aiLL 10/22/23 - Power of AI & Offers Tips During This Time of Massive Change

Description

Chapters

Transcript

0:19 welcome welcome welcome Be not Afraid
0:21 what shaking what's
0:24 happening what's happening what's going
0:27 down let's see do I have everything set
0:28 up here I think do but if not we'll wing
0:32 it love the timer awesome all right
0:35 we'll keep doing the
0:37 timer I did it from my car today I'm
0:40 like I got to do a
0:41 timer and I start recording I pull it
0:45 beside some cops they're on their phone
0:52 too oh man it's happening Brandon
0:55 jet Paulie Z what's
0:59 happening Lily good day to you Sherry D
1:04 what's
1:05 shaking what's going down everybody
1:08 Tobias checking in from the hot tub
1:10 Tobias just likes to rub it in he's like
1:12 yeah you know just live the life just
1:15 eat nachos salad and drink wine and
1:20 watch the AI learning lab from a hot
1:24 tub glint Copperhead calling the uh
1:27 Irregulars to order very nice not Lama
1:32 Curry we don't know about Tobias he's a
1:34 little he's I think I think Tobias has
1:37 some Secrets but you know we'll see
1:40 we'll
1:42 see welcome everyone wow lots of folks
1:45 in here already so do me a favor if you
1:48 would share the live with your friends
1:50 let them know how weird you
1:52 are you're not weird you're just
1:55 irregular like all of
1:57 us and we'll get started here we'll jump
2:00 in we'll talk about some AI
2:03 stuff we will I don't know we'll do
2:06 whatever we do sometimes it goes funny
2:09 sometimes it goes philosophical most
2:11 times it goes like holy crap there's a
2:12 lot
2:15 happening I did see today that open AI
2:18 cancelled they had a project called
2:20 araus which was supposed to be like some
2:23 super you know sentient gp5 rumored
2:29 going to change the world thing they
2:30 canceled the project apparently it
2:32 wasn't working wasn't working like they
2:34 they wanted to so you know this is
2:37 technology so it's good enough they
2:39 they're already changing the world there
2:41 if if if if open AI does nothing more
2:44 than what they've done to date with chat
2:45 gbt we're we're fine like the uh the the
2:49 revolution has been kicked off for us um
2:53 welcome my name is Kyle Shannon this is
2:55 the AI learning
2:56 lab um we'll talk about anything you
2:59 want to talk about about in here it's
3:00 generally generative AI stuff future of
3:03 work um using these crazy tools trying
3:06 to keep up with these crazy tools uh
3:08 answering any questions I can so if you
3:10 have questions just pop them in the
3:11 comments below uh we've got folks in
3:13 here that mod and and and do things like
3:16 that so if you have basic questions they
3:17 can answer those and and Elevate elevate
3:20 your uh your questions up if I missed
3:23 them um depending on how many rabbit
3:25 holes I run down sometimes that can
3:27 happen where you know I'll I'll I'll
3:30 down a rabbit hole and and lose a bunch
3:32 of things arachis
3:35 phobia that's pretty good um what else
3:40 so I think that's it I think that's it
3:45 what else we got got the Cashes Clay
3:48 bosot shirt on today for there seems to
3:51 be some fans of tell us this t-shirt
3:53 story so if you want to know the T-shirt
3:56 story that's
3:57 [Laughter]
3:59 it for you youngsters Cash's clay was
4:02 the name of a famous fighter named
4:04 Muhammad Ali before he changed his name
4:06 to Muhammad Ali if you don't know that
4:08 shame on
4:10 you look at your history books all right
4:13 let's get let's get rolling here um okay
4:17 so if you're new to the channel if you
4:20 if you just been been uh you know
4:22 stumbling along through Tik Tock you
4:24 stumbled on this you're like AI learning
4:26 lab I want to go learn about AI if you
4:28 haven't played with chat jbt yet go you
4:30 need to do that you need to do that and
4:33 not cuz it's like it's cool but because
4:37 it's actually important it's it's
4:38 important for everyone to at least
4:40 understand what the hell is going on um
4:43 and what these things are and why are
4:45 people talking about this stuff why why
4:47 what's the big deal you know what I
4:49 would like to know what's the big deal
4:51 you know I see all these videos that are
4:52 like you can make money with ch is that
4:55 true can you really make money with chat
4:56 jpt I watch them and I click on them and
4:58 they tell me you can make money
5:01 and it just seems too good to be true
5:03 you know what I
5:05 mean so rather than worry about those
5:08 buttheads just go there go to chat.
