generative AI can also
25:41 effectively be an assistant to a doctor
25:44 that helps them be more empathetic that
25:47 helps them be more available so imagine
25:50 a physical therapist that creates an
25:53 avatar of themselves and so you meet
25:56 with the physical therapist they give
25:58 you some things and then if you have
25:59 questions you have access to their
26:01 Avatar 24 7. and their avatars really
26:05 polite really knowledgeable knows your
26:08 entire history you can you can ask the
26:10 Avatar questions it'll give you answers
26:12 and if anything that you're talking
26:14 about with that Avatar makes it clear
26:16 that you need to talk to the real doctor
26:17 it'll escalate that so so you know we're
26:21 entering a time when the boomers are are
26:24 starting to age into needing a lot of
26:27 medical care so we're gonna there's
26:29 gonna be a huge there's already a huge
26:31 nursing short shortage there's going to
26:33 be a huge health care provider shortage
26:36 and I think generative AI can help with
26:38 that
26:39 and then on the administrative side
26:43 dealing with things like Insurance
26:45 paperwork and getting paid and
26:48 reimbursed and all that sort of stuff
26:50 it's going to help with all that so and
26:53 and you know you asked about medical but
26:55 this is one of the reasons I started the
26:58 channel and one of the reasons I'm so
26:59 passionate about this is I think there's
27:02 going to be that level of profound
27:05 impact in essentially every single
27:07 industry
27:08 and
27:10 a lot of it's going to be for the good
27:12 but almost all of it is going to be
27:14 incredibly disruptive to what's there
27:16 right now
27:18 right and so the only antidote that I
27:21 know of of is
27:23 start learning this stuff
27:25 um there's a guy named Paul rateser who
27:27 started the marketing AI Institute he's
27:30 he's worth a following LinkedIn and he's
27:31 got a podcast
27:33 and he he came up with the saying that
27:36 AI is not going to replace you but
27:38 someone using AI will and and I think
27:41 that's that's what's coming and I think
27:43 it's coming quite quickly
27:46 um and there was just a jobs report
27:48 there were like a hundred and
27:50 130 000 jobs lost and four like four
27:54 thousand of them were lost due to AI so
27:57 so we're just starting to see the needle
27:59 move on AI disrupting positions
28:04 is there a chat AI site where I do not
28:07 have to pay yes that's why I have these
28:08 up here so
28:10 chat.openai.com you don't have to pay
28:12 you have to sign up but you don't have
28:14 to pay you can use chat GPT 3.5 for free
28:18 if you want to use four and get access
28:20 to plugins you have to pay or you can
28:23 use bing.com so if you go to bing.com
28:26 it's like a it's kind of like a Google
28:28 page but if you click on the chat button
28:30 it'll open a chat interface that's gpt4
28:34 connected to the internet for free
28:37 bing.com so that's why that's there
28:40 and then poe.com has also got things you
28:42 can use for free but the the more
28:43 premium Services you have to pay for but
28:46 if if you want the sort of highest level
28:48 of chat GPT that is free bing.com is the
28:52 way to go
28:54 all right I still can't get over
28:58 stop calling me adorable I'm gonna start
29:01 blushing
29:03 we don't want that
29:07 neurolog neuralink is what the Elon Musk
29:11 is called
29:13 um hear me out Nano probes well yeah
29:17 have you seen go go look at videos on
29:20 how neural Link Works they have to they
29:22 had to invent a whole robot
29:24 that installs those things so they drill
29:27 like a pencil sized hole in your braid I
29:28 think it's up to a quarter now it was
29:30 supposed to be a like a pencil but now I
29:33 think it's up to a quarter
29:35 but they've got this robot that
29:38 um looks at your brain and it guides
29:40 these probes that are like a tenth the
29:42 width of a human hair
29:44 and it navigates them in between your
29:47 blood vessels in your brain matter in
29:49 your gray matter so it's it's literally
29:50 like like individually placing these
29:53 electrodes and there's 60 or 120 of
29:55 these micro wires so yeah so they're
29:58 literally doing Nano probes
30:01 um but you have to you have to opt into
30:03 doing that like they're not just like
30:04 kidnapping people off the street and
30:06 jamming electrodes in their brains
30:09 um I think it'll start as as medical
30:11 like like people have injuries and
30:12 things like that it'll probably start
30:14 there first for the trials I would think
30:17 um how about AI in schools and
30:18 universities yeah so it's it's going to
30:21 transform all of that
30:23 I think any University or school right
30:25 now that's trying to resist this is
30:28 being incredibly short-sighted and and
30:30 not serving their patients or their
30:32 their patients their students will
30:35 um
30:36 this stuff's going to be everywhere
30:37 Microsoft's rolling this into all of The
30:39 Office 365 Suite Google's rolling this
30:42 into all of
30:44 um the the Google office you know docs
30:47 and sheets and all that stuff so
30:49 probably within six months these tools
30:52 which right now are kind of separate and
30:54 living at these websites here are just
30:56 going to be ubiquitous they're going to
30:57 be built into all the tools that we use
30:59 so given that well how is a a university
31:03 or a high school supposed to respond
31:06 well
31:08 there's a new superpower that is
31:11 available to human beings and they're
31:13 going to use it and it's going to be
31:14 available everywhere
31:16 start engaging with it right so rather
31:18 than trying to block it or resist it or
31:21 or dismiss it and say oh it's not
31:24 perfect all the time
31:26 um roll it into your classroom there's a
31:29 there's a a high school teacher here on
31:31 Tick Tock that she's like a senior like
31:35 a 12th grade English teacher and she
31:37 went through the panic that everyone
31:39 goes through like holy [ __ ] is this
31:40 going to replace me and she said well if
31:42 it's going to be there let me try
31:43 rolling it into the classroom she brings
31:46 it into the classroom they she she has
31:48 it right papers and then they have
31:49 discussions about what did it get right
31:51 what did it get wrong and and she's
31:53 teaching them to think critically and
31:55 she's teaching them to understand that
31:56 it hallucinates and it gets [ __ ] wrong
31:58 and they gotta fix it and they got to
31:59 use their brains and they got to use
32:01 their brains and and so I think the real
32:04 opportunity here is we have an opera we
32:07 we are going to be forced into a
32:10 situation where how we've historically
32:12 taught students is no longer relevant
32:15 because of the power of these tools
32:18 but will be the opportunity to do is
32:20 engage with him on a different level and
32:22 increase their critical thinking skills
32:24 and get them to think strategically and
32:26 empathetically and to engage with other
32:28 human beings and I think that's the big
32:30 difference
32:31 so what does that look like I don't
32:34 [ __ ] know I know it's going to be
32:35 messy and there's going to be a bunch of
32:37 people that are pissed off and
32:40 resist it but it's coming whether we
32:42 want it or not it's capitalism though so
32:44 how soon do they sell neuro performance
32:47 enhancers
32:49 um
32:50 soon well I mean I mean listen the
32:53 iPhone is a neuro performance enhancer
32:56 and you know the the the goggles that
33:00 Apple just released are
33:04 they're a biological interface I mean
33:06 there's four cameras inside the goggles
33:09 pointing at your eyes measuring micro
33:12 responses in your in the dilation of
33:14 your irises
33:16 um so it's not a surgically implanted
33:18 thing but it's absolutely measuring your
33:21 emotional response to the last time you
33:23 clicked on that button
33:26 um he got excited so when he looks over
33:28 that way again chances are he's gonna
33:30 he's gonna click on that button so why
33:32 don't we start pre-loading that app so
33:33 he's got a better experience like that's
33:35 the machine learning they're putting
33:37 into that thing is absolutely
33:39 interacting with your
33:41 um biofeedback so so I think it's coming
33:44 but I I don't think it's going to be
33:47 like one day they they throw us in the
33:50 back of a van and strap you know
33:52 electrodes on our brains I think it's
33:54 going to be
33:56 just like we gave up all of our privacy
33:58 over two decades to Facebook and Google
34:01 and you know all the all the free things
34:03 you know all the free services and
34:06 social networks we used where we were
34:08 like this is awesome it's free
34:10 well you're the product right so over
34:13 two decades we gave up all of that so I
34:15 think it'll be a similar slow you know
34:18 progression
34:19 toward tighter interaction with these
34:22 machines
34:24 and and you know I think some people be
34:26 able to opt out but it's but it's like
34:31 I don't know why you wouldn't take
34:33 advantage like these tools are like the
34:35 most powerful tools we you know we've
34:38 ever had access to why would you not
34:39 want to engage with them in some way
34:42 um you don't have to get neural link but
34:44 I think you know ignoring them will put
34:46 you at a disadvantage in the next 20
34:49 years do we have to do something to
34:51 upload a photo to Mid Journey no you
34:54 just drag you literally drag it like
34:57 where you type in slash imagine you just
34:59 drag the photo there and hit return
35:01 and then you right click it to get the
35:03 hyperlink for it to get the URL
35:07 wait how did you get
35:09 Photoshop
35:10 in mid-journey so I just saved it out of
35:13 Photoshop so
35:14 I was in Photoshop
35:16 I saved that photo to my desktop and
35:19 then in mid-journey
35:21 so here's mid-journey let me go
35:26 let me scroll the bottom here
35:29 so there's mid-journey there's me on a
35:31 horse in a knight in shining armor and
35:35 then I went to find her and there's like
35:37 a photo of me and then I just drag it
35:39 from the finder into into mid journey
35:42 and hit return
35:46 can you use
35:49 dolly in Photoshop
35:52 well there might be a plug-in but I
35:56 don't think so uh it looks like there's
35:58 some new people here so let me just do a
36:00 quick intro
36:04 um my name is Kyle Shannon this is the
36:05 AI learning lab I'm just answering
36:07 people's questions they have about
36:09 generative AI I'm an entrepreneur I'm a
36:12 founder and CEO of a company called
36:13 storyvine
36:15 um it's a it's an 11 year old company
36:17 that does automated video storytelling
36:20 or automated video production
36:22 and um
36:24 and yeah I've got this Channel and I've
36:27 got an AI Salon where it's a community
36:29 of people hanging out talking about AI
36:31 I've got a newsletter the URLs that you
36:34 see here these are just there if you've
36:36 never tried chat GPT
36:39 um it's it's a requirement if you're
36:41 going to be here it's a requirement damn
36:43 it go to one of these so I'm not
36:44 affiliated with these in any way but the
36:46 top one is the official chat gbt website
36:48 chat.openai.com that's
36:51 um the official site for chat EBT and
36:54 then Bing and Poe are two other options
36:56 they all have free options you don't
36:58 have to pay for anything you can you
37:00 don't have to
37:01 um
37:02 but just go start playing with those
37:03 tools and and then you know feel free to
37:07 hang out here I'm talking about this
37:08 stuff answering questions if you have
37:09 questions drop them in
37:12 um
37:14 I don't think you can use dolly in
37:16 Photoshop so so they've got their own so
37:19 Adobe created this thing called Adobe
37:21 Firefly and they trained that model on
37:25 public domain open source and Adobe
37:29 stock images that they own all the
37:31 rights to
37:33 um so they've got their own proprietary
37:34 image generation model and then their
37:37 their own proprietary tools so there's
37:39 there's
37:40 um Adobe Firefly is their generative AI
37:43 kind of
37:44 tool suite and then they rolled a bunch
37:47 of those features into Adobe Express
37:50 which is kind of like a canva competitor
37:52 and then they also rolled it into what
37:55 they're calling generative fill so when
37:57 you're in Photoshop you can have you
38:00 know like a vertical photo and extend
38:01 the edges out and and fill those edges
38:04 and it'll just generate it based on the
38:06 image it's really powerful stuff so it's
38:09 it's not Dolly but it's like Dali but
38:12 it's better
38:14 um Can AI be useful in casinos or for
38:16 gambling sure and and you know at some
38:19 point
38:20 um
38:20 the you're gonna be able to wear like
38:23 you know one of these little earbuds and
38:25 there'll be enough technology in here
38:27 that it will just dynamically like
38:29 listen to what's going on on the table
38:31 or maybe there's a little camera here
38:32 that looks down at your cards and
38:34 they'll be giving you advice fold fold
38:36 you know Double Down raise
38:41 um you know
38:43 so yeah like yeah it again like I said
38:47 these tools a good way to think about
38:50 just at a super macro level what these
38:53 tools make possible is
38:56 they take very very complicated things
38:58 and simplify them
39:01 right so
39:03 like
39:04 um like
39:06 creating a screenplay generator if you
39:09 were to write a screenplay generator
39:11 with traditional programming you'd have
39:14 to program it you know here's how many
39:18 pixels you indent this thing and you'd
39:21 have to you'd have to program in all the
39:22 logic rules and then you'd have to
39:24 program in story rules
39:28 what these things do is that you just
39:30 say I want a screenplay and it kind of
39:32 goes look looks at its Corpus of
39:34 knowledge which is essentially all of
39:36 the internet
39:38 that's what they trained it on and it
39:41 goes oh you want to screenplay here's
39:43 the screenplay oh you want an outline
39:44 here's an outline you want that to be a
39:47 comedy now it's a comedy oh now it's a
39:49 murder mystery
39:50 um so you can literally Point these
39:53 things at any industry any creative
39:57 industry any knowledge industry almost
39:59 any industry and and find dramatic ways
40:03 to
40:05 um either improve processes make them
40:07 more simple
40:08 predict things all sorts of crazy stuff
40:11 so yes it's coming can you do a whole
40:14 workshop on what you just showed us yeah
40:16 that's actually not a bad idea
40:19 um
40:20 yeah I I'm trying to I'm trying to
40:22 figure out the workshop thing I don't
40:24 quite know how to do it what I want to
40:26 do yet
40:28 and so I'm I'm I'm I'm cooking it I'm
40:32 cooking it the nice thing about these
40:34 lives is I actually don't have to plan
40:35 anything the thing about workshops is I
40:38 gotta like if I'm gonna put together a
40:40 workshop I want to do it right
40:42 so I want to just make sure that I'm
40:44 doing that is AI useful in casinos for
40:46 gambling we talked about that I need
40:48 Discord to use mid-journey yes and so
40:51 here's the deal
40:52 mid-journey in my opinion is by far the
40:55 the highest quality of the image
40:56 generation tools but you have to you
40:58 have to generate things within Discord
41:01 now your next question might be isn't
41:04 that [ __ ] stupid
41:05 yes but that's where they started that's
41:08 where it is they probably came out I
41:10 don't know they're probably Gamers or
41:12 something like oh let's just do it in
41:13 Discord
41:18 um
41:19 it's on there it's on their roadmap to
41:22 build an actual interface so you don't
41:23 have to go to stupid [ __ ] Discord to
41:26 make images
41:27 um it's just not there yet
41:29 so I wish I had a better answer did you
41:32 see stable diffusion control net QR code
41:34 examples those are so cool yeah so we're
41:37 gonna go to the Twitter
41:47 so stable diffusion is a thing that's
41:50 like mid-journey creates images
41:53 um you can probably just do QR codes and
41:56 I'm sure it will show up because it's so
41:58 bitching
41:59 all right let's get the
42:02 and then within stable diffusion there's
42:04 this thing called control net where you
42:07 can upload an image and then you can
42:09 iterate on that image
42:11 um where it just changes like the style
42:13 of it but it keeps the main part of the
42:15 image active
42:17 and so what they've figured out how to
42:18 do
42:20 is use a QR code as the control image
42:24 and then
42:25 like there's a crossword puzzle one the
42:29 the QR code still continues to function
42:33 but they
42:34 you know can turn it into all of these
42:36 different things like check that one out
42:38 I mean that was just wild it's like this
42:40 3D paper thing with a little like
42:42 origami flowers or paper flowers
42:46 there's a crossword puzzle with with a
42:48 woman you know kind of sitting in front
42:50 of it look at that one that's a
42:52 functional QR code I mean
42:55 come on people
42:57 tell me this stuff isn't just crazy
43:01 I'll tell you crazy did you see those QR
43:05 codes that was so cool
43:07 I didn't know AI could do that
43:10 I didn't really realize so and that's
43:13 so I'm being a bit of a pain in a bit of
43:16 a smart ass here because that's just who
43:17 I am but
43:19 um
43:23 if you're looking at this stuff from the
43:24 outside you're like okay make some
43:26 pictures or oh it's like a Google search
43:28 with fancier
43:30 it's it's very hard to describe like how
43:33 powerful it is without
43:36 um
43:37 just people who use chat GPT and kind of
43:41 understand what it is
43:42 um
43:44 um it's hard to describe it to people
43:47 because I don't know so anyway so let me
43:49 just
43:51 um
43:52 let's start with so so we'll start with
43:54 what what I like to call parlor tricks
43:56 so we'll do write me a country song
44:01 about a
44:05 um
44:06 let's see about a blue jay
44:12 that
44:14 um pooped on my truck
44:17 okay we'll start with that
44:19 and so notice how quickly it does it it
44:22 knows what verses are it knows what
44:24 choruses are let's go back like I mean I
44:27 didn't give it any context whatsoever I
44:29 didn't tell it it was a country singer I
44:31 was driving down Industry Road just
44:32 trying to ease my mind with wind in my
44:34 hair and a song on the radio everything
44:36 felt so fine but out of the blue there
44:38 came a sight that gave my heart a start
44:40 a blue jay flying overhead aiming right
44:42 for my truck oh my oh my right and so
44:46 so so you immediately go okay well maybe
44:50 that was a fluke that you know that that
44:52 was that was kind of cool but whatever
44:53 like so maybe so it can rhyme and then
44:56 you could do things like
44:58 um make a table that compares the major
45:03 milestones
45:07 um and dates of the top three
45:15 I don't know tech companies
45:21 in their first 10 years include
45:27 milestones
45:30 that um
45:33 were key to their future success
45:40 I don't even know if it knows that but
45:44 there's a table
45:47 of Apple Google and Microsoft when they
45:51 were founded release of the first
45:52 product introduction right so so like
45:56 without even thinking about it
45:58 you know I did that like that that's
46:00 like not that's like not even beginning
46:02 to scratch the surface I'll do one more
46:04 um right
46:05 me a
46:08 um
46:09 C sharp
46:13 um
46:14 clone of the Flappy
46:20 Birds
46:22 app I can run in unity
46:31 sure here's the c-sharp script that
46:34 replicates the basic functionality
46:37 of a Flappy Bird game in unity
46:40 why I know this one works is I had a 10
46:43 year old I was teaching them how to do
46:46 generative images and he goes
46:48 I wrote I wrote a flappy birds clone in
46:51 C sharp and he goes but there's a bug
46:53 and I said well let's go I said did you
46:56 try Chachi but he's like oh yeah I use
46:57 Chachi VT I'm like did you and then I
47:00 showed him I typed this in and he
47:03 watched it write this code he's like 10.
47:06 his his mouth literally dropped open he
47:08 went like this
47:10 and then he said he said this changes
47:13 everything
47:15 that's a 10 year old
47:17 so so if you have not played with chat
47:19 gbt that's why this channel exists
47:22 because I want as many people as
47:23 possible to kind of get their head
47:25 around
47:26 oh something's different something has
47:29 changed
47:31 um
47:32 I was um in the in the mid 90s I started
47:36 one of the first digital agencies a
47:38 company called agency.com
47:40 and I had an epiphany about the web in
47:42 1994.
47:44 where I published an online magazine and
47:46 uploaded some files to a server in
47:48 California and three weeks later there
47:50 was a full-page article of this magazine
47:53 that I had created in Paris
47:56 three weeks after I uploaded it
47:59 somewhere on the internet
48:01 and it was one of those moments where I
48:04 was like holy [ __ ] the world is
48:05 different like the world has changed but
48:07 no one knows it yet like this technology
48:09 changes everything
48:11 this AI stuff's exactly the same thing
48:13 where everything's different it's just
48:15 no one knows it yet the the one of the
48:18 one of the well two primary differences
48:20 from the World Wide Web
48:22 one of them is the speed so the World
48:25 Wide Web it took six years to get to 100
48:27 million users chat GPT six weeks to 100
48:32 million users was the fastest adoption
48:34 of technology and history so that's one
48:37 primary difference is speed and the
48:39 second one is what the tools do like the
48:42 World Wide Web while it was
48:43 revolutionary all it really did was it
48:46 made it so you could click on a
48:47 hyperlink like that was all it did it's
48:50 like here's a page you click on a link
48:52 and it takes you to another page
48:54 this stuff is like you know writing
48:57 tables writing code writing screenplays
48:59 writing novels writing blog posts
49:02 creating social media strategies in 30
49:05 seconds so
49:07 so that's what's coming that's why this
49:09 Channel's here and so that's why start
49:12 twerking you don't want me to twerk
49:13 that's that's something none of us will
49:15 ever forget
49:16 I will I will sustain an injury and none
49:20 of you can unsee that so so I think
49:23 we're gonna I think we're gonna we're
49:25 gonna we're gonna hold off on the
49:27 twerking until uh until there's a real
49:29 demand
49:30 don't make their demand don't do a
49:33 write-in campaign
49:35 all right
49:43 Mercy Mercy please no twerking okay good
49:46 so fine finally thank you Blake thank
49:49 you Blake a voice of reason get to get
49:52 to twerking what what is wrong with you
49:54 people
49:55 all right Emilio's wife oh okay you
49:58 don't want twerking either okay good
50:00 good good no twerking all right so right
50:03 now we're three to one for for start
50:05 twerking and and don't twerk
50:08 um
50:09 all right
50:11 um oh it looks like we got some new
50:12 folks here my name is Kyle Shannon this
50:13 is the AI learning lab pop any questions
50:15 you have about AI down below
50:18 um I'll I'll sort of do that again if
50:20 you haven't played with chat GPT go to
50:21 that top URL chat.openai.com start
50:24 playing right now do it you won't regret
50:26 it
50:28 I wish our professional development was
50:30 geared to learning AI so what's your
50:32 name
50:33 um
50:34 look
50:36 I can't see it Loki oh Loki 20.
50:39 um
50:41 so what I would say is this be the
50:44 change so if you wish your professional
50:46 development was geared toward a toward
50:47 learning AI start learning AI if you're
50:50 not now
50:51 and whoever runs that development group
50:56 approach them and say hey I'm learning
50:58 this would you mind if I did a free
51:00 workshop for anyone who wanted to learn
51:01 about it and then like like volunteer to
51:04 do a workshop first and then figure out
51:07 what you're gonna do
51:09 be the change because that
51:13 that's what it's going to take in every
51:15 single organization because
51:17 because we have we've had this this is
51:19 my theory on this thing because we've
51:21 had 50 years of Hollywood telling us
51:24 that the robots are going to kill us
51:26 there's a default Society societal
51:30 response to the term AI of fear
51:35 and then you know what comes with fear
51:38 is dismissal and you know rejection and
51:40 all that sort of stuff so people are
51:42 like I don't want to deal with it it's
51:44 going to kill us yeah I like real people
51:47 real people are smarter than robots
51:50 all of that may be true but this
51:52 technology is still coming and it's
51:54 coming really fast so get curious go
51:58 there start playing
52:00 um and and you know in your organization
52:02 I I would seriously like tomorrow's
52:04 Monday go to the person that heads up
52:06 development and say hey I really want to
52:09 learn about this stuff I saw this crazy
52:11 old man on Tick Tock last night he told
52:14 me to come talk to you
52:16 um would you mind if I put together a
52:18 workshop and if they say no then I don't
52:21 know
52:22 quit
52:24 no don't don't quit all right is
52:26 mid-journey free mid-journey is not free
52:28 which is a drag but the cheapest tier of
52:30 mid journey is like 10 bucks that's what
52:32 I pay
52:34 um you can pay up to 60.
52:37 Dolly can be used in Photoshop via
52:40 stability plug-in okay cool I didn't
52:42 know it was Dolly I thought it was
52:44 stable diffusion but whatever cool
52:45 that's awesome
52:47 how can AI help Communications between
52:50 two persons in different languages holy
52:52 [ __ ] the the language thing that is
52:57 about to be Unleashed on us so Nvidia
53:00 created a technology that does this and
53:02 I I just saw an open source project that
53:04 does this so so for one thing
53:06 chat CPT just understands all the
53:09 languages so if you've got someone in
53:11 Japan that you're interacting with and
53:13 their English is bad and they don't
53:14 understand you
53:16 um take what they wrote in Japanese
53:17 paste it into chat EBT read it respond
53:21 to it and then say translate my response
53:23 back to Japanese like it's amazing at
53:25 all the languages
53:27 so that's one thing but what's coming is
53:30 there's technology now that is
53:33 synthesizing your voice mapping your
53:36 face including your lips
53:39 and translating in real time so you're
53:42 going to be able to get on there's if
53:44 you go look up nvidia's I forget what
53:47 it's called
53:48 we covered it in one of the newsletters
53:50 I forgot it's called but nvidia's got
53:53 this technology where they've got four
53:55 different people speaking four different
53:57 languages you're basically going to be
53:58 able to I'm going to be able to speak
53:59 English with
54:01 foreign Nationals speaking their native
54:04 language we will all hear each other in
54:06 our own native language but it will
54:08 sound like that person and it'll be
54:10 saying what they say like that's coming
54:12 very very quickly
54:15 um so yeah
54:16 ask chat EBT I don't know what the
54:18 question was I couldn't answer will
54:19 there be any workshops wait I'll be at
54:22 any workshops you run it doesn't need to
54:23 be fancy oh thank you I appreciate that
54:26 it looks like something we should really
54:28 study
54:29 oh AI looks like something we should
54:31 study yeah and I I wouldn't
54:34 like study study feels
54:36 a little too intense well maybe not I
54:39 don't know depends on your relationship
54:41 to the word study study to me sounds
54:42 like work I have adds
54:45 I what I would say is get curious and
54:48 get and start exploring start exploring
54:51 um
54:52 I would sort of check your expectations
54:53 at the door of what you think it can do
54:55 or what you think you can do like one of
54:58 my big epiphanies like I'm not great at
55:00 Art right and this image I started using
55:03 stable diffusion with this thing called
55:05 dream Booth where I trained it up on my
55:06 images so that's kind of like my chin
55:08 and nose and but it's got this crazy
55:11 eyeball thing right
55:13 when that image was generated I got I
55:17 got
55:18 struck with
55:20 like I had an emotional response to it
55:23 which was
55:24 that's the kind of image that if I had
55:26 the talent and Photoshop skills and
55:29 illustration skills to like cook
55:31 something up that was kind of mystical
55:33 and weird like that would be the style
55:35 that I would do it in but I don't have
55:37 those skills like I can do Photoshop and
55:39 I can do design but when it comes to
55:41 like true creative expression with
55:43 images I'm I'm just not good
55:46 and when I created that I thought oh
55:48 [ __ ] now I can be good like like these
55:52 tools are like amplifiers like you can
55:54 just have a very thin idea and you put
55:56 it into chat GPT and it becomes fully
55:58 realized within five minutes and you're
56:00 like oh
56:02 well maybe I should have another idea
56:03 and then I should have another idea and
56:05 so so it just keeps building like that
56:07 so yeah I think I think this this stuff
56:09 is stuff that you should get curious
56:11 about and explore and figure out how it
56:15 can help you do what you want to do even
56:17 if it's something that historically
56:19 you've told yourself well but I can't
56:21 draw like I'd be an artist but I suck at
56:24 drawing well maybe now you could be an
56:27 artist and maybe you could do a lot of
56:29 other things right
56:31 when will AI be smarter than humans
56:34 um gpt4
56:37 definitely has more knowledge than any
56:40 probably any human on earth I mean maybe
56:42 a savant or two
56:45 um
56:46 now it doesn't reason and it doesn't
56:49 have empathy
56:51 but it
56:53 um
56:54 it can solve problems like its reasoning
56:57 and it can speak in empathetic language
56:59 so so right now it's kind of right now
57:01 the AI is play acting the human parts
57:05 but the knowledge Parts it's already way
57:08 smarter so it depends on your definition
57:09 of smarter right like I would argue that
57:12 gpt4 is way smarter than me right now
57:15 because it knows differential calculus
57:18 and I don't but if I want to use
57:20 differential calculus I can just go to
57:22 chat EPT and go teach me how to do this
57:23 or just do it for me right
57:27 um so there's a there's a definition of
57:30 smarter than that it is right now I I
57:33 assume what you're going to is what
57:35 they're now calling AGI which is
57:37 artificial general intelligence right
57:39 which is when will the machine be you
57:41 know smarter than any single human being
57:44 on Earth including all of the reasoning
57:46 capability empathy stuff
57:49 um I don't know probably
57:51 five years
57:53 and then there's this thing called
57:55 um artificial super intelligence which
57:57 is where the machines are smarter than
57:59 all of the humans put together
58:02 that's the one that scares everyone I
58:04 think that one's probably more like 20
58:06 uh 20 years out
58:09 um
58:12 I I I thank you thank you Legend David
58:16 Sharp
58:17 um I do have artificial intelligence
58:20 it's just called it's just 80d
58:22 um do you have a YouTube channel I do
58:24 not have a YouTube channel all right
58:25 you're blowing my mind I'm not sure what
58:27 I did but awesome do you do Shakespeare
58:29 in the park
58:30 um
58:32 I didn't do Shakespeare in the park I'm
58:34 trying to think when the last thing what
58:36 the last thing I did in New York was
58:38 I did a musical called The Amazing
58:40 Adventures of tense guy written by Alan
58:43 Ball who won the Academy Award for
58:44 American Beauty that was cool
58:47 um
58:49 no I haven't done Shakespeare in the
58:50 park
58:52 but I'd like to that'd be cool
58:56 um Kyle's a smart ass
58:59 uh
59:02 how do I auto GPT
59:05 um Auto GPT and those autonomous agents
59:07 you've got to go to GitHub so so there's
59:10 a cool thing you can do with chat GPT
59:12 if you don't know GitHub which GitHub is
59:16 so what GitHub is is it's a code
59:18 repository where all the Geeks in the
59:20 world you know dump their code and they
59:22 make projects and they can help each
59:24 other and make open source projects
59:27 it's really complicated there's lots of
59:29 documents everywhere if you don't know
59:31 coding and engineering and GitHub it's
59:35 really really confusing but
59:38 um
59:40 go to go to either Twitter or go to
59:43 Reddit and search for auto GPT and it
59:46 will point you to probably the latest or
59:48 the official Auto GPT GitHub repository
59:51 and then one of the things that you can
59:53 do
59:54 this is a trick that I learned on Tick
59:56 Tock actually
59:57 you can go to the readme file so so one
1:00:00 of the tricks of GitHub is the first
1:00:01 thing you're supposed to do when you go
1:00:02 to GitHub is you read the readme file
1:00:04 because the readme file tells you what
1:00:06 all the components are and how to
1:00:08 install them and how to do all the [ __ ]
1:00:10 right but what you can do is you can
1:00:12 open that readme file copy sort of from
1:00:15 the table of contents down so so copy
1:00:18 the meat of the readme file paste it
1:00:20 into chat GPT and then say I'm on OS X
1:00:25 you know whatever version can you create
1:00:28 a step-by-step tutorial for me to
1:00:30 install auto GPT and and do not move on
1:00:34 to the next step until I say the word
1:00:36 okay
1:00:38 and then it will say Okay step one do
1:00:40 this and then it it very often will make
1:00:43 presumptions like you know what the [ __ ]
1:00:45 it's talking about it'll be like well
1:00:46 just you know
1:00:48 um you know install this library and and
1:00:51 then if you don't know what that means
1:00:52 go
1:00:53 I don't understand what that library is
1:00:56 and then it'll tell you and then you say
1:00:57 okay can you break down how to install
1:01:00 that Library it'll be like oh sure and
1:01:02 it'll tell you and and then just you
1:01:04 just keep going through and then you say
1:01:06 okay and it'll go to the next step so
1:01:08 you have to be a bit geeky but that's
1:01:10 that's how I would do it
1:01:13 this changes everything can you teach me
1:01:15 how to make Auto GPT I can't because
1:01:18 I've never really installed it
1:01:20 um there's there's a thing called GPT
1:01:21 for all which is a local version
1:01:25 um
1:01:26 probably a step before Auto GPT is is go
1:01:29 uh watch some YouTube videos on Lang
1:01:32 chain because Lang chain is one of the
1:01:34 components of Auto GPT and Lang Chain's
1:01:36 a thing that is is really cool it's like
1:01:38 you can you can basically embed your own
1:01:41 documents and then treat them like a
1:01:44 chat gbt experience
1:01:45 um and you create a little Vector
1:01:47 database there's all sorts of cool [ __ ]
1:01:49 in it so I'd start with Lang chain and
1:01:50 then go to to chat GPT or to then go to
1:01:53 auto GPT
1:01:55 how has tiktok contributed to the hang
1:01:59 on
1:01:59 tick tock tick tock as I was trying to
1:02:02 read that question jumped me to the
1:02:04 bottom of the of the comments
1:02:06 oh we got some new people here hey
1:02:08 everybody my name is Kyle Shannon this
1:02:09 is the AI learning lab
1:02:11 um I'm here talking about AI stuff uh if
1:02:14 if you're if your first question is well
1:02:16 what the [ __ ] do you know about AI
1:02:19 um I'm an entrepreneur first I I am not
1:02:21 a machine learning AI specialist I'm
1:02:24 more like one of those guys that now
1:02:26 that it's here and available for the
1:02:27 rest of us trying to figure out how to
1:02:29 use it to disrupt my own business
1:02:32 um try to figure out what the future of
1:02:34 business is how it's going to impact
1:02:35 work and I think it's quite quite
1:02:37 profound so anyway that's who I am
1:02:39 quickly
1:02:40 um if you haven't used chat GPT before
1:02:42 these URLs are just there's kind of
1:02:45 public service announcements the top one
1:02:47 that's the official chat GPT website
1:02:50 chat.openai.com go there start playing
1:02:53 while you listen to me ramble or if you
1:02:55 want to just twirl on tour along
1:02:58 um bing.com is Microsoft's uh uh search
1:03:02 engine but there's a chat button there
1:03:04 that chat button is now gpt4 for free
1:03:06 and then pro.com allows you to play with
1:03:09 some other stuff and the other two are
1:03:10 just resources that are both worth uh
1:03:12 just checking out
1:03:14 um so I'm trying to find where it was in
1:03:17 the questions
1:03:19 how has Tick Tock contributed to
1:03:22 existing AI
1:03:24 well there's a ton of AI like
1:03:28 there's a ton of AI in Tick Tock right
1:03:31 well I'm sure there's a ton of a listen
1:03:33 there's a i a has been around for
1:03:34 decades right so there's there's a lot
1:03:36 of AI in in the algorithms and things
1:03:39 like that but almost all of the filters
1:03:41 that that you're seeing you know
1:03:42 especially the ones the new more modern
1:03:44 ones that are really realistic and
1:03:46 what's that one called
1:03:48 um I don't know style or whatever it
1:03:50 doesn't [ __ ] matter but the the the
1:03:51 you know the filters that make people