5:11 open.com that's the official chat GPT
5:13 website these are the guys that started
5:15 all this mess well there's lots of
5:17 companies that started this Google
5:19 actually started it when they invented
5:20 the transformer in 2017 but then they
5:22 kind of ignored it because they thought
5:23 it would threaten their business model
5:25 as my supposition but these guys took
5:28 that little Transformer thing and ran
5:29 with it
5:30 and they did gpt1 and then they did gpt2
5:34 and then in 2020 they did gpt3 and that
5:36 was the one where people were like oh
5:37 this thing looks pretty interesting and
5:40 then November 30th of last year they
5:43 launched jet GPT and that's it and it's
5:45 gotten better and better and better and
5:47 better and better and better and in the
5:50 past 3 weeks it's gotten even more
5:52 betterer so that that chat GPT at this
5:55 point is the one if you're going to just
5:57 play with one of these things if you're
5:59 just going to explore one thing start
6:00 with that one um Bard Claude and Bing
6:05 are also kind of in the top four they're
6:07 they're kind of the next tier down just
6:09 in terms of
6:10 quality although Google is rumored to be
6:14 releasing a thing called Gemini soon
6:17 which is apparently as good or better
6:19 than gp4 I'll believe it when I see it
6:22 um Claude is is good Claude used to only
6:27 be available in the US and Canada or the
6:30 us or the UK they've now rolled Claude
6:33 out to 95 countries so if you couldn't
6:36 get to Claude before you likely can now
6:39 Hey Ron what's
6:40 happening um all the time you're an
6:42 irregular now Be not Afraid oh be not
6:44 afraid's definitely an irregular and if
6:46 be not afraid if you're on the
6:48 um Discord server just let me know and
6:51 I'll I'll give you the irregular roll
6:53 over there
6:56 um um what was I gonna say oh yeah so
6:59 Claude Claude claude's A a big deal and
7:02 and now is available in 95 countries
7:05 it's um the big deal about Claud is that
7:08 it can take really big prompts so if
7:10 let's say you're working on the Great
7:12 American novel but you're kind of a
7:14 crappy writer and and you want some
7:16 editing help with Claude you could
7:18 upload your whole novel or or just paste
7:20 it in there and then say help me edit
7:23 this and just go chapter by chapter and
7:25 have it give you suggestions so um so
7:28 it's a big deal um those of those so
7:32 here's another pile that's what I call
7:33 the kids table so Pi is is by a company
7:38 called
7:39 inflection they um they just raised $1.3
7:42 billion to buy
7:44 22,000 big gpus um to to make their
7:49 thing
7:52 gooder and but what's cool about Pi is
7:54 download the iOS app and put it in
7:56 conversation mode there's a little
7:58 telephone at the bottom in the chat
7:59 window window click on that and just
8:01 start talking to it and uh it's it's
8:04 it's pretty
8:05 slick um and then perplexity doai is
8:09 worth checking out PO is worth checking
8:11 out um here are some of the image
8:13 generation tools dolly3 mid Journey
8:16 Leonardo and Adobe go ahead and
8:18 screenshot that um the nice thing about
8:21 do is it's now incorporated into GPT 4
8:25 if you've got a GPT Plus
8:27 subscription you should have access to
8:30 DOI now it seems like it's rolled out
8:31 pretty wide at this point if you don't
8:33 have access to it it's probably coming
8:34 very very shortly um I can demo that
8:37 it's quite
8:39 remarkable
8:43 um Sher D my boyfriend who's your
8:46 boyfriend piie or
8:49 dolly3 uh Leonardo is built on top of
8:53 stable defusion it's a nice web front
8:55 end adobe's really good if you're in a
8:56 corporate situation and then here are
8:59 some video things Runway ml hey Jen
9:02 blockade Labs did CSM is I just threw
9:06 that in there as a placeholder it's a 2D
9:08 image to 3D object thing it's not great
9:11 right now uh it may get better um and
9:14 then
9:16 um if you're trying to figure out where
9:18 are all the apps uh there's directories
9:21 so Matt wolf has futur
9:28 pedia.com David Shapiro and Paul rer I
9:31 think are all worth following so one
9:33 useful thing is Ethan mik's blog he's a
9:36 Wharton business school Professor he
9:39 just put out a really interesting piece
9:41 today um David Shapiro will blow your
9:44 mind and Paul rater if you're in
9:46 business you should absolutely be
9:47 following Paul rater and the marketing
9:50 AI Institute uh it's also worth dude I
9:54 still have comp you
9:57 serve this is listen this early AI stuff
10:01 we we are definitely right now we're
10:03 going to look back at these days and you
10:05 know one of these companies like Claude
10:08 is probably comp you serve and open AI
10:10 is like Netscape and it's definitely
10:14 going to be like that it's definitely
10:16 going to be like that um the other thing
10:19 I I'll talk tonight a fair amount I'll
10:22 I'll sort of refer to people over on the
10:24 Discord the Discord is the AI Salon
10:28 Discord um so so so it's a community of
10:31 people we meet every other Tuesday so
10:33 this coming Tuesday is the next Salon
10:35 meeting so you can sign up for the
10:37 Meetup to get notifications for that and
10:40 then um there's also an invite for the
10:43 Discord so if you want to join that
10:44 Community that's that's what's going on
10:46 there all right so that's that let's get
10:48 started we we'll we'll Jump Right In
10:50 we'll we'll answer some questions a few
10:52 weeks ago someone brought me brought up
10:55 perplexity what's your knowledge about
10:57 it um my know
11:01 my knowledge about it is that my
11:03 knowledge about it is
11:05 low I know it's good I I haven't used it