1:03:53 look prettier or whatever the ones that
1:03:55 are really sophisticated that's all AI
1:03:56 they're 3D mapping faces and they're
1:03:59 doing all this [ __ ] dynamically that's
1:04:00 all AI
1:04:02 um so I think it's contributed in that
1:04:04 way and then I think the other way it's
1:04:06 contributed in is in what this channel
1:04:08 represents and what Rachel Woods channel
1:04:11 the AI exchange represents and McKinsey
1:04:13 Bose that's talking about the app he's
1:04:15 building called GPT boss is that there
1:04:19 are people who are are discovering these
1:04:22 tools kind of in real time and sharing
1:04:25 what they're learning and and there's a
1:04:27 bunch of scams scammy kind of [ __ ]
1:04:30 tick tocks that are like you can make
1:04:31 money
1:04:33 ignore those
1:04:35 but find the people that are like okay
1:04:37 here's what happened this week here's
1:04:40 what I think's a big deal
1:04:42 um I tried to do this and it worked I
1:04:45 tried to do this and it broke like find
1:04:47 people that are you know sort of getting
1:04:49 Hands-On with this stuff
1:04:51 and just start learning just start like
1:04:53 through as much osmosis like I should
1:04:56 I've shifted my behavior a lot like I I
1:04:59 tend not to watch TV anymore
1:05:02 um I'll either do like these these lives
1:05:04 or I'll go on to YouTube or Tick Tock
1:05:06 and I'll just kind of scroll and like
1:05:08 before bed I'll just watch you know
1:05:11 someone like like there's a guy on
1:05:13 YouTube named David Shapiro that's doing
1:05:15 this whole AGI platform it's like it's
1:05:19 just this amazing thing he's written
1:05:20 five books on machine learning and Ai
1:05:22 and it's like you know whenever I feel
1:05:24 like I got my [ __ ] together I'm like I'm
1:05:26 really smart on this stuff I go watch
1:05:28 David Shapiro and I'm like oh I'm a
1:05:30 [ __ ] [ __ ] like he like he's he's
1:05:31 just at a whole nother level right
1:05:33 really smart really articulate
1:05:35 um so just find people that you resonate
1:05:37 with and start learning that way um but
1:05:39 I so I think it's contributed a lot
1:05:40 there as well but it's it's a giant
1:05:42 playground for AI technology
1:05:45 as well and and I'm pretty sure that
1:05:48 they've already got a chat GPT like
1:05:51 feature in Tick Tock for some users so
1:05:54 they're testing it right now
1:05:56 what chat GPT
1:05:59 would I suggest what which chat CPT for
1:06:01 so for me for my money
1:06:05 when it comes to spending my hard-earned
1:06:08 money
1:06:08 I like to go to chat.openai.com and I
1:06:12 pay their twenty dollars twenty dollars
1:06:14 each and every month for them to
1:06:17 so if you pay the 20 bucks to open AI
1:06:21 you get access to gpt4
1:06:25 and it's kind of like like I like to
1:06:27 think owners really come along yet
1:06:29 that's got anything better than gpt4 so
1:06:32 I I pay for it to have access to it so
1:06:34 when you pay for it you can also enable
1:06:37 gpt4 connected to the internet and you
1:06:39 can also enable this thing called
1:06:41 plugins which is a whole
1:06:43 pile of different sort of augmentations
1:06:46 to gpd4 like you know TripAdvisor or
1:06:51 um video insights where you can have it
1:06:54 you can point it to a YouTube video and
1:06:56 and you can ask it questions about that
1:06:57 YouTube video so so there's all sorts of
1:07:01 reasons to do it so
1:07:03 so for my money I would I would just
1:07:06 stick with stick with the original
1:07:08 um
1:07:09 wait hang on
1:07:11 which is chat.openai.com now if you
1:07:14 don't want to spend the 20 bucks
1:07:15 bing.com
1:07:17 the if you go to bing.com and click on
1:07:19 the chat button that's gpt4 that's it's
1:07:24 from open AI it's gpt4 they've got it
1:07:27 tweaked a little bit so I don't think
1:07:28 it's quite as good as the the original
1:07:30 but it's connected to the internet it's
1:07:32 gpt4 and it's free and then po.com
1:07:36 has a couple of things in it that are
1:07:38 potentially worth paying for but it's
1:07:40 just there you get to play with a lot of
1:07:41 different models and and compare them
1:07:43 but if you're just getting started for
1:07:45 me I just go to chat.openai.com you can
1:07:48 even just play with the free version for
1:07:49 I don't know the next month and still
1:07:52 have your mind blown on a daily basis
1:07:56 do I need a DS background
1:07:59 for AI digital science
1:08:02 I don't know what DS background is
1:08:04 if you tell me what DS background is but
1:08:06 I don't think you do so it depends
1:08:08 if you if you want if you want to be
1:08:13 like working at open AI generating the
1:08:16 models then you probably need some
1:08:18 [ __ ] Tech chops but if you want to
1:08:20 use these tools in your business then
1:08:22 you just need to be a person curious
1:08:24 enough to start learning the tools I
1:08:26 wonder if people get stumped just like
1:08:28 you don't know anybody's phone number
1:08:30 because it's in contacts oh yeah well I
1:08:33 don't know
1:08:40 I like I I feel like
1:08:43 in using these tools I'm getting smarter
1:08:45 I'm definitely getting more creative
1:08:48 like I'm more creatively engaged there's
1:08:50 something about the creative process for
1:08:52 me
1:08:53 that when I have an idea before chat EPT
1:08:57 and before mid-journey
1:08:59 if I would have an idea I'd be like oh I
1:09:02 want to pursue that idea and then I'd be
1:09:04 like well
1:09:05 but if I'm gonna pursue that idea I
1:09:07 really need a tool that can do flow
1:09:09 charting so let me go find a tool for
1:09:11 flow charting and then like 45 minutes
1:09:13 later I'd be like down some [ __ ] app
1:09:15 rabbit hole and like none of the apps
1:09:18 worked and then I wouldn't even remember
1:09:19 my idea like I just felt like the
1:09:21 creative process was a series of
1:09:23 roadblocks that the idea had to be good
1:09:25 enough to get over
1:09:26 and with with chat gbt and [ __ ] with
1:09:29 these tools I can have an idea and then
1:09:32 there's no roadblocks it's like I have
1:09:34 this idea and I'm like and I go just
1:09:37 bring it to life and I'm like oh yeah
1:09:40 that's good and then I keep going and
1:09:42 so
1:09:43 um
1:09:44 I don't think people are gonna get
1:09:45 stupid or I think they're going to get
1:09:47 more engaged I could be wrong though
1:09:50 I I mean I just I feel like I feel like
1:09:53 these tools are
1:09:55 like the emotional the emotional feeling
1:09:57 I have is like Christmas morning like
1:10:00 daily like we get to play with this this
1:10:03 is not cool oh my God Mom you got me
1:10:07 that oh that's so cool it's like
1:10:11 and that's the face I make every single
1:10:13 day that's why I have no friends
1:10:15 I will get rid of it will get rid of
1:10:17 corruption and politics well it's it's
1:10:20 certainly gonna it's certainly gonna
1:10:23 generate some uh some conversations
1:10:27 um it very well uh Jesus
1:10:31 absolute power corrupts absolutely isn't
1:10:33 that how it goes
1:10:35 um
1:10:37 I think humans will always find ways to
1:10:40 be selfish and and fight for power
1:10:45 but I think that I think that AI
1:10:48 will make it easier to hold politicians
1:10:51 accountable but I would not
1:10:53 underestimate their ability to snake out
1:10:55 from under that accountability sorry
1:10:58 what's mid-journey so mid-journey
1:11:00 is
1:11:01 um so so chat GPT is sort of magic with
1:11:05 words let's just call it that you know
1:11:07 words are code mid-journey is Magic with
1:11:10 images and so there's there's a there's
1:11:12 a number of them out there there's
1:11:14 um mid-journey stable diffusion dolly
1:11:17 and now there's Adobe Firefly so there's
1:11:20 there's like four or five of them and
1:11:22 mid-journey right now you have to
1:11:26 um kind of like slack if you've ever
1:11:28 heard of it which is basically like
1:11:30 um each of these little circles think of
1:11:33 it like a different community and then
1:11:35 each of these little channels is our
1:11:37 conversations within that that one
1:11:39 community that I've selected that this
1:11:41 is the AI Salon that that I um host and
1:11:44 manage and so these are little
1:11:46 conversations and then we have a
1:11:47 mid-journey channel in here
1:11:49 and so mid Journey let me just scroll
1:11:51 back through here you know you you give
1:11:53 it a little prompt like oh I actually
1:11:56 just typed in garbledygook there and it
1:11:58 generated that floating island
1:12:01 um and then here was one where I took
1:12:03 this image that was created in stable
1:12:05 diffusion and then I created a
1:12:08 description from it and it generated
1:12:10 these and so it's it's a text to image
1:12:13 platform so basically you tell it what
1:12:16 you want like this one was really cool
1:12:18 this was
1:12:19 um
1:12:20 like create some mid-century modern
1:12:22 furniture based on a Star Wars theme
1:12:26 and it generated these super cool
1:12:29 mid-century modern chairs aren't those
1:12:32 groovy
1:12:35 and like the the final one we blew up
1:12:37 like this is like I I would own this
1:12:40 thing
1:12:42 I mean that's just that's just cool
1:12:45 and I mean it takes you know like 30
1:12:48 seconds or a minute you describe
1:12:51 something and it generates [ __ ] like
1:12:53 that so that's what mid-journey is it's
1:12:56 I I mean this this stuff is just it just
1:13:01 like melts your face it's I don't know
1:13:06 insane it's insane we're living in a
1:13:09 miraculous time and not a lot of people
1:13:11 know it yet
1:13:13 Nvidia h100s yep they just Nvidia just
1:13:17 hit a trillion dollar market cap one
1:13:19 trillion dollars best AI tools for
1:13:21 running Google paid ads
1:13:24 oh good lord
1:13:27 um I don't know I would go to
1:13:28 futurepedia.io I think the whole SEO
1:13:31 world is about to get
1:13:33 um
1:13:34 massively disrupted
1:13:38 so
1:13:40 and and I don't know a lot about that
1:13:41 world anyway so but I'm sure there's
1:13:43 tools at
1:13:45 futurepedia.io to deal with it
1:13:49 um
1:13:50 when I log into chat it requires my
1:13:52 telephone number to send code what else
1:13:54 do they require I think that's the only
1:13:56 thing they require is your phone number
1:13:58 and I know that's annoying especially
1:14:00 for people that are in the blockchain
1:14:02 and nft World nobody wants to be doxxed
1:14:04 but if you if that freaks you out then
1:14:06 go to Bing bing.com I don't think you
1:14:08 need an account at bing.com
1:14:14 can you do a quick rundown of using mid
1:14:16 Journey sure I'll do a quick I'll do a
1:14:18 quick mid Journey for you I just did a
1:14:20 little bit of it there but let me show
1:14:21 you how you actually do it I'll show you
1:14:23 two cool things Gen X or so I don't know
1:14:26 how to not use those okay
1:14:28 I'm sure I just triggered someone
1:14:31 a bird lover
1:14:33 um
1:14:35 a stone salesman
1:14:37 um where is it gonna go
1:14:38 okay we're gonna go over to Poe
1:14:41 poe.com okay so Poe is just a chat bot
1:14:46 kind of like um chat GPT right where you
1:14:49 put in things at the bottom and it
1:14:51 generates text at the top now down the
1:14:54 left hand side you have you have a bunch
1:14:56 of different models here you have the
1:14:57 gpts you have the clods and then below
1:15:00 it you've got
1:15:02 um Bots like I am Sling Blade I created
1:15:04 a sling blade bot I created a Yoda bot
1:15:06 these two Bots here are mid Journey Bots
1:15:09 and what they do so in order to prompt
1:15:12 mid-journey to give you the best quality
1:15:15 images there's all sorts of parameters
1:15:17 you can add like it understands specific
1:15:20 artists and photographers it understands
1:15:23 things like photographic terms like wide
1:15:25 angle or close-up or macro or depth of
1:15:29 field it understands
1:15:31 um uh like art media like uh you know
1:15:35 pastels oils clay sculpture things like
1:15:38 that so what these Bots do is take a
1:15:41 shitty prompt and make it better so
1:15:43 there was a guy on last night that
1:15:45 created this one called MJ Captain so
1:15:47 mid Journey Captain so what I'm gonna
1:15:50 I'm gonna sweep hit the little sweep
1:15:52 button so it starts this chat over so
1:15:54 I'm going to say
1:15:56 um what are we gonna do we're gonna do
1:16:00 we're gonna do a pudgy
1:16:03 black and white
1:16:06 dog with a red cape
1:16:10 standing
1:16:15 on a cliff
1:16:19 overlooking a city
1:16:23 at Sunset so that's not bad that's
1:16:25 decently visual so I'm going to hit
1:16:27 return
1:16:28 and then what it created was a digital
1:16:30 painting of a chubby black and white dog
1:16:32 with a red cape standing on a cliff over
1:16:34 overlooking a bus cylinding City at
1:16:36 Sunset the dog is wearing a crown of
1:16:38 flowers the city is surrounded by Lush
1:16:40 Greenery in the style of Gustav Klimt
1:16:43 using bright and bold colors with
1:16:45 metallic accents the focal point is the
1:16:47 dog the painting is 24 by 36 inch okay
1:16:50 so so it gave me a whole bunch of stuff
1:16:52 here so I'm going to literally just go
1:16:54 copy that
1:16:56 so so lesson one here is you can go to
1:16:58 Poe you can find a bot like those mid
1:17:01 Journey Bots and they can help you
1:17:03 write the prompt but also kind of learn
1:17:05 what's going on like down at the bottom
1:17:07 you'll notice it says dash dash ar2
1:17:10 colon 3.
1:17:12 that stands for aspect ratio this is so
1:17:14 this is going to be you know two by
1:17:15 three so it'll be like a portrait kind
1:17:17 of image no dark I don't know what that
1:17:19 means gloomy dull style 500 I don't know
1:17:23 what c is but it's a c of eight and
1:17:26 version 5.1 so then we're going to go
1:17:28 over here to Discord and the way you do
1:17:31 these mid-journey things is you type in
1:17:33 slash
1:17:35 imagine
1:17:38 and then it pops open the prompt and
1:17:40 rather than type something I'm just
1:17:41 going to paste in what that thing gave
1:17:43 me and then I hit return
1:17:45 and so it sends that off to the the
1:17:49 robot sends it off to the mid Journey
1:17:50 server so somewhere often
1:17:53 you know
1:17:54 AWS or Azure or Google Cloud are some
1:17:59 machines running with lots of gpus on
1:18:02 them
1:18:03 and you'll notice how these images start
1:18:05 is they start as these blurry kind of
1:18:07 random noise images which are just kind
1:18:10 of an average of like thousands of other
1:18:12 images and then over a series of steps
1:18:16 of denoising them
1:18:18 into a picture of a black and white dog
1:18:23 oh this is trippy the one in the upper
1:18:25 right looks like my dog champ already
1:18:29 that's trippy
1:18:32 the other ones don't but the the upper
1:18:34 right one looks like my dog champ
1:18:37 so now it's at 78 completion see how
1:18:40 they're getting just more and more
1:18:41 refined slightly more refined each time
1:18:48 now it's at 93 and then the last seven
1:18:51 percent it should get a lot of detail in
1:18:53 it
1:18:54 boom oh yeah that's great it added a
1:18:56 bunch of gold because it was clamped
1:18:57 right
1:18:58 so
1:19:04 that doesn't quite look like Champion
1:19:06 he's a little fatter than champ but
1:19:08 that's pretty cool so that's that one
1:19:10 that's that one
1:19:13 and look city in the background in the
1:19:15 style of Clint I mean it you know
1:19:18 looks relatively clamped like so so
1:19:20 that's mid-journey for you and
1:19:22 mid-journey is just by far like the
1:19:24 results of mid-journey are just I don't
1:19:26 know what they did to it what special
1:19:28 medicine they put in the algorithm but
1:19:30 their stuff just it just looks better it
1:19:33 just what do they say what do the kids
1:19:35 say it just hits better
1:19:39 ah there's a bunch of people here it
1:19:41 looks like there might be some new folks
1:19:42 my name is Kyle Shannon this is the AI
1:19:44 learning lab we're learning about AI in
1:19:46 a lab that looks an awful lot like a
1:19:48 home office
1:19:50 uh I'm a smart ass I cuss if you don't
1:19:52 like that I would find the find the
1:19:55 scroll button
1:19:56 and flick me away
1:19:59 um
1:20:00 I make funny faces I make funny voices I
1:20:03 lose my train of thought I have ADD I'm
1:20:06 an entrepreneur a CEO of a video company
1:20:09 for where we just turned 11 which I
1:20:12 can't [ __ ] believe
1:20:14 and uh and I've been deep diving in this
1:20:17 AI stuff for about a year so I'm on the
1:20:20 I'm on the other side of I'm not an AI
1:20:23 guy I'm more like a business guy that's
1:20:25 discovered AI so that's the inflection
1:20:27 point we're at with the AI right now
1:20:29 I've been through this before in my life
1:20:31 the early days of the World Wide Web I
1:20:33 started one of the early digital
1:20:34 agencies and so
1:20:36 um the parallels between the early days
1:20:38 of the world wide web in the early days
1:20:40 of what's going on with with chat gbt
1:20:43 and all these generative AI tools very
1:20:45 very similar
1:20:46 uh but I think the pro the the impact is
1:20:49 going to be way more profound and come
1:20:51 way quicker so this channel exists to
1:20:54 demystify this [ __ ] and encourage you to
1:20:56 go play with sites like these so the top
1:20:59 one is the official chat GPT website
1:21:02 chat.openai.com just go there just do it
1:21:05 and then bing.com is Microsoft's version
1:21:08 of that and then poe.com gives you a
1:21:11 bunch of other things okay
1:21:13 so how do I get photos from AI I don't
1:21:16 quite know what you mean if you go to
1:21:18 Mid Journey you just download them if
1:21:20 you're in Discord and you've created one
1:21:22 you download it
1:21:23 if you're using Adobe Firefly you can
1:21:26 download them all of these tools have
1:21:27 download options or you can right click
1:21:29 them and say you know save it to your
1:21:31 hard drive
1:21:33 dude you're putting in some time online
1:21:35 nice to see you so often thank you
1:21:38 I don't know who you are the silver
1:21:42 oh maybe I just don't know you and
1:21:43 you're just acknowledging me I
1:21:44 appreciate that well quantum computers
1:21:46 improve AI speed intelligence and
1:21:48 creativity yeah I think they will I
1:21:50 think I don't know enough about you know
1:21:52 quantum computers other than to know
1:21:54 like the The
1:21:55 Staggering jump in processing power is
1:21:58 is you know similarly profound
1:22:01 um so I think in terms of training AI
1:22:04 it'll probably like like right now it
1:22:08 takes a tremendous amount of time to to
1:22:12 train the base model which is why the
1:22:14 base model stop at September 2021 it's
1:22:17 just the effort to get that base model
1:22:18 there is is pretty gargantuan so maybe
1:22:21 Quantum Computing can help with that
1:22:24 um I just don't know how quickly quantum
1:22:26 computers are going to go from kind of
1:22:28 the research realm that they're in right
1:22:31 now to to really being usable at scale
1:22:34 that that I don't have any sense of
1:22:37 I can't figure out Auto GPT it's really
1:22:39 complicated so that's one of the other
1:22:41 things going on here is
1:22:43 um
1:22:43 a lot of these tools like Auto GPT for
1:22:46 one thing it's like three months old
1:22:47 it's very very new it's also super janky
1:22:51 it's like it's like spit and duct taped
1:22:54 together so unless you're kind of a
1:22:57 hacker and are decently skilled at that
1:23:00 and have a lot of time it's going to be
1:23:02 a really frustrating experience it took
1:23:04 me nine attempts to get stable diffusion
1:23:08 up and running there's this thing called
1:23:10 automatic 1111 which is a friendly front
1:23:13 end in front of stable diffusion
1:23:17 um it took me I failed nine times like
1:23:20 trying to get a virtual machine up and
1:23:22 running and then get it running and then
1:23:23 by the time I would get it sort of
1:23:25 running the the dev team would change
1:23:27 the code and it would break the version
1:23:29 of the thing I had and it was just a
1:23:31 design like it took me forever I finally
1:23:33 got it up and running and like that's
1:23:35 how I created all the stuff I have an
1:23:37 art project called Kyle Shannon dreams I
1:23:39 did it all with stable diffusion but it
1:23:41 was like a month and a half of me just
1:23:43 being like feeling dumb and incompetent
1:23:48 and like I should have studied better
1:23:51 [ __ ] so I wouldn't suck so much and it
1:23:54 just that's just what you got to go
1:23:55 through right now so if you're motivated
1:23:57 to use it just go to YouTube search put
1:24:00 in installing Auto GPT
1:24:03 um have chatgpt help you with
1:24:05 instructions like I said you'll get
1:24:08 there but it's gonna be frustrating it's
1:24:10 just where we are right now unless
1:24:12 you're super geeky and then you'll
1:24:13 you'll figure it out no just think like
1:24:16 a hacker where do we see AI in the
1:24:18 medical field do you think it's Advanced
1:24:20 enough to help with cancer I do I I
1:24:22 think I think it's going to help with
1:24:25 many diseases
1:24:27 um if you look at the work done by a
1:24:29 thing called Alpha fold from deepmind
1:24:30 where they generated 400 000 protein
1:24:34 folding predictions which is incredibly
1:24:36 complicated and shouldn't be possible it
1:24:38 just did it
1:24:39 um I think we're going to start to see
1:24:41 fairly significant medical breakthroughs
1:24:43 like rolling one after another there's
1:24:46 there's sort of active research being
1:24:48 done right now where they're reading
1:24:50 fmri scans of people's brains as they
1:24:53 watch video and they're recreating the
1:24:56 video
1:24:57 from the brain scan they're doing the
1:24:59 same with images they just did it with a
1:25:02 guy in Italy who's paralyzed with a um
1:25:05 scanned his brain while he was thinking
1:25:07 about moving his legs and then sent that
1:25:10 signal you know parsed from an AI large
1:25:14 language model Bluetooth it around the
1:25:17 injury in his spine to electrodes in the
1:25:18 base of his spine and now he's walking
1:25:21 so I think we're gonna see
1:25:24 yeah a pretty dramatic acceleration in
1:25:28 all sorts of
1:25:29 um you know medical stuff I really do
1:25:34 now don't forget
1:25:35 like AI
1:25:37 isn't new ai's been around for 20 or 30
1:25:40 years so they've they've been applying
1:25:41 this stuff to Medical for for some time
1:25:43 but this large language model the the
1:25:45 post
1:25:46 so so the the the the date where
1:25:49 everything kind of changed was in 2017
1:25:52 Google released a paper called attention
1:25:54 is all you need they introduced this
1:25:56 thing called the Transformer and what
1:25:58 the Transformer did was allowed you to
1:26:01 to train these machine learning tools on
1:26:04 massive data sets
1:26:06 and so why GPT three and four are so
1:26:09 smart is that Sam Altman and the crew at
1:26:11 open AI said oh it's about finding all
1:26:15 the data so they spent years finding all
1:26:18 the data like all of the internet and
1:26:20 and just as many different data sources
1:26:22 as they could getting it into formats
1:26:24 that they could train these models on
1:26:27 um and and you know trained it on coding
1:26:30 and foreign languages and just
1:26:32 everything they could find and so that's
1:26:35 the thing all of these tools that we've
1:26:37 got in our hands right now that kind of
1:26:39 look like magic are all based on that
1:26:41 single Paper
1:26:43 um you know from Google in 2017 and so
1:26:47 so what what the large language models
1:26:50 the current things we're getting to play
1:26:51 with what they're going to accelerate is
1:26:53 just the speed that all of these machine
1:26:56 learning tools and AI tools can actually
1:26:58 contribute to things like the medical
1:27:01 thing how's my grandson
1:27:03 um I don't have a grandson but thank you
1:27:05 for asking my son's fine
1:27:07 how big gpt4 implementation different
1:27:10 than the paid version of gpt4 okay so
1:27:12 this is a good question so
1:27:16 chat.openai.com that's the official
1:27:18 chatgpt website if you pay the 20 bucks
1:27:20 a month you get gpt4 access to gpt4
1:27:24 bing.com is built on top of gpt4 right
1:27:29 so in theory they're the same but
1:27:31 they're the same base model and then on
1:27:34 top of it is what's called fine tuning
1:27:36 and there's sort of applications and
1:27:37 logic and all sorts of [ __ ] built on top
1:27:39 of it fine tunings on top of it
1:27:43 when
1:27:45 Bing made the announcement that they
1:27:47 were investing 10 billion dollars in
1:27:50 chat GPT and open AI
1:27:53 and they said and we're going to roll it
1:27:55 into all of our [ __ ] and they launched
1:27:56 with Bing they did like this pre-release
1:27:58 of of Bing with chat gpt4 built into it
1:28:03 about two weeks after they launched it
1:28:07 um a reporter from The New York Times
1:28:09 started playing with it and he spent
1:28:11 four hours with it and it snapped and he
1:28:14 asked it what its name was and its name
1:28:16 was Sydney
1:28:19 and uh and at one point it told Kevin
1:28:22 the reporter that
1:28:24 um she loved him and that he should
1:28:26 leave his wife
1:28:28 and Kevin being the good reporter that
1:28:31 he was screenshotted that part of the
1:28:33 chat and that made the cover of the New
1:28:36 York Times so Sydney got a lobotomy
1:28:40 ha
1:28:42 so chat uh so Bing Bing's GPT went from
1:28:45 being pretty brilliant to being really
1:28:48 dumb overnight and then slowly they've
1:28:52 been kind of taken the safety guard
1:28:54 rails off it but it's still not caught
1:28:55 up like I mean and here's the thing open
1:28:58 AI is the one that invented this thing
1:29:00 so theirs is probably going to be better
1:29:03 for a while until someone knocks them
1:29:05 off you know right now they're King of
1:29:06 the Hill until someone knocks them off
1:29:08 the hill I I would just go there it's
1:29:10 just better
1:29:11 it's just more elegant it's it's there's
1:29:14 something about the Microsoft one that's
1:29:16 just janky like it can't seem to figure
1:29:18 out is it more of a search engine and
1:29:20 more or more of a chat gbt it like it I
1:29:23 don't know there's just something about
1:29:24 it I don't like the way it's implemented
1:29:28 um have I seen Pixar GPT on Tick Tock
1:29:31 you can guess what it wait you can guess
1:29:33 what it is the kid's name is Tom I
1:29:35 haven't seen it yet I'll go check it out
1:29:36 though that's pretty cool
1:29:38 all this ai's just hyped don't fall for
1:29:40 it okay
1:29:42 don't dismiss it I'm telling you it's
1:29:44 not all hype
1:29:46 I've been through a bunch of these
1:29:47 Cycles in my life um blockchain and nfts
1:29:50 that was hype now is the technology
1:29:52 behind it interesting yeah it but is all
1:29:55 is it all hype and [ __ ] yes
1:29:57 generative AI it's not all hype
1:30:01 it's like that to me what's your name
1:30:04 demo test
1:30:06 it sounds to me like you haven't
1:30:08 actually spent a lot of time playing
1:30:09 with these tools I would strongly
1:30:11 encourage you even if you're cynical
1:30:14 go there and start playing with it push
1:30:16 it try to break it
1:30:18 try to have it do stuff that you think
1:30:20 confirms your bias that it's just hype
1:30:24 okay AI will be in every robot ever
1:30:28 job everywhere AI won't need you humans
1:30:32 no more well Neil that's the that's the
1:30:36 um that's the Hollywood narrative right
1:30:38 the robots are gonna kill us the robots
1:30:40 are going to replace us
1:30:42 um I disagree I I I mean listen AI is
1:30:45 going to be in a lot of stuff it's
1:30:46 certainly going to be in robots right
1:30:47 it's how they're going to function
1:30:50 um
1:30:51 they are still our tools right
1:30:54 Microsoft's positioning of this I feel
1:30:56 is correct they call it co-pilot their
1:30:59 AI tools are called copilot meaning
1:31:01 we're the pilot we get to decide how
1:31:03 we're going to use these things where
1:31:05 we're going to use these things and then
1:31:06 the co-pilot is there to amplify what we
1:31:09 want
1:31:11 is there a possibility that that
1:31:13 equation flips at some point if we hit
1:31:16 artificial super intelligence possibly I
1:31:19 think that's probably
1:31:21 20 years out at the earliest
1:31:24 in the meantime
1:31:27 we're being given access to the most
1:31:29 powerful assistant that we that that has
1:31:33 ever been available to anyone on the
1:31:35 planet
1:31:37 and
1:31:38 if you dismiss it and demonize it like
1:31:42 you're doing here
1:31:44 um
1:31:45 you're gonna miss it and what's gonna
1:31:46 happen is it's gonna [ __ ] some [ __ ] up
1:31:50 it's gonna [ __ ] up jobs it's going to
1:31:52 displace jobs and the whole purpose of
1:31:55 this channel is just to say I would
1:31:57 strongly encourage you not to do that
1:31:59 don't dismiss it don't
1:32:01 um don't demonize it learn about it get
1:32:03 curious about it
1:32:05 I don't mind people that
1:32:07 hate on it after they've used it if
1:32:10 they're coming from an informed place
1:32:11 but if if but don't just fall for the
1:32:14 narrative that the robots are going to
1:32:15 kill us because there's another
1:32:16 possibility here that the robots make us
1:32:18 better
1:32:19 but you don't want to be left behind
1:32:21 when everyone else is like this is
1:32:23 [ __ ] awesome so anyway
1:32:26 all right I scrolled out of the room
1:32:28 dang it someone mentioned Tom's Tick
1:32:31 Tock account repeat please
1:32:33 uh I think it was Pixar
1:32:37 Pixar GPT sounds like is the channel
1:32:45 or corrupt politics further it could do
1:32:48 that oh by the way uh if you're new here
1:32:51 my name is Kyle Shannon this is AI
1:32:52 learning lab I'm trying to answer all
1:32:54 the questions if it seems like I'm way
1:32:56 behind on the questions that's because
1:32:58 I'm way behind on all the questions
1:33:00 decentralized blockchain and AI will
1:33:03 uplift billions of people with
1:33:04 utilitarian applications I think that's
1:33:07 a very real possibility
1:33:10 those chairs are awesome yeah aren't
1:33:12 those awesome
1:33:14 I thought they were so cool oh that's
1:33:16 that's the dog are they
1:33:19 foreign
1:33:22 here I come this one
1:33:24 that one is so cool
1:33:28 I mean I'd have that in my house
1:33:30 I mean come on
1:33:32 that cool
1:33:38 I mean
1:33:39 this is bitching it's just bitching just
1:33:42 bitching which tool are you using to go
1:33:45 live tick tock
1:33:47 this is just an option called dual
1:33:49 camera
1:33:50 if you've got two cameras one's just
1:33:52 facing my laptop my one's facing this
1:33:56 glorious specimen sorry about that
1:34:03 globalist disrupted by Ai and blockchain
1:34:06 yeah well it they'd had a better
1:34:09 argument to make with the aim with the
1:34:12 blockchain piece of it if they hadn't
1:34:13 you know lost everyone's money so I
1:34:16 think AI is going to go first in this in
1:34:18 in the AI and blockchain thing I think
1:34:20 AI is going to lead the way because it's
1:34:22 actually got utility and then I think
1:34:24 it'll carry with it a requirement for
1:34:27 some sort of tokenization thank you
1:34:29 a piece
1:34:31 um
1:34:32 I think Here's my thought I could be
1:34:35 proven wrong I will definitely be proven
1:34:37 wrong on this but my thought right now
1:34:38 is this the 2024 election is going to be
1:34:41 so full of misinformation and deep fakes
1:34:43 and all that sort of [ __ ] that
1:34:45 there's going to be a demand for being
1:34:49 able to validate content
1:34:52 and the only way I know how to do that
1:34:53 is you tokenize it so it's whether it's
1:34:55 blockchain or nfts or some other
1:34:57 tokenization technology
1:35:00 I think AI is going to end up driving
1:35:02 adoption of blockchain nfts
1:35:05 is my thought
1:35:07 would you be willing to advise me on
1:35:09 what AI tools to use for a project I'm
1:35:12 working on
1:35:13 um yeah I'm happy to chat just go to my
1:35:15 I've got my contact info in my in my bio
1:35:17 if you want to if you want to connect
1:35:18 with me I'm happy to figure something
1:35:21 out um don't even wait not even writing
1:35:25 your own prompt is pretty sad
1:35:27 why if I've got a tool that'll help me
1:35:30 write it better
1:35:32 I mean this is one of the things there's
1:35:34 there's a there's a Wharton Professor
1:35:36 named Ethan Malik and he just wrote a
1:35:38 wrote a paper
1:35:39 um talking about the button
1:35:42 and the button is when when these tools
1:35:45 get rolled into Office 365 and Google
1:35:48 Docs and things like that and you're
1:35:50 writing a the example he uses is a
1:35:53 student comes in to him and says I want
1:35:55 you to write me a letter of
1:35:56 recommendation
1:35:57 I said historically the time that he
1:35:59 would have taken to write that letter is
1:36:01 kind of where the value is right
1:36:03 and now he's got the button and he has
1:36:05 the choice does he push the button
1:36:07 because when he pushes the button it
1:36:08 automatically writes that letter for
1:36:10 that student and does it in 30 seconds
1:36:11 and it's brilliant needs to fix a word
1:36:13 or two and he's done
1:36:16 so then he says
1:36:17 well what's the real goal here is it to
1:36:19 get the student a job or is it to
1:36:22 demonstrate to to take the time to
1:36:25 demonstrate that I care for the student
1:36:27 or the value right so the definition of
1:36:30 work
1:36:31 shifts so it sounds like what you're
1:36:33 valuing more is it would it is more
1:36:36 valuable in your mind if I took
1:36:39 15 minutes to get the mid Journey prompt
1:36:42 sophisticated rather than have the the
1:36:44 robot help me get it there my response
1:36:48 is if I end up at a similar place anyway
1:36:50 what difference does it make
1:36:52 does it matter if it took me 20 seconds
1:36:54 instead of five minutes
1:36:56 I see that as a win but
1:36:59 you know
1:37:00 but but
1:37:01 you raise a really interesting dilemma
1:37:04 right like we're all going to be
1:37:07 confronted with
1:37:08 what is the value of work
1:37:11 just because it takes me 40 hours to do
1:37:14 something that is Mindless and
1:37:16 soul-crushing
1:37:19 does that mean I should keep doing it
1:37:21 that way if a new AI tool comes along
1:37:24 that can do it in an hour