11:08 a ton but what perplexity does really
11:11 well is it provides citations so so you
11:15 can limit your search to just say just
11:17 look on Wikipedia or just look on you
11:19 know certain sites and then when it
11:21 gives you an answer back it brings back
11:23 um links to the the places that it
11:26 searched to get you that answer um so
11:29 it's good for research it's good for
11:30 things like that people that use it a
11:32 lot swear by it and it's one of those
11:34 sites that I keep meaning to like spend
11:37 the couple of hours just getting my head
11:39 around it I haven't done that yet so um
11:42 I think it's absolutely worth worth
11:44 getting in there Saturday morning ads
11:46 got inspired by your system last night
11:48 and Ed chat GPT to start making a word
11:51 game with python
11:54 awesome
11:56 awesome that's the whole point of the
11:58 channel is just start playing it doesn't
12:03 matter what you do it literally doesn't
12:05 matter what you do because these tools
12:07 are changing on a weekly basis they're
12:10 evolving they're getting more
12:12 powerful
12:13 um at at this point um gp4 with all of
12:18 its it's got all these new
12:21 multimodal tools the these different
12:25 modes
12:26 um and and it's it's really quite
12:29 something so um yeah so start playing
12:32 start making word games if you had no
12:35 money and wanted to start a business
12:36 from A to Z where would you
12:39 start
12:42 H well a service business
12:46 is a service business requires no
12:50 investment and no infrastructure right a
12:52 service business requires your brain and
12:55 some ability to execute [ __ ]
12:59 so here's the really good news your
13:02 ability to execute [ __ ] just got
13:05 dramatically Amplified right but I'm not
13:08 good at Art Direction well now you are
13:10 but I'm not good at copyrighting well
13:12 now you
13:13 are but I'm not good at business
13:15 strategy now you are so what I would say
13:17 is take take where you've got some
13:19 expertise and where you know there's a
13:21 pain
13:22 point and you know it could be
13:27 around you know maybe maybe you you
13:30 understand an industry that's
13:32 particularly afraid of AI right now and
13:36 you could start a business around
13:37 educating people on AI there's going to
13:39 be a lot of that over the next 18 months
13:41 holy crap there is so much to learn
13:44 there is so much to try to keep up with
13:47 and a
13:49 shockingly
13:51 shockingly lown number of Business
13:54 Leaders are up to speed on this at
13:57 all and so
14:00 that's kind of what this channel is it's
14:02 just saying hey this shit's coming do
14:05 something I mean I haven't this is not a
14:06 this is not a service business but but
14:09 could be I'm I'm the CEO of a of a video
14:11 technology company so I've got a I've
14:13 got a day gig but I in the mid 90s I
14:16 started agency.com which was one of the
14:18 very first digital
14:20 agencies and what's powerful about
14:23 starting a service business if you can
14:25 do something that people don't know how
14:27 to do like back then it was we can make
14:30 you a
14:32 website and we happen to be good enough
14:35 at understanding business needs to be
14:37 able to you know walk
14:39 into Fortune 500 companies as a bunch of
14:42 you know 30s something cocky little
14:45 pricks and go we can do
14:47 this they're like
14:51 okay you know we charged them a lot of
14:53 money and they pay you money and then
14:55 you go do the work and if you don't
14:57 suck then they give you more money the
15:00 next time and more money the next time
15:02 and then if you do suck you go find more
15:04 clients and so that's probably where I'd
15:07 start so could be doing workshops it
15:09 could be doing education it could be
15:10 doing problem solving Kyle what is AI
15:14 and chat GPT generating about the
15:16 probability of a third world
15:19 war um well so
15:23 nothing um so
15:28 zindel
15:31 chat GPT and these generative AI tools
15:34 are just tools it's so so I I could also
15:37 answer your your your question would be
15:40 sort
15:41 of
15:42 um like like a parallel question would
15:45 be what's Photoshop doing about creating
15:49 images for World War II nothing like
15:51 it's
15:52 Photoshop these tools are not
15:54 autonomously taking action on their own
15:57 so there's there's a couple of
15:58 misconceptions about Ai and and I think
16:00 probably one of the biggest ones is that
16:02 the robots are blank the robots are
16:07 predicting World War III the robots are
16:09 going to take us out the robots are no
16:12 these things right now think of of chat
16:16 GPT is kind of like a super fancy
16:20 calculator like like like from a
16:23 technical perspective it's almost
16:25 exactly what it is you type in your
16:27 prompt and as you're typing in your
16:29 prompt when you submit it it turns it
16:31 into a mathematical
16:33 probability of what's the next most
16:35 likely word from this prompt to respond
16:39 to that then takes all that and says
16:42 what's the next most likely word after
16:43 that so when it gives you this big long
16:45 response that looks like it's some
16:46 sensient being it's got It's got no
16:50 understanding of what it's written which
16:52 is why there's a thing called
16:53 hallucination where these large language
16:55 models will just [ __ ] lie to your
16:57 face they're like here here's your
16:58 answer and you're like uh pretty sure
17:00 that's wrong buddy oh sorry about that
17:02 here's your answer and it'll do it wrong
17:04 again um why is that cuz it it actually
17:07 has no idea what it's writing now
17:10 there's ways that you can prompt it that