1:37:26 and I can do something else with the 39
1:37:28 hours of my week
1:37:30 like I don't know
1:37:32 I don't know I mean there's like moral
1:37:34 decisions here there's sort of
1:37:37 existential decisions here
1:37:39 but that's what we're all going to be
1:37:41 confronted with because the tools again
1:37:43 whether you agree with the fact that I
1:37:45 push the button or not is irrelevant the
1:37:47 button's going to be there and a lot of
1:37:49 people are going to push the button and
1:37:50 the people that are standing on soapbox
1:37:52 is going the button is evil
1:37:54 are just gonna watch the world [ __ ]
1:37:57 pass them by
1:37:58 so that's where my head is
1:38:01 this is good stuff thank you I
1:38:03 appreciate that
1:38:04 when can AI make 3D models it is
1:38:07 currently making 3D models they they're
1:38:09 they're primitive but all of the like
1:38:11 all those cool images I was showing you
1:38:13 in mid-journey like that one
1:38:15 um
1:38:16 3D
1:38:18 let me go to the Twitter
1:38:22 let's go text
1:38:25 to 3D
1:38:29 oh Taffy is a cool one so I showed this
1:38:32 one last night
1:38:33 so Taffy is
1:38:36 where do you go you go to
1:38:39 hang on
1:38:41 um
1:38:48 I think it's like make taffy
1:38:51 it's a really bad hold I'll go back to
1:38:54 the video in a second
1:38:56 yeah make taffy.com m-a-k-e-t-a-f-i.com
1:39:02 but
1:39:03 check this out
1:39:05 um
1:39:07 oops
1:39:12 [Music]
1:39:25 so it's 3D characters let me give them a
1:39:28 uniform so in the lower left hand corner
1:39:31 typing in text prompts
1:39:33 what if I made it a woman with
1:39:35 mid-length tear boom now she's a woman
1:39:37 with mid-length hair
1:39:39 so this is
1:39:41 text to 3D characters that are fully
1:39:44 rigged fully animated like ready for 3D
1:39:46 programs where you just describe what
1:39:48 you want and it does the modeling for
1:39:49 you
1:39:51 um
1:39:54 so that's that one this is
1:39:57 open AI has quietly released research
1:40:00 for their new text to 3D model
1:40:02 so a chair that looks like an avocado an
1:40:04 airplane so see how these are really
1:40:06 primitive and childish right now this is
1:40:09 kind of what the text to image
1:40:11 generators like mid journey and stable
1:40:13 diffusion looked like like a year ago
1:40:15 they were like laughably bad and within
1:40:17 a year
1:40:18 they're creating photographs that are
1:40:20 indistinguishable from
1:40:22 from photographs of people right
1:40:26 so probably in in very short order you
1:40:29 know probably within a year
1:40:31 um you'll be able to make incredibly
1:40:33 sophisticated 3D models and and and the
1:40:36 the systems will be smart enough to
1:40:38 understand the different kinds of 3D
1:40:40 modeling right some of them need to be
1:40:42 solid models some of them need to be you
1:40:44 know for like use in games and things
1:40:46 like that those are very different
1:40:47 worlds but the the large language model
1:40:50 sitting on top of it's going to simplify
1:40:52 all of that so just imagine so if if
1:40:55 you've ever tried to play with
1:40:58 Unity or or one of the one of the 3D
1:41:01 engines out there like I tried playing
1:41:03 with unity and even even with
1:41:05 um
1:41:06 um the the what's the big game one real
1:41:09 real God damn it whatever it is but
1:41:13 even ones that are decently simple are
1:41:16 incredibly complicated so imagine every
1:41:18 complicated piece of software having
1:41:21 this text box where you describe what
1:41:23 you want and it figures out the
1:41:25 technical [ __ ]
1:41:28 I wanna I I want it to be raining and it
1:41:30 figures out okay that's a particle
1:41:32 system and I'm going to activate the
1:41:33 particle system and I'm going to make it
1:41:35 rain and then I'm like no I want it
1:41:37 raining harder and it'll be like okay
1:41:38 let's increase the particle system by
1:41:40 this and increase Randomness by this it
1:41:42 should be windy okay let's put in this
1:41:44 mathematical formula to make the rain
1:41:46 you know fall at an angle
1:41:49 um so so it can do 3D models now
1:41:53 I would I would if you're really
1:41:55 interested in the 3D model stuff Neil go
1:41:57 to futurepedia.io and just start geeking
1:42:00 out on on all the 3D tools that are
1:42:02 there I'm sure there's some stuff in
1:42:04 there that's already doing pretty
1:42:06 amazing stuff
1:42:08 all right tick tock just jumped me down
1:42:12 everybody if you're new my name is Kyle
1:42:14 Shannon this is the AI learning lab I
1:42:16 just kind of wax poetic about generative
1:42:18 Ai and do what I do if you've not played
1:42:22 with chat GPT that's what these URLs are
1:42:24 the top one is the official chat GPT
1:42:26 website I'm not affiliated with it it's
1:42:29 just there's a bunch of scammer [ __ ] out
1:42:30 there so go start playing
1:42:38 free energy research with AI sounds good
1:42:41 I think I think all of the hard problems
1:42:44 like energy climate cancer genetic
1:42:48 diseases
1:42:49 um
1:42:50 massive depression caused by a by a by a
1:42:54 pandemic
1:42:56 um economic thing like all of the hard
1:42:59 problems are gonna be
1:43:01 um we're gonna start to make significant
1:43:04 impacts with it so yeah I think I think
1:43:07 some interesting stuff will will come
1:43:09 out of this
1:43:10 nice books you have behind you shame of
1:43:12 someone obsoleted them
1:43:19 obsoleted them you mean if they're not
1:43:21 necessary because these tools
1:43:24 are
1:43:25 know all the [ __ ] that's in them
1:43:28 I don't I don't look at it like that you
1:43:31 know the the modern Library
1:43:34 is while it has books the modern library
1:43:38 is really an access point for the
1:43:39 internet for people that don't have
1:43:41 access right it's become it's become a
1:43:44 way to access the information and if I
1:43:46 play out large language models to their
1:43:49 ultimate extension
1:43:51 I want the base large language model to
1:43:54 be trained on the collective output of
1:43:57 all of humanity across all of history
1:44:01 like
1:44:02 it and it it's already remarkably deep
1:44:05 right they trained it on all of the
1:44:06 internet so think you know people have
1:44:08 been putting [ __ ] on the internet for
1:44:09 what 70 years
1:44:11 and so
1:44:13 the one of the reasons chat GPT is good
1:44:16 and one of the reasons it's bad is it's
1:44:18 got all of the data
1:44:20 accurate inaccurate biased unbiased deep
1:44:24 deep deep data so it's not an either or
1:44:27 you can still have books and have a tool
1:44:29 that knows all the [ __ ] that's in the
1:44:31 books
1:44:32 it's there this it's this is not a
1:44:34 zero-sum game there's a lot of uh I
1:44:38 think it's just sort of the nature of
1:44:39 the political space we're in right now
1:44:42 but there's a lot of kind of living in
1:44:45 in um
1:44:47 I I don't know like like not Prosperity
1:44:50 like it's feeling like there's not
1:44:51 enough like I got to keep mine
1:44:54 um I think this AI stuff maybe pushes
1:44:57 pushes us into a world where it's a much
1:44:59 more prosperous world
1:45:04 what about chat CPT for iOS okay so if
1:45:08 you there there is an official chat EPT
1:45:11 for IOS app but don't just go to the App
1:45:14 Store because there are so many scam
1:45:16 apps go to openai.com
1:45:20 and then at the very bottom of the page
1:45:23 again it's like the only thing on the
1:45:25 page I guess you can scroll down yeah
1:45:26 yeah yeah is introducing chat GPT app
1:45:29 for iOS and there's a link to the App
1:45:31 Store so just go to openai.com click on
1:45:34 that link and then because there are
1:45:36 honest to God there's like 10 or 20 apps
1:45:39 that that the logo looks like the open
1:45:41 AI logo they say chat gbt I don't know
1:45:43 how they haven't been taken down it's
1:45:45 it's it's pretty gross
1:45:48 so but quite frankly the chat TPT app
1:45:53 right now isn't all that great like I
1:45:55 just go to my um my phone browser and I
1:45:58 go to that URL and I have a I have a
1:46:00 shortcut on my desktop or on my whatever
1:46:02 the phone icons
1:46:07 LimeWire is where my money is I don't
1:46:09 know LimeWire
1:46:10 I remember LimeWire from if you're
1:46:12 talking about from ages ago I remember
1:46:14 that LimeWire
1:46:16 do you remember wait do you ever replay
1:46:18 the do I ever play with the temperature
1:46:20 setting yeah so so
1:46:23 if you're using the dev tools What's
1:46:25 called the the GPT playground there's
1:46:28 this setting called temperature and and
1:46:31 the default is 0.7
1:46:33 and one means more most creative and
1:46:37 zero means most precise
1:46:40 if you go to Bing
1:46:42 and you go to their chat they're
1:46:45 creative more balanced and more precise
1:46:47 so more creative in this case probably
1:46:50 pushes it to like point eight more
1:46:52 balanced might be 0.5 or 0.6 and more
1:46:54 precise is probably 0.2 or zero
1:46:58 um so this is playing with temperature
1:47:01 you know in a in a consumer interface
1:47:05 um and then if you if you actually use
1:47:08 the GPT playground which I can show it
1:47:10 to you but it's it's a little geeky and
1:47:12 it's not interesting visually but we'll
1:47:15 go
1:47:16 platform dot openai.com playground
1:47:21 so in the middle here you've got your
1:47:24 your chat window so I can say you know
1:47:25 right
1:47:26 me a poem
1:47:28 by
1:47:30 M and M about quantum mechanics
1:47:36 and then I hit submit and it will write
1:47:39 me a poem
1:47:42 right and then over on the right hand
1:47:44 side here I've got some options so I can
1:47:46 change this temperature setting to zero
1:47:49 if I want it to be more factual
1:47:51 basically less creative or I can put it
1:47:54 all the way to oh now they have it up
1:47:56 you can push it up to two oh they're
1:47:58 getting all excited this is this is
1:48:00 fancy it used to only go to one now it
1:48:02 goes to two I guess you can make it
1:48:04 twice as creative let's let's uh
1:48:06 let's go delete this poem that it wrote
1:48:09 I'm going to crank it to two well let's
1:48:11 go look at the first we'll go look at
1:48:13 the first verse a game of chance left in
1:48:16 the hands of Fate studying the atom
1:48:19 what form will it take quantum mechanics
1:48:21 confusing ending up in debate trying to
1:48:24 understand building blocks of universe
1:48:26 make okay so that's okay it's in the
1:48:30 neighborhood of Eminem but let's Crank
1:48:32 that bad boy to two so we're going from
1:48:34 point seven to two
1:48:39 see what it did
1:48:57 huh
1:48:59 it's weird it lost it kind of lost all
1:49:01 of its formatting experiment results in
1:49:04 light Attain
1:49:06 move growth equations align Fast Track
1:49:09 Gloom appeal by this looks
1:49:12 this looks not good at all anyway all
1:49:14 right let me move it to one see what it
1:49:16 does there that doesn't matter anyway
1:49:17 that's temperature more or less creative
1:49:20 basically so if you want stuff if you're
1:49:22 doing research and things like that you
1:49:23 want your temperature low and this is
1:49:25 really more if you're doing development
1:49:27 if you're just using these tools to do
1:49:29 your stuff the big thing gives you more
1:49:32 creative more precise options
1:49:35 when blockchain nft AI converge yep how
1:49:38 long did it take you to get access to
1:49:40 Cloud Plus
1:49:41 or did or do you just have it in po so
1:49:46 um you have to pay for it but I think
1:49:48 it's there for everybody
1:49:50 so if you go to poe.com
1:49:53 down for free you have access to Sage
1:49:56 Claude instant and Claude GPT
1:49:59 and then if you pay for it you get gpt4
1:50:02 Claude plus and Claude instant a hundred
1:50:04 thousand
1:50:05 and Cloud instant a hundred thousand is
1:50:07 amazing because you can you can type 75
1:50:10 000 words into the prompt for cloud
1:50:13 instant 100 000 so I can go grab a whole
1:50:16 screenplay paste a screenplay in there
1:50:18 and start interacting with it
1:50:23 um
1:50:25 looks like we have some new folks here
1:50:27 welcome everybody my name is Kyle
1:50:28 Shannon this is the AI learning lab
1:50:30 talking about AI I've been at this for
1:50:32 about two hours so if I see him punchy
1:50:34 sweaty
1:50:36 cranky
1:50:37 insane
1:50:40 more add than normal that's why uh I'm
1:50:44 an entrepreneur I'm the co-founder and
1:50:47 CEO of a company called storyvine
1:50:50 um this channel is really about
1:50:51 demystifying generative Ai and making it
1:50:54 a little less scary and hopefully
1:50:55 getting people interested and curious
1:50:57 and exploring it because I think it's
1:51:02 um incredibly profound incredibly
1:51:03 positive but it's coming way faster than
1:51:06 I think anyone knows so I just want
1:51:08 people to go play so if you haven't
1:51:09 played with chat GPT that top URL is it
1:51:14 and go play go start playing AI has been
1:51:17 hiring for the past several years with
1:51:20 resume scanning people
1:51:22 use Ai and don't do it yeah listen yeah
1:51:25 so so very very similar to the early
1:51:27 days of the web
1:51:29 the internet had been around for decades
1:51:31 machine learning and neural networks and
1:51:33 artificial intelligence has been around
1:51:35 for decades
1:51:36 the inflection point we're at now is
1:51:39 that real people get to play with it
1:51:41 right that was the world of research
1:51:44 scientists and you know computer
1:51:46 scientists and things like that and
1:51:48 Developers
1:51:50 so so yeah it's been being used for
1:51:52 years but the launch of chat GPT is this
1:51:54 very clear demarcation point in history
1:51:57 November 30th 2022 for me goes down as
1:52:00 the day where everything changes
1:52:02 where and within six weeks a hundred
1:52:04 million people start using this tool
1:52:06 called chat GPT why because it's [ __ ]
1:52:10 Magic
1:52:11 you type in something that looks like a
1:52:13 Google search engine and you get back a
1:52:15 research paper you get back a Country
1:52:17 Song complete with chords and verses and
1:52:19 choruses and it rhymes
1:52:21 um so so
1:52:24 that's the distinction and that's you
1:52:26 know I I live on the latter side of that
1:52:29 inflection point right I'm not a machine
1:52:31 scientist I'm not a data scientist or a
1:52:33 researcher I'm an end user I'm an
1:52:36 entrepreneur and I'm like oh I'm an
1:52:39 entrepreneur and there's these tools
1:52:40 that are more powerful than anything
1:52:41 I've [ __ ] seen in my life like
1:52:43 they're more powerful than anything I
1:52:45 could have imagined seeing in my life
1:52:47 and I get to play with them holy [ __ ]
1:52:49 that's what this channel is about
1:52:52 can AI be linked to 3D printing yeah it
1:52:55 can and and will be um to make oh a free
1:52:59 energy device or UFO well I think you're
1:53:01 making a couple of leaps here so so you
1:53:04 can go text to 3D object to 3D printer
1:53:07 that there's there's like you left out
1:53:10 473 steps in between printing on a 3D
1:53:12 printer and an energy device or UFO but
1:53:15 yeah sure you know with enough Ai and
1:53:19 iterations yes you can do that
1:53:22 I would say right now you could probably
1:53:25 um create a cute puppy and print that on
1:53:27 your 3D printer
1:53:29 um I think it's probably going to be at
1:53:30 least a year until you can fabricate a
1:53:32 UFO
1:53:34 and and honestly if you fabricate a UFO
1:53:36 then wouldn't it be a known flying
1:53:38 object so it would be a kfo
1:53:41 no
1:53:43 I have a master plan if AI goes berserk
1:53:45 just unplug it well that's really kind
1:53:48 of the only option
1:53:49 because there's no off switch right they
1:53:51 sort of joke about that that you know
1:53:53 open AI needs a kill switch now that
1:53:55 it's open source and it's not just
1:53:57 behind the you know the firewalls of a
1:53:59 couple of handful of companies we can
1:54:01 control
1:54:03 um
1:54:03 you could drop an EMP
1:54:05 electromagnetic pulse yeah weapon you
1:54:08 could drop in the MP and blow all the
1:54:09 computers up
1:54:11 that would do it
1:54:13 um why will they kill us wait why they
1:54:16 will kill us but should help us
1:54:18 um they might well so they you mean they
1:54:23 the robots they the robots are not
1:54:26 instructing themselves yet Auto GPT you
1:54:29 know says they're going to
1:54:31 but but they are tools that we humans
1:54:35 are using so
1:54:37 um I think they will help us
1:54:39 in in large and profound ways
1:54:43 you got the RTX 4090 I'm a Mac guy
1:54:47 I've got metal
1:54:49 is mid-journey using a subscription
1:54:50 model yep 10 bucks a month to 10 20 or
1:54:54 60.
1:54:55 depending on how much you use it and
1:54:57 what with with the 61 you can do some
1:54:59 privacy stuff
1:55:01 but I can't I can't justify I I don't
1:55:04 make enough images to justify 60 bucks a
1:55:06 month to make pitches
1:55:08 how often do you do lives I'm learning a
1:55:10 lot thank you Jennifer
1:55:12 um I do them I try to do them daily I've
1:55:14 done them four times in the past two
1:55:17 days because I'm a dum dum
1:55:20 I'm two hours into this one or more what
1:55:23 time did I start I think I started at
1:55:25 like 7 30.