17:12 make it do less than that and
17:14 when when you get your prompting right
17:17 what these things do is actually quite
17:20 remarkable and they seem sensient and
17:23 they seem human and they seem like they
17:25 could
17:26 predict you know the stuff you're
17:29 talking about but what I would say is I
17:32 I I'll toss it back on you go ask it go
17:36 ask chat GT what it thinks say here's
17:38 what's going on in the world give me
17:39 your thoughts on this now what you'll
17:41 find with with a lot of the commercial
17:44 things like chat GPT and Bing especially
17:47 um but even Bard and probably Pi is
17:50 they've got what are called Safety guard
17:51 rails so if you go in there and try to
17:54 do something it that it could perceive
17:56 as hurting other people it'll go yeah
17:58 we're not going to talk about that I
18:00 don't like that that's not safe so
18:03 they're a bit of a pain in the ass that
18:04 way
18:07 um wait it's not
18:09 prediction it's probability is not
18:12 prediction it's pro pro well the the way
18:14 it works is it's a probability engine
18:16 but it's it's not a probability like the
18:18 probability of World War I it's a
18:20 probability engine like what's the next
18:22 word that answers the question that you
18:25 asked what that that answers the prompt
18:27 that you asked so that's what I mean by
18:28 probability engine it's not it it it is
18:32 not a it is not a prediction machine but
18:34 you can get it to act like one Go Tell
18:35 it it's a it's a a futurist uh that
18:38 specializes in in uh
18:41 geopolitical uh
18:43 Intrigue and have it have it do
18:46 something for you have it predict
18:47 something for you it will it'll predict
18:49 it um is that prediction based in
18:52 anything no not
18:54 really all
18:56 right let's see let's go go get some
18:59 let's go get some questions here get
19:01 some questions
19:05 going I appreciate some teachers now
19:08 embracing a AI but encouraging the kids
19:10 to question the critique and and
19:12 critique the AI Anthony y absolutely I I
19:16 think one of the most important things
19:18 right now for educators there there's
19:20 two things one is just don't ignore it
19:22 and don't pretend it's not coming don't
19:24 pretend it's not here right so if you're
19:27 an educator
19:29 especially if you're a [ __ ]
19:31 administrator take down the chat GPT
19:34 detectors and the blocks they don't work
19:37 and we're moving into a world where
19:39 these tools are going to be ubiquitous
19:40 they're going to be everywhere so take
19:42 down the the
19:44 detectors and start getting into
19:46 critical
19:47 thinking there's a there's a a Tik
19:50 tocker named I forget her name it's it's
19:52 Gibson like I am Gibson or the real
19:54 Gibson or something like that she's a
19:56 like a 12th grade high school teacher
19:59 and she brought chat GPT into her
20:02 classroom like last January and she was
20:05 totally freaked out by it and and like
20:07 now she's doing things like you know the
20:09 the kids for their semester this
20:11 semester they've got they have to write
20:13 one
20:14 paper by hand and then they have to
20:17 write one paper with chat GPT and then
20:20 there's different rules of engagement
20:21 the one they have to write with chat GPT
20:23 they then have to critique it and say
20:25 what did it get right what did it get
20:26 wrong what would you do different
20:29 and and she's she's working with them to
20:31 help help them understand where are the
20:33 boundaries with these
20:35 tools and that's I I feel like that's
20:37 the only thing all of us can do right
20:39 now is just kind of like like you know
20:42 what it is right now we're we're all in
20:45 a dark room and there's all these weird
20:48 noises and there's walls and we don't
20:50 know if the walls are closing in or
20:51 expanding we don't know what the shape
20:53 of this thing is and we're all just kind
20:55 of feeling around in the dark trying to
20:57 figure out what the [ __ ] is this stuff
21:01 right what's machine learning compare
21:04 contrast
21:05 AI um that's I I would say I would say
21:09 chat gbt is going to be much better at
21:11 that than I am um here's the deal with
21:15 is chat GPT AI or not no it's not it's
21:18 machine learn it's a machine learning
21:19 tool um it will never be called machine
21:23 learning it's always going to be called
21:25 AI why because Hollywood called it AI 30
21:29 40 years ago and it's all we talk about
21:32 now and machine learning sounds like
21:35 Technical and we got to deal with it and
21:36 AI just sounds like oh it's the AI so
21:39 it's always going to be called AI it is
21:41 technically not AI um AGI is coming
21:44 artificial general intelligence if you
21:46 go listen to David Shapiro who I I
21:48 recommended earlier um he believes that
21:52 AGI will be here within 12
21:54 months the amazing thing is that it's
21:57 allowing certain kids with certain
21:58 learning difficulties to adapt Anthony
22:00 an the I think I think it it that's the
22:05 that's one of the things I'm most
22:06 excited about is that people with
22:08 disabilities and people with access
22:10 issues and people um just with with
22:13 different languages just communication
22:15 barriers in general those communication
22:17 barriers are just going to start
22:19 dropping and
22:21 and one of the things we talk about a
22:23 lot in here and one of the things I've
22:25 seen there's some conversations going on
22:27 in the AI salon distri ORD
22:31 about these
22:33 tools very very quickly
22:36 take just an idea in your head and bring
22:39 it to
22:40 life and so imagine that if if you're if
22:44 you've got some neurological disorder or
22:46 disability or some way that that you're
22:49 you know you're cognitively there but
22:51 you've got some Block in expressing
22:53 yourself these tools are going to
22:56 unblock that and free all that that
22:58 stuff up dragonfly Alchemy my daughter's
23:01 been playing with with character AI
23:03 interesting lots of autistic characters
23:07 interesting she feels she feels
23:09 celebrated yeah perfect
23:12 perfect I like yeah exactly and and you
23:17 know character AI is one of those things
23:19 where they're creating all these
23:20 different personalities that you can
23:21 interact with po also does that and you
23:24 can also create your own bot at po so if
23:27 you go to po.com
23:29 you can play with other people's Bots
23:30 there's thousands of them and they have
23:32 a bot maker there it's really easy you
23:34 basically just pick a model and you
23:36 write a prompt and see so like I made
23:39 one called I am Shakespeare and then
23:41 last night I modified it to say I am
23:43 Kramer from from
23:46 Seinfeld and you could just create a bot
23:48 that acts like a certain thing so it
23:50 could be something that understands
23:52 autism or you you know somehow responds
23:55 in a way that resonates in a very
23:57 specific way
24:00 um I I you know I I think the idea of
24:04 digital
24:06 companionship and not kind of the creepy
24:09 you're having sex with a robot like you
24:11 know there's going to be all that but
24:14 but like get past the creepy part and
24:17 then go to how many
24:22 people don't have people they can
24:24 communicate with can't communicate don't
24:27 know what to say or how to say it or
24:30 like all that stuff there's going to be
24:32 massive amounts of people that are
24:34 supported by these things and if you if
24:37 you want to experience this go download
24:39 the PI app
24:41 Pi if you go to the Apple App Store I
24:44 don't think they have one for um Android
24:48 yet but they'll have one but pi. a that
24:54 thing is a conversational bot in fact
24:58 here here let me show you
25:00 so and you can you can now do this with
25:03 chat the chat GPT app as well all right
25:06 so you go to
25:08 Pi and it you know it loads here's some
25:11 conversations you can have but in the
25:13 bottom in the bottom corner there's a
25:15 little phone icon and if I click that it
25:18 turns into a conversational
25:27 bot
25:31 [Music]
25:34 come on my internet must be slow am I
25:37 glitching am I am I hang on let me quit
25:40 this and restart it that seems a little
25:50 slow I'm neurod diverg and have ADHD and
25:53 I'd like to better understand the autism
25:57 spectrum can you help
26:04 absolutely neurod Divergence is a broad
26:07 term that includes many conditions and
26:08 experiences including ADHD and
26:12 autism understanding the autism spectrum
26:14 can be a great way to better understand
26:16 neurode Divergence as a whole here's a
26:18 quick overview of autism autism is a
26:21 neuro so and and you
26:24 can like like that particular bot is
26:27 designed
26:28 um to be very
26:30 conversational Pi is actually a really
26:32 crappy tool like if you want to get
26:34 something done go to chat GPT but if you
26:36 want to have a
26:38 conversation oh you're good Kyle kids
26:40 downloading again no I think I'm good I
26:42 think the I think just the appid crapped
26:44 out
26:46 um you know if you want to have a
26:48 conversation about you know a very
26:50 specific topic Pi is Pi can have it I've
26:55 I've asked it about books like I was
26:57 driving and I someone had told me about
26:58 a book and I said oh can you tell me
27:00 what this book is about and then um and
27:04 then it it it talked about the book and
27:06 I I said well can we explore that
27:07 subject and then it it just we just went
27:09 into a whole conversation not about the
27:11 specific book but about that whole topic
27:14 and so you you can and and that's just a
27:16 generic conversational but you these
27:18 things are also going to get quite
27:20 specific and you'll be able to to make
27:22 them quite specific in fact
27:25 um so the chat G
27:28 app um
27:31 now has a couple of really interesting
27:33 things in it so if I go new chat
27:35 here uh let's see do I have custom
27:38 instructions in here
27:41 settings yeah so in settings in the app
27:46 there's a thing called custom
27:48 instructions and so I can set up chat
27:52 GPT so that I want it to act like a a
27:55 particular kind of personality or I
27:57 wanted to respond in a certain way and
28:01 what what it's got now is it's got a
28:03 similar mode as Pi does so see these
28:05 little headphones here I put it into
28:07 headphone
28:15 mode can you help me understand neuro
28:19 Divergence or what I like to call neuros
28:26 spiciness of course I'd love to help
28:30 explain neurod Divergence sometimes
28:33 playfully referred to as neuros
28:34 spiciness it's a concept where
28:36 neurological differences are chat GPT is
28:39 so
28:41 supportive I like your neuros
28:45 spiciness um so so with chat GPT alone
28:48 just with the chat GPT
28:50 app using custom instructions and using
28:53 that conversational mode you could have
28:56 it behave in a certain way could could
28:58 you could whoever just pinned that last
29:00 comment pin it back I was just about to
29:02 um respond to it and it disappeared um
29:05 the thing about the uh the state
29:08 representative or whatever it
29:09 was
29:12 um so so yeah so the so like like that's
29:16 something to play with like like is it
29:19 going to be a replacement for a
29:21 therapist
29:22 no is it anywhere near that right now no