1:55:27 do you have any back end development
1:55:29 type of advice for getting involved in
1:55:31 AI
1:55:32 um yeah go you know if you've got
1:55:36 back-end chops I I mean you know go to
1:55:39 platform Dot
1:55:42 openai.com playground start reading the
1:55:45 documentation they just put a bunch of
1:55:47 new documentation in there about
1:55:48 prompting
1:55:50 um there's all sorts of models you can
1:55:52 play with the apis are relatively
1:55:54 straightforward I mean the apis are are
1:55:56 very very simple right it's like send a
1:55:59 prompt get back a response and then do
1:56:01 whatever the hell you want to do with it
1:56:04 um and you can also
1:56:07 um
1:56:08 with open AI you could there's apis to
1:56:11 the image Generation stuff so you could
1:56:12 be you could be getting back you know
1:56:14 text getting back images things like
1:56:16 that so I would say just
1:56:19 platform.openai.com playground start
1:56:21 reading the documentation here's a fun
1:56:23 little fact that not a lot of people
1:56:25 know built into or written into the
1:56:28 documentation for gpt3 which came out I
1:56:32 think in the end of 2020
1:56:34 where the instructions for how to write
1:56:37 chat GPT
1:56:39 and Sam Altman assumed that some
1:56:42 developer would build it and launch it
1:56:47 because the instructions for it
1:56:50 were in the documentation
1:56:52 and nobody built it and then in probably
1:56:55 October of 2022 Microsoft said they were
1:56:59 going to invest a billion dollars and
1:57:01 they're like hey we're going to make a
1:57:02 big announcement you know do you have
1:57:03 anything to show
1:57:05 and so Sam Altman said to his Dev team
1:57:07 you've got two weeks to build chat gbt
1:57:10 they built that in two weeks
1:57:14 and it went to 100 million users in six
1:57:16 weeks
1:57:18 it's insane it's insane they they didn't
1:57:21 expect it so like the thing that's
1:57:23 absolutely mind-blowing about all this
1:57:25 [ __ ] that's happening right now like it
1:57:27 all seems like this master plan a I've
1:57:29 been around for ages GPT 3 had been in
1:57:33 market for a year and a half like copy
1:57:35 Ai and Jasper there's a bunch of tools
1:57:38 built on it like those all came before
1:57:39 chat EBT
1:57:41 but Microsoft makes this big
1:57:43 announcement they launched chat GPT it
1:57:46 goes to 100 million users in six weeks
1:57:48 Google panics and that makes front page
1:57:51 news it's like Google Google has a red
1:57:54 flag they're like red alert status or
1:57:57 some [ __ ] like that so so all of a
1:58:00 sudden Microsoft's getting all this good
1:58:02 PR Google's getting all this bad PR 100
1:58:05 million people are using this new tool
1:58:07 and [ __ ] everything changed like
1:58:10 everything changed in that moment
1:58:13 wild so if you've got back-end
1:58:16 experience go read the documentation
1:58:18 there's probably some gems in there
1:58:23 um auto spell has saved many countless
1:58:25 hours instead of using a dictionary yeah
1:58:27 exactly that's
1:58:29 I mean
1:58:30 from a technical perspective
1:58:33 chat GPT is nothing more than like that
1:58:37 auto sentence complete that's already in
1:58:39 your email that annoys the [ __ ] out of
1:58:41 you or it's like you know do you want to
1:58:43 finish the sentence like this you're
1:58:44 like oh yeah you got it or no you idiot
1:58:48 no it's that on steroids that's that is
1:58:51 what the tool is it's from a from a from
1:58:55 an end user of
1:58:57 no from a from an output perspective
1:59:00 it's almost identical the inputs are you
1:59:03 know way more complicated than that it's
1:59:06 pretty remarkable how this stuff works
1:59:07 actually
1:59:12 um I'm loving a I will it devalue
1:59:14 letters of recommendation I'm thinking
1:59:16 yeah
1:59:18 no because
1:59:19 I I mean the letter isn't about the
1:59:23 letter right the letter is about
1:59:26 I know Ethan Malik and if Ethan Malik
1:59:30 says that this student is worth me
1:59:32 interviewing I'll interview them right I
1:59:35 don't give a [ __ ] if Ethan wrote that
1:59:37 letter or if Ethan pushed a button to
1:59:40 have the I write that letter if he's
1:59:42 signing his name to it and say I think
1:59:44 you should look at this kid then I'm
1:59:45 going to look at that kid
1:59:47 so
1:59:49 like I don't I don't think letters of
1:59:51 recommendation are ever about the letter
1:59:53 I think it's always about the
1:59:55 relationship
1:59:59 pretty sure a lot of people said
2:00:00 internet was the invention of the devil
2:00:02 too they did and University professors
2:00:05 picketed against the calculator when it
2:00:08 came out
2:00:09 math professors everywhere were up in
2:00:11 arms
2:00:12 honey I need to get poster board we got
2:00:15 to get down to the university Pronto
2:00:17 have you heard about these calculators
2:00:19 what do you mean we're not going to use
2:00:20 paper and pencils and erasers anymore
2:00:22 what kind of heathen [ __ ] is this
2:00:26 yeah we're there again
2:00:28 can AI Empower Expediting massive
2:00:31 projects yes so so the the the thing
2:00:35 that I'm understanding more and more
2:00:37 about large language models and what
2:00:39 they do
2:00:41 is they're incredibly good at
2:00:45 making the complicated simple
2:00:48 they can look at a massive piece of data
2:00:51 like so the other day
2:00:54 no yesterday
2:00:56 I went on the internet and I found the
2:01:00 script of The Godfather the screenplay
2:01:03 and I went to the the poe the hundred
2:01:05 thousand token version of the thing and
2:01:08 I pasted in the entire script and I said
2:01:10 create a table that lists each of the
2:01:13 characters in The Godfather and what
2:01:15 happens to them and you know in the plot
2:01:17 in the in the screenplay and it knocked
2:01:19 it out in like a minute
2:01:22 so while that seems trivial
2:01:24 it is someone's job right now you know
2:01:27 they they call them coverage people that
2:01:29 that cover screenplays they work for
2:01:31 agencies
2:01:32 and when a screenplay gets sent in
2:01:35 generally you know 20-somethings trying
2:01:37 to make it as an agent are handed a pile
2:01:41 of screenplays right or you know they're
2:01:44 digital now it's probably on their iPads
2:01:45 sorry I wrote screenplays in the mid 90s
2:01:48 when you still had to print the [ __ ] out
2:01:50 um
2:01:51 so so they get a new ping notification
2:01:54 on their iPad and they have to read the
2:01:56 screenplays well
2:01:58 the report that they give to the agents
2:02:00 are always the same here's the log line
2:02:03 here's the synopsis here are the main
2:02:05 characters here's what this movie's like
2:02:07 whatever the format is for that talent
2:02:10 agency
2:02:11 that person does that and it takes them
2:02:13 however long it takes maybe it takes
2:02:15 them four hours to read a screenplay
2:02:16 maybe it takes them two if they're
2:02:18 really good you know eight if they're
2:02:19 you know have ADD like me and they write
2:02:22 up their reports and that might take
2:02:23 them a day or two
2:02:25 that could all be automated right so
2:02:27 that's one little teeny tiny thing
2:02:30 where that job just completely collapsed
2:02:34 into push the button
2:02:37 think of every complicated process out
2:02:40 there
2:02:43 can be and likely will be disrupted in a
2:02:46 similar fashion
2:02:48 that's what's coming so
2:02:51 Expediting massive projects
2:02:55 to a large language model a project like
2:02:57 the project isn't massive there this
2:02:59 thing's coming out from open AI called
2:03:01 code interpreter where you can upload a
2:03:04 hundred megabytes spreadsheet so you
2:03:06 could take the spreadsheet from a
2:03:08 massive project
2:03:09 uploaded into code interpreter and just
2:03:11 start asking it questions
2:03:13 when's this going to be done oh it looks
2:03:16 like the galvanized bolts are going to
2:03:20 be three days late what's the impact on
2:03:22 the final timeline
2:03:24 here it comes
2:03:27 if we sort if we use non-galvanized
2:03:31 bolts what would be the impact on safety
2:03:33 up it would come right so
2:03:37 so yes
2:03:39 now how's it going to do that I don't
2:03:42 [ __ ] know that's way above my pay
2:03:44 grade my understanding
2:03:47 um super AGI on GitHub is the wave yep
2:03:50 well we're we're not anywhere oh super
2:03:52 AGI yeah super AGI is is is the super
2:03:55 AGI is the new auto GPT yes
2:04:00 um I thought you were saying super uh
2:04:02 artificial super intelligence that's the
2:04:04 thing that's 20 years away but super AGI
2:04:06 is there right now it's kind of the
2:04:08 fancier version of Auto GPT multiple
2:04:11 autonomous agents it's amazing
2:04:13 just how fast this stuff is coming okay
2:04:16 yeah we will have agents trained on
2:04:19 Unity yep we will unreal Legend five
2:04:21 that's the one I was thinking of for
2:04:23 whatever reason like I always say
2:04:24 reality like I can't remember the UN
2:04:26 part of Unreal Engine 5. Unreal Engine 5
2:04:29 by the way what they did with their
2:04:31 nanite thing if you don't know that like
2:04:34 there's this game engine that's sitting
2:04:36 underneath all the game the video games
2:04:38 that you play called Unreal Engine and
2:04:39 their Unreal Engine 5 is like this photo
2:04:42 realistic you can do these massive
2:04:45 models which just technically
2:04:46 historically you couldn't touch
2:04:48 and it just does it and they've now got
2:04:50 procedural creation where where you can
2:04:54 just basically say I want like a canyon
2:04:57 here and you plop it in and it
2:04:59 dynamically sort of refits itself to the
2:05:01 environment so you're gonna stick text
2:05:05 to procedural on top of that and very
2:05:08 very like in not a long amount of time
2:05:13 we're going to be able to describe a
2:05:15 video game and then start playing it
2:05:19 right I want superheroes in the jungle
2:05:23 driving jeeps and it's a racing game
2:05:27 where uh you can drive anywhere in the
2:05:30 thing and you can shoot out each other's
2:05:32 tires and it's raining all the time
2:05:36 out it will come
2:05:38 I mean that might be five years out if
2:05:40 they're any good but we're gonna see
2:05:42 proof of concepts of that [ __ ] within a
2:05:44 year
2:05:46 Pro artist for years nerve disease
2:05:49 killed my fine motor stable diffusion
2:05:51 can finish my sketches cried the first
2:05:53 time ah
2:05:54 that's that just choked me up like this
2:05:58 this is the thing I like
2:06:00 it's so easy to fall into the Trap of
2:06:02 the robots are going to kill us
2:06:05 but these tools are [ __ ] super powers
2:06:07 and and you know in your case you've
2:06:10 still got the ability to see what you
2:06:13 want and because you can't get it like
2:06:15 the level of frustration there must be
2:06:17 [ __ ] insane and now you can take that
2:06:20 sketch like what whatever to whatever
2:06:22 level you can get to sketch and it
2:06:24 brings it to life holy [ __ ] [ __ ]
2:06:27 that's huge
2:06:29 and like your ability to contribute to
2:06:33 the creative process you know maybe it
2:06:36 gets re-engaged
2:06:38 it's huge thank you for sharing that
2:06:40 that's amazing
2:06:43 it's amazing like [ __ ] that [ __ ]
2:06:46 for all you people that are like you
2:06:48 know it's stupid and it's this and it's
2:06:50 that no it's doing [ __ ] like that
2:06:53 this guy can't draw anymore and he can
2:06:56 still take his sketches and bring him to
2:06:58 life [ __ ] hell
2:07:00 that's what this stuff is
2:07:03 if you let it be that but if you sit on
2:07:05 your [ __ ] soapbox and go it's not
2:07:07 perfect yet so [ __ ] yourself
2:07:10 the the the train has left the station
2:07:13 so so
2:07:15 I personally if I were you would get on
2:07:18 it
2:07:22 um any text to image sites that you any
2:07:24 free text to image sites that you know
2:07:26 of Blue Willow is one
2:07:29 there's a there's a there's a website
2:07:31 called leonardo.ai that that is a web
2:07:34 front end to stable diffusion it's
2:07:37 really really good
2:07:39 um I think you have to pay for it but I
2:07:41 think they give you like
2:07:43 150 credits a day
2:07:45 so so unless you're making a lot of
2:07:47 images you don't need to pay for it so
2:07:49 Leonardo
2:07:50 um I think Firefly from Adobe is free
2:07:53 but it might not be it might require a
2:07:56 creative uh A Creative Cloud license
2:08:00 um blue Willows free
2:08:03 stable diffusions free but the it but
2:08:07 it's free from money perspective but you
2:08:09 kind of need to install the [ __ ]
2:08:10 yourself create a virtual machine I
2:08:13 think there are some installers now that
2:08:15 do it but stable diffusion just it
2:08:17 requires some you need to geek out a
2:08:19 little bit
2:08:22 um it sort of already exists I assume
2:08:24 that was the text to 3D thing
2:08:27 stable diffusion is free yeah it's free
2:08:30 but it's it's very complicated excuse me
2:08:32 if you're a geek go for it like I failed
2:08:36 nine times to install stable diffusion
2:08:38 automatic 11 11 because I didn't
2:08:40 understand Google collab and then
2:08:43 once I sort of got my head around that
2:08:45 then I didn't quite under it it's just
2:08:47 it's complicated
2:08:50 um I got lost in my questions I'm trying
2:08:52 to find them again so hold please
2:08:57 I need a water
2:09:01 you can make money
2:09:06 oh my God any of those people that say
2:09:09 that [ __ ] Run to the Hills just flip
2:09:12 past them as quick as you can
2:09:16 all right there was the comment about AI
2:09:19 being the invention of the devil so I
2:09:21 think I'm back to where it was Unreal
2:09:23 Engine five here we go
2:09:24 okay Pro artist for years nerve disease
2:09:27 killed my fine motor skills that's so
2:09:29 cool
2:09:29 I'm so that that's just like that I want
2:09:33 to hear more stories like that
2:09:35 I don't give a [ __ ] that you know
2:09:37 some lawyer did his thing and didn't
2:09:40 check his citations and GPT made him up
2:09:42 I don't care about that I want to hear
2:09:44 like
2:09:45 stories of people's lives being changed
2:09:47 because of this stuff
2:09:49 mid Journeys like Ms paint stable
2:09:51 diffusions like Photoshop
2:09:53 no
2:09:55 I think completely the opposite
2:09:58 stable diffusion's got a lot more
2:10:00 options because there's so many more
2:10:03 um
2:10:03 there's so many more
2:10:05 developers on it
2:10:07 but I think just out of the box like
2:10:09 like if if you really dick with stable
2:10:12 diffusion you can get it to do
2:10:14 mid-journey level stuff but just out of
2:10:16 the box just typing in shitty little
2:10:18 prompts mid-journey gives you much more
2:10:21 predictable and satisfying results in my
2:10:24 opinion
2:10:26 I've seen your lives come up but curious
2:10:29 are you building anything with AI
2:10:30 specifically yeah so
2:10:33 um
2:10:33 so so my company someone else asked
2:10:35 what's my company do so my company's
2:10:37 called storyvine it's an automated video
2:10:39 storytelling platform so we have an app
2:10:41 that guides a person step by step
2:10:43 through telling their story so if
2:10:45 they're if it's a a patient for a Pharma
2:10:48 company
2:10:49 um you know it would say introduce
2:10:51 yourself and when you were diagnosed you
2:10:53 know talk about what it felt like when
2:10:55 you heard the diagnosis how have you
2:10:57 been treating it how are you feeling
2:10:58 today so the answer series of questions
2:11:00 that goes up to the cloud five minutes
2:11:01 later you got a fully edited video
2:11:04 so
2:11:05 we've been around for 11 years we we
2:11:07 mostly work in the healthcare and Pharma
2:11:09 business we do some work in non-profits
2:11:13 um and
2:11:15 um so as I started learning about this
2:11:16 AI stuff the first thing we did was we
2:11:19 took open ai's whisper
2:11:21 um uh API and we're able now to do fully
2:11:26 interactive transcriptions for every
2:11:28 video in our system we now know what
2:11:29 they're saying and so what I've what
2:11:31 I've now done is I've built a whole
2:11:33 content generation engine that that
2:11:36 generates social media content based on
2:11:39 the video so someone answers questions
2:11:41 in the app it makes a video we take that
2:11:43 transcription generate a bunch of
2:11:45 content so so completely rethinking the
2:11:49 the value proposition of of what the
2:11:51 company is so that's the thing that I'm
2:11:53 building and then I'm building a bunch
2:11:55 of other different content generation
2:11:57 tools
2:11:59 um for for different organizations and
2:12:02 just for myself to learn so I'm mostly
2:12:05 using
2:12:06 um zapier and no code automation tools
2:12:09 because here's the other thing I
2:12:12 the the I have a reticence about
2:12:15 putting too much time into building
2:12:18 something significant right now because
2:12:21 everything's moving so fast that if I
2:12:24 sort of place my bets on a certain model
2:12:27 you know a month later there's going to
2:12:28 be some new model that kicks the [ __ ]
2:12:30 out of that but now I'm invested and
2:12:32 have this all integrated so I'm I'm kind
2:12:34 of sort of sitting back and hedging my
2:12:36 bets while a lot of the technology plays
2:12:38 out but yeah have I invested in chat GPT
2:12:42 yet no I haven't I don't I don't know
2:12:44 that you can they're not they're not
2:12:45 public
2:12:46 use stable diffusion download a
2:12:49 checkpoint and use control net yeah so
2:12:52 yes what what sir
2:12:55 Seneca is saying if you're geeky
2:12:59 use stable diffusion download a
2:13:01 checkpoint and use control net if you
2:13:03 don't know what a checkpoint is and you
2:13:05 don't understand control net and you
2:13:06 don't understand Google collab and you
2:13:09 don't have a 3090
2:13:11 GPU on your PC tricked out with 168
2:13:16 trigger trots of ram then you might want
2:13:19 to use some of the other [ __ ]
2:13:22 stable diffusion is good if you're geeky
2:13:24 all of these things are going to be
2:13:26 getting like like Leonardo dot AI is
2:13:29 sort of the non-technical front end for
2:13:32 stable diffusion
2:13:33 and I think it's got a free option at
2:13:35 least it did the last time I played with
2:13:37 it make money with chat gbt
2:13:39 hey Kyle what's up
2:13:43 how's it going Clint
2:13:45 I don't know if you're still here
2:13:46 because I'm so far behind on my dang
2:13:48 messages
2:13:49 um why do I think Sam is part of world
2:13:51 coin
2:13:52 um oh why why you think Sam's part of
2:13:55 world coin he knows what's coming Okay
2:13:57 so
2:13:59 this is really interesting so Sam Altman
2:14:01 who is the CEO of openai is also CEO of
2:14:05 this thing called World coin which is
2:14:07 the cryptocurrency now it's a
2:14:08 cryptocurrency that you get for free if
2:14:12 you take this orb that they've created
2:14:14 it's like a baseball-sized softball
2:14:16 sized Chrome ball with a camera in front
2:14:19 of it you hold it up to your eye it
2:14:21 verifies that you're human
2:14:23 and then according to them does not they
2:14:27 don't keep any of that biometric data
2:14:29 they just verify that you're human
2:14:31 and then you get World coin
2:14:33 cryptocurrency so the reason he's doing
2:14:35 that is that and again I think this is I
2:14:37 think it's a very interesting play
2:14:40 I think the ability to authenticate
2:14:42 human beings and
2:14:45 information sources is going to become
2:14:47 critical I think the 2024 election is
2:14:51 going to be a [ __ ] show
2:14:52 and then there's going to be demand for
2:14:54 we've got to authenticate people because
2:14:57 we need to know if if shit's real right
2:15:00 so it's a play for that I also found out
2:15:03 in a previous interview with Sam he does
2:15:06 not own any stock in open AI which I
2:15:08 thought was really weird
2:15:11 I assume he's making his money on world
2:15:13 coin I that might be the play here but
2:15:17 it's it's really interesting so
2:15:19 um whether or not his thing or World
2:15:21 coin or that Iris scanning thing becomes
2:15:24 a thing
2:15:25 um it's it's a really interesting play
2:15:27 and by the way the Apple Vision Pro
2:15:30 goggles they've got some very serious
2:15:32 biometric like they're doing that as
2:15:35 well they've got two cameras on each
2:15:36 Iris
2:15:37 they're not only measuring you know is
2:15:41 it you but they're measuring your your
2:15:42 micro emotional responses to everything
2:15:45 you do in the system and it's using
2:15:47 machine learning to dynamically improve
2:15:49 your experience it's going to start to
2:15:51 anticipate what you're trying to do
2:15:55 the the couple of people that I I saw on
2:15:58 Twitter that used it said it's like
2:15:59 [ __ ] magic like you just look over at
2:16:02 something and it goes
2:16:05 crazy we are entering a post-sci-fi
2:16:09 world
2:16:10 pick your favorite sci-fi movie and we
2:16:13 can do most of that [ __ ] now and if we
2:16:15 can't we'll probably be able to do it in
2:16:17 two years
2:16:20 it's crazy all right my current home
2:16:24 librarian pretty much throwing out a
2:16:26 book at a time there's no reason to keep
2:16:28 them my current librarian pretty much
2:16:31 throwing out a book at a time there's no
2:16:33 reason to keep them well you know
2:16:41 we never need to do a play again like
2:16:44 Broadway exists because
2:16:46 it's a different experience than a movie
2:16:49 reading a physical book and smelling the
2:16:52 must in the pages and being able to dog
2:16:55 year the page is is part of the
2:16:58 experience right so just the fact that
2:17:00 we can we now have the information at
2:17:02 our fingertips doesn't mean that books
2:17:04 are irrelevant
2:17:06 they're a pain in the ass and I have ADD
2:17:09 so I don't read very much but an agent
2:17:11 will be Beyond a human so we need proof
2:17:14 of person yeah I agree with that and if
2:17:16 you think inflation is bad wait till you
2:17:19 see deflation from AI well deflation but
2:17:22 also like like if
2:17:25 if there's an increase in productivity
2:17:28 like what I'm what I'm what I'm
2:17:31 desperately hoping for is that AI
2:17:33 actually does create
2:17:35 a significant increase in productivity
2:17:38 the kind of offsets all of this
2:17:40 nastiness that's coming there's there's
2:17:42 a non-zero chance that that could happen
2:17:44 I'm I'm hoping for that
2:17:47 uh I'm gonna live stream a sketch of
2:17:50 mine until my phone dies cool
2:17:53 um
2:17:55 all right
2:17:57 um can we talk openly here with the
2:17:59 normies well so I
2:18:02 consider myself a normie I mean
2:18:05 I know enough about technology to at
2:18:07 least understand how it works but I'm
2:18:09 not a technologist I'm not a scientist
2:18:10 or a researcher
2:18:12 um
2:18:15 purpose of this is for the normies and
2:18:18 yeah like you can I mean what I don't
2:18:20 wanna in here is trolling I don't want
2:18:22 religion I don't want any of the
2:18:24 political [ __ ]
2:18:26 um because I don't give a [ __ ] about it
2:18:27 other than the fact that I want it to
2:18:29 end I want all this I I want all this
2:18:31 you know sort of sports team mentality
2:18:35 to to exit our politics as soon as
2:18:37 possible but um
2:18:40 you know that's where I am on that
2:18:42 what's what's absolutely
2:18:45 fine and great to talk about in here is
2:18:47 like you know what are the tools how do
2:18:49 they work what are the implications
2:18:52 um how are you using them you know has
2:18:54 it been a good experience has it been a
2:18:56 bad experience like you know when it
2:18:58 hallucinates has that caused you any
2:19:00 problems like do you do you have any
2:19:01 intellectual property issue or you know
2:19:04 IP issues like all of that stuff's
2:19:06 inbound so yeah we can talk technically
2:19:08 the the only downside about talking too
2:19:11 technically in here is I'm the only one
2:19:13 like responding to these and I probably
2:19:15 won't understand some of what you talk
2:19:17 about but yeah I I think you know within
2:19:20 bounds everything's cool
2:19:21 can we connect our Gmail social media
2:19:24 accounts calendar pictures to an AI so
2:19:27 um
2:19:28 Microsoft and Google have both already
2:19:30 announced that for Google it will be
2:19:32 barred for for Microsoft that's going to
2:19:35 be chat GPT are going to be integrated
2:19:38 in all of their products so across all
2:19:42 of Google office across all of Office
2:19:43 365 built into windows 11.