29:27 but could it help you learn something so
29:29 I've been working with these with my
29:31 campaign for state representative here
29:33 in Florida my goal is
29:38 to put the serious and confusing topics
29:41 in a more digestible way awesome great
29:44 great I I'll tell you Anthony one of one
29:47 of the things that I would suggest
29:49 is
29:52 um like the purpose of this channel is
29:54 really to demystify this stuff and and I
29:57 think that
29:58 to what I found doesn't work is there
30:02 there's a lot of emotion around AI right
30:04 now right it's dangerous the robots are
30:06 going to kill us it's going to take my
30:08 job there's there's a lot of fear around
30:10 it and and part of that is just 30 40 50
30:13 years of Hollywood telling us that the
30:16 robots are going to kill us right there
30:18 there's also a thing that there's a very
30:20 real
30:23 um there's a very real well founded
30:28 misconception that in order to do AI
30:30 you've got to be super technical right
30:34 because prior to November 30th last year
30:38 the launch of chat
30:40 GPT prior to that you kind of had to be
30:42 a mathematician or an engineer or a
30:46 researcher um because the tools were
30:48 very technical and it's I mean it is
30:50 incredibly mathematically heavy uh I'm
30:53 not a mathematician I have a degree in
30:55 acting yeah
31:00 um so one of the things that I found is
31:03 that the minute it the minute you try to
31:06 tell someone oh AI is good AI is
31:08 important their eyes glass over and they
31:10 either get angry or they get afraid or
31:12 they get
31:14 dismissive almost all the time when I
31:17 experience someone that that is one of
31:19 those three mad afraid or dismissive
31:22 they're all they're all flavors of the
31:24 same thing they've they've not tried it
31:26 and so so so one of the things that I
31:28 try to do in this channel is just try it
31:30 just play so if you're trying to um
31:33 demystify it for people you know if if
31:36 you're talking to someone say hey what
31:37 are you into oh I'm into um a bridge
31:41 infrastructure great let's go let's go
31:42 to chat GPT and see what it knows about
31:44 Bridge infrastructure and then have them
31:46 seed you and just make it relevant to
31:48 them and let them see in real time how
31:53 deep these tools are and how how how
31:55 quick they are how deep they are
31:58 how profoundly different they are than a
32:01 Google search Pi is my therapist oh [ __ ]
32:03 pop that back up I was just about I was
32:05 just reading that pop that back up again
32:07 whoever pinned that that looked like a
32:08 good
32:10 one
32:11 um that's awesome though um p is my
32:15 therapist he's free I probably owe him a
32:17 dinner or a charger that's awesome Genie
32:21 and a bottle uh Genie in a Bottle 10 out
32:23 of 10 on the
32:25 name awesome awesome awesome
32:34 awesome I can't type today I can't type
32:37 any
32:38 day um Kyle I feel like I should be re
32:42 receiving college credits credits for
32:43 attending your lives every night I got
32:45 to get
32:47 accredited here here Ted looks at life
32:50 made an image for me of this
32:54 one Kyle
32:56 stock
32:58 this is right now it starts as a Tik Tok
33:00 Channel but you know my goal is the
33:02 Vegas sphere we we'll do these in the
33:05 Vegas sphere and everyone gets
33:08 University
33:09 [Laughter]
33:12 Credit what did you learn I don't know
33:15 but I keep going
33:16 [Laughter]
33:20 back oh
33:22 man AI is the
33:25 future what's funny about that one is
33:28 the the future is here it's that's a
33:30 thing that I'm
33:31 increasingly kind of smacked in the face
33:33 with with this AI stuff is that it's not
33:36 coming it's
33:38 here and it's what what what ends up
33:40 happening is the the the biggest
33:43 converts of this stuff are very often
33:47 the ones that were most resistant to it
33:49 and were the biggest
33:50 attractors because people that are very
33:54 dismissive of
33:56 it for the most part haven't experienced
33:59 it or they might be very educated in the
34:03 mathematics and Engineering side of it
34:05 and so they understand it too much and
34:07 they're like well it's just a prediction
34:08 engine and so they dismiss it that
34:11 way and then you say but try
34:14 it yeah yeah
34:17 whatever what are you interested in I'm
34:19 interested in this boom Oh that's
34:23 interesting and then before you know it
34:25 they're like did you know this thing was
34:28 amazing yeah what I'm trying to tell you
34:32 it's it like I the only thing that I'm I
34:34 am absolutely confident on is that
34:36 there's no way to explain how profound
34:39 It Is by explaining how profound it is
34:42 it's only when someone has what I call
34:44 their Kevin mallister moment this
34:48 one oh be not afraid you made that image
34:50 awesome yeah so I go to the I go to the
34:53 showand tell channel in the Discord and
34:55 I I dump them into my little image
34:56 Library here that's
34:58 awesome and what's what's been really
35:00 cool here is is a lot of the people that
35:02 are in here on a regular basis that we
35:05 nicknam they nickname themselves the
35:06 Irregulars which I think is [ __ ]
35:08 hilarious um as we're talking about
35:12 stuff they're using the new um do within
35:15 chat GPT to make these images and and so
35:17 just in real time I'm talking about
35:20 stuff I'll make a funny joke or
35:22 something like that and all of a sudden
35:23 there's these images showing up that are
35:24 that and and there's there's ation and
35:28 how did you do that and what was the
35:29 prompt and that da and and there's so
35:31 it's like real time learning real time
35:35 self-expression um real
35:38 time
35:41 um capability expanding right like Sher
35:45 D who's in here has has been a nurse all
35:48 her life and she's just diving in to the
35:52 world of of generative art and she's
35:54 making beautiful stunning images and
35:56 she's like I don't know if she's
35:58 Reinventing herself but she's certainly
35:59 exploring a whole new Avenue that's
36:01 that's not part of her core skill set
36:05 and it looks like she's having a blast I
36:07 assume you're having a blast the work is
36:10 amazing
36:12 so that's kind of what's going on here
36:15 that's the whole idea giant llamas
36:16 running around with roses in their
36:20 teeth and rainbows coming from behind
36:22 them oh that was oh wait I've got that
36:24 image oh I don't think I moved it over
36:26 here though
36:28 yeah there was a llama yesterday with
36:30 there was a llama running down a road
36:31 with like rainbow farts coming coming
36:33 out of its butt I assume that was that
36:36 one on a giant 8K
36:41 screen oh that's
36:43 [Laughter]
36:45 awesome all right we'll leave we'll
36:46 leave this one up Be not
36:49 Afraid this is a good one here look at
36:51 this one isn't that
36:55 cool it's very um it's very big isn't it
36:59 from the movie Big use chat GPT for
37:02 writing College papers oh using chat GPT
37:04 for writing College papers five-page in
37:06 an intensive course
37:09 20page Renee Steinman recommendations um
37:14 so my recommendation Renee would be
37:20 um I don't know how much you've been
37:22 using it but my experience with chat GPT
37:26 is don't think that it's Google so so so
37:30 and here's what I mean by that Google
37:32 kind of trained us just give us the
37:34 minimal amount of prompt possible just
37:36 give us the keywords and we'll do all
37:38 the work right and I and I think I think
37:41 some of the some of the early mistakes
37:44 when people are using these large
37:45 language models is they're like write me
37:47 a paper about dogs whatever right and
37:50 then what they get back is this very
37:52 generic often poorly written poorly
37:55 structured poorly researched you know
37:57 highly inaccurate paper that looks like
38:00 a paper but when you read it it's
38:01 actually a piece of [ __ ] so what I would
38:04 say is um here let me you know what let
38:06 me let me go let me go play a little and
38:10 and we'll we'll I'll show you how how I
38:13 would how I would approach
38:16 it and I'll just do it with um chat gbd
38:20 3.5 so so um I I firmly believe that gp4
38:26 is is the chat GPT plus is worth the 20
38:30 bucks if you're if you're a poor college
38:32 student you might not be able to afford
38:33 that if you can I would do it cuz the
38:35 the writing for GPT 4 is much better and
38:39 in fact if you can't afford it you could
38:41 also use Bing chat. bing.com so that's
38:45 Microsoft's uh version of of chat gp4
38:48 it's not as good as open ai's version of
38:51 it even though it's based on the same
38:53 model um but it's it's it's good writing
38:56 so so if you can't afford 20 bucks go
38:58 use Bing but but what I would do is um
39:03 chat jpt works much much
39:06 better when you give it lots of context
39:10 right so with Google it's like you know
39:11 go find me a website that da and then
39:14 it's it's like that's
39:16 it with chat jpt it's kind of like think
39:19 of it like this IM imagine that we've
39:22 all been given an idiot savant as a as
39:26 an
39:27 intern and that's what chat gbt
39:30 is the Savant part is it knows [ __ ]
39:34 everything it's been trained up on all
39:36 of the public internet so so or as much
39:39 as they could get their hands on right
39:41 and so it kind of knows everything the
39:44 idiot part is that it doesn't know how
39:47 to do anything it doesn't have any
39:49 context so you you have to give it all
39:51 the context so what I would do is I
39:52 would start out with something like um I
39:55 don't know let's say we're going to do
39:57 we're going to do a um a a a paper on um
40:04 the the history of Gummy Gummy
40:07 Bears I want you to act as a
40:10 confectionary
40:12 historian and assist me
40:17 in understanding the history of gummy
40:22 bears and how we got to where we are
40:25 today in the gummy tradition
40:30 period I have to write a five-page
40:34 paper and I love gummy bears so much
40:38 that I want my paper to be about that in
40:42 particular I like the
40:46 hobo gummy bear
40:52 company it's spelled that
40:55 wrong hair
41:03 IO but I can only imagine that there
41:07 are lots of other interesting
41:11 companies that got us to where we are
41:13 today
41:21 period my professor is a
41:24 stickler for his hisorical details and
41:28 accuracy so I also want you to include
41:32 anecdotal stories that might bring my
41:34 paper to life period before we write
41:39 anything
41:41 can you help me understand how we might
41:45 structure such a
41:48 paper
41:50 so giving chat GPT the chance to a
41:55 understand what the hell you're talking
41:56 talking about is really important and
41:59 don't just dive into write me a paper on
42:02 like like have it like like this outline
42:05 let's let's see what it
42:07 did um certainly writing a five-page
42:10 paper on the history of gummy bears is a
42:11 great idea the evolution of gummy bears
42:13 a sweet Journey Through History it's not
42:16 bad CH bt4 would probably be a little
42:18 better but it's not bad abstract a brief
42:21 summary of the paper's main points
42:22 introduction hook your readers with an
42:24 in interesting fact or an OTE about
42:27 gummy bears introduce the topic and its
42:29 significance provide an overview of the
42:31 structure origin of gummy Confections