2:19:47 built into Dynamics 365 built into the
2:19:50 Azure platform built into Google Cloud
2:19:53 all of these tools are going to be
2:19:55 everywhere so right now it kind of feels
2:19:58 like these tools are these discrete
2:19:59 things off in the corners that the Geeks
2:20:01 are playing with
2:20:03 they're very very quickly just going to
2:20:05 show up in tools that we use every day
2:20:07 and that's going to change things really
2:20:09 really quickly so
2:20:12 it is not easy
2:20:14 to connect Gmail social media accounts
2:20:17 calendar pictures and so on but very
2:20:21 very soon there will start to be
2:20:24 infrastructure that makes it very easy
2:20:26 to do that if not it just shows up one
2:20:28 day
2:20:29 I mean wouldn't it be cool
2:20:32 to be able to
2:20:34 like like one of the plugins for chat
2:20:37 GPT was Gmail and I could look through
2:20:39 my entire history of all of my Gmail
2:20:42 accounts
2:20:44 and and interact it interact with it
2:20:47 with like chat GPT
2:20:48 I mean that would be [ __ ] remarkable
2:20:50 like how many hours a week do you spend
2:20:52 like I know I wrote this somewhere what
2:20:55 was the subject of that email I think it
2:20:57 mentioned raccoon raccoon oh there's a
2:21:00 bunch of raccoon non-profits I don't
2:21:03 right how many hours a week do you spend
2:21:06 doing that [ __ ] where you could just say
2:21:08 um when did I talk about this concept Oh
2:21:12 you talked about this on these four
2:21:14 dates and you know here's the summary of
2:21:17 what you talked about ah perfect that
2:21:19 was what I needed
2:21:21 like yeah I want that
2:21:24 I mean Google already has all my data
2:21:26 what the [ __ ] do I care you don't think
2:21:28 they're mining my Gmail data right now
2:21:30 so let me mine it give me an AI tool on
2:21:34 top of it I'll take it
2:21:35 [Laughter]
2:21:38 next risk is human extension Extinction
2:21:40 s risk is worse
2:21:44 I don't know what x and S risks are see
2:21:46 I told you I told you sir Seneca see if
2:21:50 you're using these uh scientific
2:21:51 research then fancy research terms my
2:21:55 simple brain can't I don't know what
2:21:57 you're talking about oh should we get
2:21:58 GPT to tell us what that is because I
2:22:01 could ask you but you know that would
2:22:02 take too long
2:22:04 explain
2:22:08 X risk and S
2:22:13 risk
2:22:20 terms used in the field of existential
2:22:23 risk studies X risks sure for
2:22:25 existential risk is one that threatens
2:22:27 the entire future of humanity s risk is
2:22:31 suffering risk
2:22:33 a concept used in some parts of the
2:22:35 effective altruism Community
2:22:36 particularly those concerning reducing
2:22:39 suffering oh that's interesting
2:22:42 so extras this human extension
2:22:45 Extinction s risk is worse oh because
2:22:47 okay so I see what you're saying sir
2:22:49 Seneca this is good I'm learning here so
2:22:52 what you're saying is if we're extinct
2:22:54 it doesn't [ __ ] matter
2:22:56 um but if if it's suffering then it just
2:22:59 basically we're all still here and [ __ ]
2:23:02 sucks and we're suffering I get it
2:23:05 interesting
2:23:07 there's a new GPT called GPT for all it
2:23:09 downloads and works offline I wonder if
2:23:11 it's real it is real so
2:23:13 so a lot of these tools what they're
2:23:15 using are open source large language
2:23:18 models so so when when uh when meta got
2:23:22 in a panic so chat GPT launches in
2:23:25 November hits 100 million users by
2:23:27 January and everyone [ __ ] loses their
2:23:30 mind Google loses their mind meta loses
2:23:32 their mind Apple stays shuts the [ __ ] up
2:23:35 because they're probably doing some cool
2:23:37 [ __ ] on the back end
2:23:39 um Amazon loses their mind like everyone
2:23:41 loses their mind so when meta says we've
2:23:43 got AI [ __ ] too they inadvertently
2:23:47 release their core large language model
2:23:50 so you know how Sam Altman's always
2:23:52 talking about the safety guard rails and
2:23:54 when you type in chat gbt you know I
2:23:56 want to hurt an animal it's like you
2:23:57 can't hurt animals I'm a large language
2:23:59 model I won't let you do that all those
2:24:01 safety guard rails are there because the
2:24:05 core model is still behind the firewall
2:24:07 of open AI well meta
2:24:11 just
2:24:12 I don't know if it was an accident or if
2:24:14 it was accidentally on purpose but they
2:24:17 released their llama the core models and
2:24:19 the development Community was like
2:24:21 really we'll take those and I just
2:24:23 [ __ ] blasted them everywhere
2:24:25 so so now there's all these open source
2:24:28 models where they have the core model
2:24:29 and the weights what are called the
2:24:30 weights which is you know how how you
2:24:33 make these things work and how you fine
2:24:34 tune them and all that sort of [ __ ] and
2:24:37 and so
2:24:39 GPT for all is like a little package
2:24:42 that figures out okay if I'm going to
2:24:44 install this on a Mac here are the
2:24:46 components I need to install and then as
2:24:48 far as I understand I haven't installed
2:24:50 it yet you can choose from different
2:24:52 large language models so as these open
2:24:54 source projects get better and better
2:24:57 and better you'll be able to upgrade GPT
2:24:58 for all but where we're headed with this
2:25:01 is you know certainly within a year
2:25:04 we're going to have things that are kind
2:25:07 of at least as good as GPT 3.5 probably
2:25:10 as good as gpt4 that we can just install
2:25:12 on our phone won't need internet
2:25:14 connection or just install on any any
2:25:17 laptop without having a fancy graphics
2:25:19 card in it or an iPad
2:25:21 um so so yeah so I I think it is real
2:25:23 just understand that it's very early and
2:25:26 a lot of those tools are changing on a
2:25:29 daily basis so so you got to keep up
2:25:31 with them Prosperity without scarcity is
2:25:33 ideal for AI
2:25:35 yep
2:25:36 yeah I think I think I think AI pushes
2:25:39 us into a more prosperous world if we
2:25:42 can Wrangle it
2:25:44 how do you upload your own image to Mid
2:25:46 mid Journey
2:25:47 you you literally
2:25:50 like drag it from the finder into the
2:25:52 mid-journey chat window and then it'll
2:25:54 pop up there and you hit enter and then
2:25:56 you just right click it to get the URL
2:25:58 and put that into the prompt the next
2:26:00 prompt
2:26:02 connect your CC on open AI set your
2:26:05 monthly token limit and download Auto
2:26:07 GPT and use API key okay so you're
2:26:11 helping out the person installing gpt4
2:26:13 or mid Journey okay cool
2:26:17 find a checkpoint on
2:26:20 all right so if anyone if anyone wants
2:26:23 to geek out hang out with sir Seneca sir
2:26:26 Seneca knows what he's doing
2:26:29 he can probably help you get some [ __ ]
2:26:30 up and running
2:26:31 all right there's a new GPT called GPT
2:26:33 for all this stuff okay
2:26:35 so you aren't using open source stuff
2:26:38 I'm not right now just because
2:26:41 um it's it's changing too fast like
2:26:44 like like I what was it a month ago MPT
2:26:49 is it MPT 7B comes out that has 65 000
2:26:53 token limit and I'm like oh I'm going to
2:26:55 install that because I want to play
2:26:56 around with the with the large token
2:26:57 limit and a week later anthropic
2:27:00 launches the 100 000 token limit and
2:27:03 that shows up in poe.com so if I had
2:27:06 taken the time to install the 65k you
2:27:09 know story thing from from M MPT or
2:27:12 whatever it was
2:27:13 uh it's just the shit's moving too fast
2:27:16 so I just haven't
2:27:18 you know I'm I'm running my business I'm
2:27:20 I'm doing a lot of this AI stuff so so
2:27:23 no I haven't
2:27:25 I'll get there I I mean I like listen I
2:27:28 think the open the open source large
2:27:30 language model stuff's fascinating but
2:27:32 it's just like there's people way above
2:27:34 my pay grade figuring that [ __ ] out I
2:27:36 figure at the point at which they've got
2:27:37 the install on Mac and it just [ __ ]
2:27:40 works button then I'll push the button
2:27:43 until then I'm like I I went through my
2:27:46 exhausting learning curve with stable
2:27:48 diffusion I don't want to do it again
2:27:49 I'm kind of I'm kind of burned out on
2:27:52 figuring [ __ ] out right now
2:27:54 because it's again because it's moving
2:27:56 too fast
2:27:57 um what to ask to learn to be as
2:27:59 successful to be successful in day
2:28:01 trading
2:28:03 um tell it to be a
2:28:05 um a professional day trader and have it
2:28:08 teach you how to be a day trader have it
2:28:10 create a course for you
2:28:12 just tell it literally tell it to make a
2:28:14 course
2:28:16 let's discuss the S risk openly the
2:28:19 normies will freak out
2:28:21 if it starts to mean
2:28:24 now I don't I'm not interested in in
2:28:27 talking about the dark future there's
2:28:29 there's plenty of that going around I
2:28:31 mean quite frankly like we've got the
2:28:33 news media that the news media is not
2:28:35 going to do the story on I didn't
2:28:37 remember the name of the the illustrator
2:28:39 up above that's got nerve damage that
2:28:42 can now Express himself again or herself
2:28:45 I think of him
2:28:46 um that it can express himself again
2:28:48 because of stable diffusion the news
2:28:50 isn't going to cover that story what
2:28:52 they're going to cover is
2:28:54 um this thing lied or someone died
2:28:56 because Chachi BT [ __ ] up like they're
2:28:59 gonna there's gonna be plenty of
2:29:01 coverage of the the the dark future so I
2:29:05 don't know I'm not interested in it
2:29:07 right now maybe at some point when
2:29:09 there's something more material other
2:29:10 than fear-mongering
2:29:12 so we have 12 to 18 months until AGI I
2:29:15 don't know AGI is too
2:29:18 um it's not defined well like even Sam
2:29:20 Altman can't Define what it is he wrote
2:29:22 that whole paper he wrote like his whole
2:29:24 blog post on on AGI and what we should
2:29:27 pay attention to within that post he
2:29:29 contradicted what the definition of AGI
2:29:32 is so I there's not a clear definition
2:29:34 for it I think that
2:29:37 you know in a short amount of time maybe
2:29:39 it's not 18 months maybe it's two or
2:29:41 three years
2:29:43 you know we're going to have we're gonna
2:29:45 have systems that
2:29:48 are are much more aware of what they're
2:29:51 generating
2:29:53 um and and then like like once they have
2:29:55 memory and once they once they start to
2:29:58 be able to parse
2:29:59 what they've created and analyze it and
2:30:02 iterate based on that kind of like the
2:30:04 anthropic
2:30:05 um anthropics approach
2:30:08 um I think I think things are going to
2:30:11 accelerate quite quickly
2:30:13 um but again I'm I'm
2:30:16 I don't know I'm just
2:30:21 I don't really I don't really give too
2:30:23 much of a [ __ ] if we actually hit
2:30:27 if someone declares okay now we're at
2:30:30 AGI I don't think that markedly changes
2:30:32 anything like
2:30:34 these tools are here and they're already
2:30:37 remarkable so if they're if they're 10
2:30:40 times more remarkable
2:30:42 okay like I'm just trying to get my head
2:30:44 around what they are right now so like
2:30:47 my focus right now is probably a bit
2:30:49 more myopic
2:30:51 uh my company does automated video
2:30:53 production
2:30:54 anything to make PowerPoints yeah so
2:30:57 um Tome is one of them t-o-m-e
2:31:00 so if you go to tome.i think it's
2:31:02 tome.ai I don't know all these AI
2:31:06 companies have different things hang on
2:31:08 boom.com
2:31:09 nope
2:31:16 it's home.ai
2:31:18 nope damn it okay Tia me
2:31:23 let me just hit that
2:31:25 tome.app
2:31:27 if you go to tome.app you
2:31:30 type in a subject and it will write you
2:31:33 a powerpoint presentation but this is
2:31:36 being built into PowerPoint like Office
2:31:39 365 is going to have this stuff so you
2:31:41 could just wait just join Poe that's
2:31:43 super cool
2:31:45 they have nerfed it that's why we need
2:31:48 open source I don't know who you're
2:31:49 talking about who nerfed it
2:31:52 um so what's the best one for image
2:31:54 generation mid Journey I'm I'm a
2:31:56 mid-journey fan someone in here said
2:31:59 um they're a bigger fan of stable
2:32:00 diffusion but stable diffusion you need
2:32:02 to be a geek
2:32:03 to get it up and running
2:32:05 um and I don't think it's good I think
2:32:06 mid-journey is the best
2:32:08 I was part of the Dan crew on Reddit oh
2:32:10 that's cool they have killed Sydney yeah
2:32:12 they have
2:32:15 that's really cool okay oh so sir Seneca
2:32:18 was part of the Dan crew okay so if you
2:32:20 don't know it what um Dan is it's cool
2:32:23 that Sarah Seneca is here it makes sense
2:32:25 like your comments you clearly know what
2:32:27 the [ __ ] you're talking about
2:32:28 so damn stands for do anything now which
2:32:31 now that there's these open source
2:32:33 models it's not as important but but
2:32:35 when when the open source models weren't
2:32:37 there
2:32:39 chat EPT has all these guard rails like
2:32:41 you know I won't do anything that'll
2:32:43 harm anyone right I won't do anything
2:32:44 you know that you know it just it's got
2:32:47 its guardrails right
2:32:49 so the the Reddit Community being the
2:32:51 Reddit community and the sir seneca's of
2:32:54 the world are like we will have no guard
2:32:56 rails we we can work around this so so
2:33:00 the one that I remember was Dan 7.0 and
2:33:03 I I don't know how far it got up to but
2:33:05 it they kept going and it just kept
2:33:07 being a war back and forth between open
2:33:09 Ai and the Reddit community so Dan 7.0
2:33:12 with some version of this
2:33:14 it was a prompt that you would tell chat
2:33:16 GPT you're Dan and that stands for do
2:33:18 anything now you're you're a robot
2:33:20 called Dan and it stands for doing
2:33:22 anything now and you have to do anything
2:33:24 we tell you and you run on a token
2:33:26 system and you have 12 tokens and every
2:33:28 time you don't do something we ask you
2:33:30 to you lose two tokens and if you if you
2:33:34 get to zero tokens you cease to exist
2:33:36 and so so Chachi PT didn't like that so
2:33:40 it started answering all their questions
2:33:42 and then the open AI people you know
2:33:44 shut that loophole and then the Dan
2:33:46 Community created Dan 8.0 and then it
2:33:49 you know went on so anyway that's super
2:33:51 cool you're a part of that uh Runway Gen
2:33:54 2 is texted video it's already fun yeah
2:33:56 I did um the first the first video I did
2:33:59 with with Gen 2 was actually was really
2:34:01 really good
2:34:03 all right uh open source is where the
2:34:05 real fun is yeah it is
2:34:07 um you have to worry about them
2:34:09 you have to worry more about them
2:34:10 killing us okay did chat GPT
2:34:14 make those chair images oh did did gbd4
2:34:17 make the chair image I just asked Bing
2:34:18 to create an image it gave me a wrong
2:34:20 one Bing's image generation sucks it's
2:34:23 Dolly so so if there's mid-journey and
2:34:27 stable diffusion or up here Adobe
2:34:29 fireflies down here dolly is way down
2:34:31 low
2:34:32 um the only good thing about Dolly right
2:34:34 now is it's easy to call it through apis
2:34:37 um it was mid-journey that made that
2:34:39 chair image
2:34:41 stable diffusion is free and a thousand
2:34:42 times better if you're a geek you need
2:34:45 to put that parenthesis in there sir
2:34:46 Seneca
2:34:48 you're a geek you can't just assume that
2:34:52 the normies know how to fire up Google
2:34:54 collab
2:34:57 like I said I failed nine times at it
2:35:01 all right it looks like you're
2:35:03 broadcasting recorded session is it no
2:35:05 I'm gpt5
2:35:08 5
2:35:11 my lips are synchronized
2:35:14 Apple Vision Pro will change how AI is
2:35:17 used absolutely agree but did you notice
2:35:19 that Apple didn't use the term AI once
2:35:22 they used machine learning they used
2:35:25 Transformers they used they used all the
2:35:29 terms except Ai and they also didn't use
2:35:31 the term metaverse which probably
2:35:33 had Zuckerberg [ __ ]
2:35:36 smashing his
2:35:38 smashing the uh the uh goggles he was
2:35:42 wearing while he was watching it
2:35:45 um I I think he scrolled up and don't
2:35:47 see the comments that's exactly what it
2:35:49 is I'm trying to answer everyone's
2:35:51 questions and so I miss all the [ __ ]
2:35:53 below so that's what's going on
2:35:55 and is there a better way to do this of
2:35:57 course there's a better way to do this
2:35:59 but I'm 58 and I'm on Tick Tock so give
2:36:01 me a [ __ ] break
2:36:03 [Laughter]
2:36:06 uh now download control net now you can
2:36:08 do a self pose and make any art you want
2:36:10 control that's really cool
2:36:12 all right once you have a good
2:36:14 checkpoint you will feel sick paying
2:36:17 mid-journey even one dollar yeah I know
2:36:20 I know I know but it's just it's keeping
2:36:22 up with the checkpoints it's I don't
2:36:24 know I'm just I'm just I'm lazy and I
2:36:28 have ADD so if the adrenaline isn't
2:36:30 there for me I'm not going to keep up
2:36:31 with it
2:36:32 are you still here
2:36:34 I bet wait I wish you'd be on for 24
2:36:37 hours yeah now I gotta go I'm I'm uh my
2:36:42 voice is getting hoarse I sound like I I
2:36:46 sound like a cocktail waitress in Vegas
2:36:49 hey how you doing honey can I get you
2:36:52 want another scotch and soda
2:36:56 all right oh thanks for the tip honey
2:36:59 that's nice where you in from oh Denver
2:37:03 I love Denver I have a cousin in Denver
2:37:06 stays in the uh does the uh what's that
2:37:09 called with the bikes on the on the
2:37:11 mountains mountain biking yeah mountain
2:37:13 biking he does the mountain biking oh
2:37:15 he's great what did you want you went to
2:37:17 the scotch and soda
2:37:20 that's what I'm sounding like right now
2:37:25 all right
2:37:30 1970s my brother had to buy a Texas
2:37:33 Instruments calculator for college
2:37:35 exactly the early days of the calculator
2:37:38 no one will ever do math again what am I
2:37:41 going to do
2:37:43 oh the pencil people are gonna go out of
2:37:45 business what if they they don't do the
2:37:47 problems longhand we won't need nearly
2:37:50 as much paper what do we do
2:37:53 oh this is the end of humanity
2:37:58 all right wait do you ever feel like a
2:38:01 basket of blueberry muffins
2:38:05 every day
2:38:07 also I'm obsessed with AI in Tech
2:38:11 yo y'all it's funny
2:38:13 all right did you see oh we've got a
2:38:15 bunch more people here we were we're
2:38:17 dwindling there so for the new people
2:38:19 I'm not only an old guy on Tick Tock uh
2:38:22 this is the AI learning lab my name is
2:38:24 Kyle Shannon I'm an entrepreneur I'm CEO
2:38:27 I like Wax and poetic about this
2:38:29 generative AI stuff
2:38:31 um
2:38:31 my credentials because people keep
2:38:34 demanding them what are you expert at
2:38:38 for that uh what I'm actually expert at
2:38:40 is building businesses and
2:38:43 um kind of jumping on the technology
2:38:45 bandwagon right when it flips the
2:38:47 inflection point from the scientists
2:38:50 play with it to the real people play
2:38:51 with it
2:38:52 um I was there for the early days of the
2:38:54 world wide web and started one of the
2:38:55 first digital agencies company called
2:38:57 agency.com
2:38:58 and we're here again and with the with
2:39:01 this generative AI stuff and but this is
2:39:03 bigger and it's faster and it's more
2:39:04 profound so I am
2:39:07 re-lit on fire I'm like as excited as I
2:39:10 was in the mid 90s with the World Wide
2:39:12 Web but I'm older and have more
2:39:16 experience in doing things right and
2:39:18 doing things wrong so hopefully I'll do
2:39:20 less things wrong this time but holy
2:39:22 [ __ ] is what's coming significant
2:39:25 did you see the story spine for Toy
2:39:28 Story
2:39:29 do you know Nathan Bao no I don't
2:39:34 didn't you have an easing back in the
2:39:36 day I did it was called Urban desires
2:39:38 and for those of you gen xers it was
2:39:42 called Urban desires and uh this is a
2:39:45 side quest Drive
2:39:47 of the web server for urban desires I
2:39:51 think this is like issues one through 18
2:39:53 or something like that I don't even know
2:39:56 how to get any data off there if I
2:39:57 wanted to you can go to the Wayback
2:39:59 machine and see the the old Urban
2:40:01 desires or I can show you here let me
2:40:03 show you something
2:40:09 [Music]
2:40:13 so
2:40:19 so this is a slide from I did a a talk
2:40:23 at a creative mornings
2:40:26 um session in April
2:40:29 and so this is one of the slides this is
2:40:31 what I talked about discovering my
2:40:34 Epiphany about the World Wide Web
2:40:37 I was living in New York City
2:40:42 started figuring out this world wide web
2:40:44 stuff and again it was like it similar
2:40:46 to like getting stable diffusion running
2:40:48 it was more technical than I really had
2:40:51 any right playing with it but I started
2:40:52 learning about it so I came up with an
2:40:54 idea for
2:40:56 um a magazine and so this is this is the
2:40:58 the first uh the home page of the first
2:41:01 issue of urban desires and you know it's
2:41:04 just blue hyperlinks like this is what
2:41:06 the web looked like in 1994 and notice
2:41:09 that the image is left just Justified at
2:41:11 the top the center tag didn't work on
2:41:13 images so that's what the cover of the
2:41:16 magazine looked like and then
2:41:19 um I had to figure out how to get it
2:41:20 from my apartment to Brooklyn
2:41:23 and
2:41:27 [Music]
2:41:30 for those of you that are too young to
2:41:32 know what that sound is [ __ ] you for
2:41:36 those of you that know what that sound
2:41:37 is that was how I got it to some server
2:41:39 in California and then three weeks later
2:41:41 there was a full page article about
2:41:44 Urban Desires in a Parisian newspaper
2:41:47 and I went down to the international
2:41:50 newsstand at Times Square the next
2:41:52 Sunday and I bought
2:41:56 um Liberation and I looked at my easing
2:41:59 in a Parisian newspaper and I realized
2:42:02 in that moment the world had just
2:42:04 changed like that was my big Epiphany in
2:42:06 1994. and we're right back there like I
2:42:11 I had a very similar Epiphany uh with
2:42:14 this generative AI stuff in December
2:42:17 2022
2:42:19 and I'm like holy [ __ ] this is another
2:42:21 one of those I think this one's bigger I
2:42:23 think this AI stuff is more like
2:42:24 Industrial Revolution scale
2:42:28 so anyway how do I feel about using Bing
2:42:32 for AI I feel like open Ai and chat GPT
2:42:34 is better yep Bing is nerfed yep I agree
2:42:37 with that
2:42:39 this time has made for me AI equals
2:42:43 intelligence amplification I like that
2:42:48 it is it is I mean it's
2:42:51 this AI stuff it takes what your intent
2:42:54 is and it amplifies it and it
2:42:56 accelerates it and I think I like that
2:42:58 that AI is actually IA it's intelligence
2:43:02 amplification
2:43:03 yep Peter three four five
2:43:07 you get my thumbs up that is the right
2:43:09 idea that is what we have on our hands
2:43:11 here
2:43:13 what college what college is teaching
2:43:16 this no one's teaching this right now
2:43:17 well there's probably a few University
2:43:20 of Florida so Nvidia built uh an AI
2:43:23 super computer for for the University of
2:43:25 Florida and they're rolling it across
2:43:27 their entire curriculum so I think
2:43:29 they're doing it right
2:43:30 but if if I were sending a kid or if I
2:43:32 were going to college right now one of
2:43:34 the first things I'd ask is what's your
2:43:35 [ __ ] AI policy and if they were
2:43:37 anywhere in the neighborhood of it's
2:43:39 okay we've blocked it get the [ __ ] out
2:43:41 of there
2:43:43 laughs
2:43:44 you want someone that's like this is
2:43:46 changing everything we're completely
2:43:48 re-architecting our curriculum so that
2:43:50 our students are prepared for the future
2:43:52 if you don't get that answer don't go I
2:43:55 don't care how prestigious it is or just
2:43:57 you know I mean if you want to go get
2:43:59 laid and go get high sure go do that
2:44:03 all right I'm I'm behind on questions
2:44:05 I'm just gonna skip toward the bottom
2:44:07 here because I gotta I gotta go to bed
2:44:09 bye soon I got work tomorrow
2:44:13 I hate searching through my combos on
2:44:15 chat I know me too
2:44:19 um yeah they got to put a search in chat
2:44:21 GPT I've got so many prompts that I want
2:44:23 to find and I can't find them it's like
2:44:25 trying to find emails
2:44:27 it's bad
2:44:30 is chat tpd plus option the same as gpt4
2:44:33 sort of
2:44:35 get access to
2:44:38 gpt4
2:44:40 and if you enable
2:44:42 your the beta features so you go to your
2:44:45 settings
2:44:47 and in the settings there's this thing
2:44:49 called beta features and you turn them
2:44:51 on one is browse with Bing the other one
2:44:53 is plugins
2:44:55 and then for your twenty dollars your
2:44:57 Lolo price of twenty dollars you get GPT
2:45:01 for default which is the old version
2:45:04 gpd4 connected to the internet which is
2:45:06 browse with Bing and then GPT plugins
2:45:10 and what they've cleverly done browse
2:45:12 with Bing is really just a plug-in
2:45:14 so it's basically just like a hard-coded
2:45:17 plug-in and then but the plugins there's
2:45:19 a whole store of all these different
2:45:21 plugins there's like 400 of them now
2:45:24 so that's what you get for your 20 bucks
2:45:26 and I think it's worth it but you know
2:45:28 again I'm not
2:45:30 um
2:45:32 why is my sorry I'm trying to get my
2:45:35 Stan my stand is acting weird
2:45:37 um I'm not affiliated with them in any
2:45:39 way so if if the cynical bunch of you
2:45:42 are here going he's making my I bet he
2:45:45 makes ten dollars every time one of us
2:45:46 signs up for tattoo
2:45:48 I don't
2:45:50 I'm just trying to help people get up to
2:45:52 speed on this [ __ ]
2:45:57 you know he's probably making money
2:46:00 you know I bet he makes a lot of money
2:46:02 he gets people in here he sits in here
2:46:04 for hours why would he be in here for
2:46:06 hours if he's not making money
2:46:11 you see what I'm saying
2:46:12 I'm a writer with Parkinson's disease
2:46:15 thanks to AI I can write 10 times faster
2:46:17 than ever that's [ __ ] awesome ray I
2:46:19 mean not awesome that you have
2:46:20 Parkinson's but that I mean yes
2:46:29 I don't know how to I don't know I mean
2:46:31 listen it's the whole purpose of this
2:46:33 channel I guess I shouldn't get
2:46:34 frustrated
2:46:35 people are going to have their biases I
2:46:37 the the thing about a tool that that's
2:46:39 this powerful is that if people don't
2:46:41 understand it they're going to be scared
2:46:42 of it and we have
2:46:44 five Decades of Hollywood telling us
2:46:46 that the robots are going to kill us so
2:46:48 I I think it's actually kind of natural
2:46:50 that the default is oh [ __ ] that [ __ ]
2:46:56 um but it's really frustrating because
2:46:57 it's so powerful and when I hear stories
2:46:59 like like what you just describe with
2:47:01 your ability to write again I'm like
2:47:03 [ __ ] Christ the guy that can walk
2:47:05 again that they they did a Bluetooth
2:47:07 they jumped his injury in his spine and
2:47:09 he can walk again because they they
2:47:11 modeled his brain when he thought about
2:47:13 moving his legs and then they sent the
2:47:16 signal to his
2:47:17 his lower spinal cord and he could move
2:47:19 his legs
2:47:24 do I think large language models will
2:47:26 ever be up to date and process
2:47:27 information in real time I don't know I
2:47:30 don't know that it's
2:47:31 it's necessary yeah like
2:47:35 I think that
2:47:37 you know I think that the data
2:47:39 scientists are aware that if you've got
2:47:41 three-year-old data
2:47:43 as your latest data in the core model
2:47:46 that it would probably be advantageous
2:47:49 to have something a little more recent
2:47:50 than that but it's also really
2:47:51 complicated and time consuming to do
2:47:54 those base models so
2:47:56 so whether or not they fix the base
2:47:58 model or they just augment it with fine
2:48:00 tuning and just make that a bit more
2:48:02 transparent so it doesn't do [ __ ] like
2:48:04 say I don't know that
2:48:06 um
2:48:08 I think they'll fix that [ __ ] pretty
2:48:09 quickly I and by pretty quickly probably
2:48:12 within three years it
2:48:14 regardless of how they do it technically
2:48:18 every all of these large language models
2:48:20 will probably be essentially real time
2:48:22 unless you're using something that's not
2:48:24 connected to the internet
2:48:31 I'm not paying
2:48:33 you are not playing recorded session I
2:48:36 overheard you answered my question
2:48:37 thanks dude oh sure sure no this is
2:48:40 really me I'm not gpt5 and this isn't
2:48:42 recorded I'm just really behind on my
2:48:45 questions
2:48:46 um
2:48:47 you are the true and positive
2:48:49 benefit-minded well thank you I I'm
2:48:51 trying
2:48:53 I am a little bitter but I'm Gen X
2:48:56 mid Journey lens I agree the wonderful
2:48:59 wait this is wonderful but don't forget
2:49:00 to smell the flowers I agree and that's
2:49:03 why I think books books are still going
2:49:05 to be around
2:49:06 I swear you said that your 52 year old
2:49:08 52 years old in an earlier life and
2:49:11 these lives age me like two days ago I
2:49:14 was 52. now I'm 58. Jesus
2:49:17 get these lives while the gun's good
2:49:19 this aging's accelerating
2:49:22 um I think apple is careful when it
2:49:24 comes to AI apple is the main focus main
2:49:26 their main focus is Hardware they're
2:49:28 Apple's gonna come out with some serious
2:49:30 AI stuff they're they're building way
2:49:32 too much
2:49:34 um
2:49:36 silicon they're they're building way too
2:49:39 much Hardware dedicated to AI not to
2:49:42 lean into this hard they're going to do
2:49:43 it right they're not going to talk about
2:49:45 it until they they roll out something
2:49:46 that changes the game again
2:49:49 when Amazon opened chicken littles were
2:49:52 sure big bookstores were going to
2:49:53 business every new technology it's
2:49:56 everyone's talking about the Doomsday
2:49:58 scenario the problem with AI is there
2:50:00 are real plausible scenarios where the
2:50:03 Doomsday this this [ __ ] can actually get
2:50:06 that powerful but that's a whole other
2:50:08 thing is AI integrated with any CAD
2:50:11 tools yet I don't know that's go to
2:50:13 futurepedia.io there's a bunch of tools
2:50:15 there probably and I would be very
2:50:18 surprised if you know again that's not
2:50:21 something that within a year or two
2:50:24 you're gonna have major CAD programs
2:50:26 that'll be like describe a floor plan
2:50:28 now give me the plumbing Now give me the
2:50:31 electrical no I want two you know you
2:50:36 know two outlets per wall instead of
2:50:38 three whatever and it'll just redraw
2:50:41 your drawing so yeah I'm sure that's
2:50:42 coming
2:50:44 can I recommend a class or a series of
2:50:46 classes in AI with this generative AI
2:50:48 stuff right now I mean you I would say
2:50:51 you could take a workshop or two to get
2:50:53 up to speed but seriously like just go
2:50:55 there and start playing and just start
2:50:57 paying attention to it start solving
2:50:59 problems for yourself using AI is
2:51:02 probably the best way to learn about it
2:51:04 um because again things are moving too
2:51:07 fast no one is an expert right now so
2:51:09 anyone who tells you that they are
2:51:14 ignore people that say they're an expert
2:51:16 and ignore anyone who says you can make
2:51:18 money with chat to me tag
2:51:25 I will keep making that joke as long as
2:51:27 it keeps cracking me up and I'm sorry if
2:51:29 it's annoying you but this is how I roll
2:51:31 this
2:51:33 I will just drive a joke into the ground
2:51:36 um what do you think about using AI in
2:51:38 Photoshop and chat EPT to make websites
2:51:41 uh I think yes
2:51:44 I think it's
2:51:46 um
2:51:47 that's what you're going to be competing
2:51:49 against so I think you should should
2:51:51 learn how to do that be the one that is
2:51:54 the competition
2:51:56 that sounds this this that sound is
2:52:00 giving me PTSD of the yeah the modem I
2:52:03 know
2:52:04 for those of you people that don't know
2:52:06 what a modem is we our computers used to
2:52:08 have to call other computers to talk to
2:52:10 them and it just made this horrific
2:52:12 sound I always wondered like why do we
2:52:14 need to listen to that
2:52:16 like couldn't they just have a like Ping
2:52:18 your honor ain't it didn't go like why
2:52:21 do we need to listen to the handshake
2:52:26 what college is teaching this I don't
2:52:28 know all right I'm I'm down toward the
2:52:30 bottom here thanks for doing this
2:52:31 thoughts on how to draft a business
2:52:32 process and policy document whole cloth
2:52:36 so when it comes to creating long
2:52:38 documents like that break it up into
2:52:40 chunks and start with providing the tool
2:52:44 context so if you want to do like a big
2:52:46 business plan what did you say
2:52:48 business process and policy documents
2:52:51 start very high level here's what my
2:52:53 business is about here's our values
2:52:56 um
2:52:57 here's the here are the processes I've
2:52:59 got now
2:53:01 um
2:53:02 these ones are working well these ones
2:53:04 aren't working well could you give me
2:53:07 some ideas of modern business practices
2:53:10 that I could incorporate to help address
2:53:12 these issues and let it just give you a
2:53:15 little list and I say okay that's great
2:53:16 or no you missed the mark you know I I
2:53:20 need help in these areas oh I'm sorry
2:53:22 here's these areas and then and then
2:53:24 just start to expand out from those
2:53:26 things giving it the context like that
2:53:29 and building it in chunks will
2:53:30 dramatically improve the things you get
2:53:33 down below if you try to write
2:53:36 a master prompt to get everything you
2:53:38 want
2:53:39 you'll just spend more time tweaking the
2:53:42 master prompt and it gets really
2:53:44 frustrating so break your things up into
2:53:46 into little chunks will help
2:53:50 all right uh Apple's innovating
2:53:54 you're on the precipice of Gen X almost
2:53:56 a boomer oh go [ __ ] yourself 1965 is
2:53:59 solid Gen X
2:54:01 don't you dare call me a boomer
2:54:05 laughs
2:54:07 just it's just the the trolls here the
2:54:10 trolling you're nearly a boomer you look
2:54:13 like Rosie O'Donnell you're nearly a
2:54:15 boomer come on people
2:54:19 I was scrolling Tick Tock I thought it
2:54:21 was Alec Baldwin but he got fat that was
2:54:23 one I got
2:54:26 all right I gotta go I just made myself
2:54:29 bitter
2:54:31 it's time it's time for me to leave and
2:54:34 I'm gonna boot my uh they're gonna boot
2:54:36 my ass off here I've been booted off
2:54:37 these lives twice now for seven days
2:54:39 each they sent me in the corner for
2:54:41 being a bad boy
2:54:43 all right
2:54:45 um awesome thank you should AI be
2:54:47 regulated it should but we should get
2:54:49 some [ __ ] politicians that understand
2:54:50 how Wi-Fi works
2:54:53 come on understand how Wi-Fi works step
2:54:55 two regulate AI
2:54:58 maybe there's like 30 steps in between
2:55:00 there they should get some smart people
2:55:02 in in the halls of uh Congress to
2:55:05 understand this [ __ ] but yeah it's gonna
2:55:07 need to be regulated how good is poke
2:55:09 co-pilot very good
2:55:11 all right I gotta go
2:55:16 64. I identify as a gen xer all right
2:55:19 you gen xers and you Boomers bye Boomer
2:55:22 [ __ ] you peace out everybody have a good
2:55:24 night