42:33 explore the okay so so that's great so
42:37 I'm going to
42:38 say okay that's fine can you
42:41 now give me that same outline but give
42:45 me three bullet points under each of
42:48 your bullet points that's actual
42:51 historical detail so I can get a sense
42:54 if the paper is what I have in mind
43:00 of course here you go hook all right so
43:02 let's see what it does introduction hook
43:05 start with a captivating story explain
43:07 the purpose okay let's see origins of
43:10 gummy Confections discuss the history of
43:13 a Turkish delight a precursor to gummy
43:15 candies dating back to the Ottoman
43:17 Empire explore how gummy candies made
43:20 with ingredients like fruit juice sugar
43:22 and gelatin emerged in the 19th
43:24 century um let's
43:27 see Hans rle in his vision describe how
43:31 Hans rle the founder of herbo started
43:33 making gummy candies in his kitchen in
43:35 Germany in in the 1920s herbo the
43:38 haribo's first gummy bear uh so so let
43:41 me say
43:44 um did Hans Riel create the first gummy
43:47 bear or did someone create a gummy bear
43:50 before he
43:54 did why a bear
44:00 this is actually kind of curious to me
44:01 Hans rle the founder Who is credited
44:05 wait come back come down
44:10 stop while there were various forms of
44:12 gummy candies before hun ral's times the
44:15 gummy bear as we know it today is
44:17 typically attributed to him and his
44:20 company Hans rle and the invention of
44:23 the gummy bear okay cool inspiration
44:25 from dancing bears the inspiration for
44:27 the gummy bear shape is said to have
44:29 come from the popular Street
44:30 entertainment at the time which often
44:32 featured dancing
44:35 bears
44:37 so I think I'd like to
44:41 change the topic of the
44:44 paper to start
44:46 out with the history of dancing bears
44:50 and how that led to the invention of the
44:52 gummy bear so that the paper ends with
44:55 Hans ryel inventing the gummy bear can
44:58 you give me a new
45:00 outline that's
45:02 [Laughter]
45:07 that um Can it give you the chemical uh
45:10 composition of gummies okay we'll get to
45:12 that let's see let's see what our new
45:13 paper is um history of dancing bears the
45:17 inspiration behind gummy bears Hans rle
45:19 and the gummy bear gummy bears go
45:21 Global um let's see yeah anyway that so
45:26 we we go on like this so this is what I
45:27 would do it um let's let's see can you
45:30 give me the chemical composition of
45:32 gummy bears and the difference between
45:36 soft and chewy gummies and harder
45:42 gummies gelatin sugar colorings citric
45:46 acid soft and chewy gummies are
45:50 typically made for a higher proportion
45:51 of gelatin and a lower proportion of the
45:53 other
45:54 ingredients they have a higher moisture
45:56 content which contributes to their soft
45:58 impliable texture so yeah like like
46:01 that's the thing about chat GPT is you
46:03 can very very quickly run down rabbit
46:05 holes and explore so for your paper what
46:08 I would say is
46:09 like like rather than having to kind of
46:13 lock yourself into an
46:15 idea like run down rabbit holes explore
46:19 and and and maybe you find you know
46:21 interesting hybrid kind of things that
46:23 make it really interesting you know and
46:25 I would say
46:28 set your expectation even if you're
46:30 using chat
46:31 gp4 set your expectation that what
46:34 you'll get out of generative AI if if
46:37 you're really good at prompting if
46:39 you're really good at prompting and do a
46:40 lot of the stuff that I'm talking about
46:42 here will be about
46:44 80% of what you're probably comfortable
46:46 with like these
46:49 things you know my experience with this
46:51 is it's very much a collaboration where
46:54 I use chat GPT to get me to 80% it gives
46:58 me the structure it gives me the basic
47:00 ideas it gives me some research and then
47:02 I have it write a draft or two of
47:04 paragraphs or sections and then I copy
47:06 and paste that into word and I'm like oh
47:08 God this sucks right this oh this middle
47:11 paragraph it completely miss this and
47:12 I'll completely rewrite a middle
47:14 paragraph and then the one below it's
47:17 okay and and it doesn't even necessarily
47:19 save me time but what it does is if I
47:23 have two hours to write the thing I'll
47:25 get to 80 % in 10 minutes and then I've
47:28 got two hours to refine the
47:31 paper
47:33 so I'm curious about how AI will change
47:36 school life within the next few years
47:39 you and me both what's your name son
47:41 wait oh Sun you've been in here before
47:43 uh you and me both like I I think I
47:47 think I think the impact on education is
47:50 going to be profound because I think our
47:52 educational system right now is
47:54 essentially designed to get people
47:58 into specialized tactical execution
48:02 training for the most part there are
48:05 exceptions to that obviously the gummy
48:08 encyclopedia this is
48:16 awesome that's Sherry
48:20 D that's really
48:23 awesome um ask if gummies contain an
48:26 food dies that are detrimental to human
48:28 health oh well we know the answer to
48:30 that uh please tell me what chemicals
48:34 are in gummies that there's evidence
48:37 that they're detrimental to human
48:43 health it might we might get one of the
48:46 safety guard rail things here that seems
48:49 very controversial to me Kyle I don't I
48:51 don't like that qu I don't like your
48:52 line of questioning
48:55 here
48:56 no such thing as a hard
49:02 gummy the last supper with gummies y'all
49:06 are
49:07 weird um sugar artificial flavors
49:11 caloric content let's see artificial
49:14 flavors food colorings many gummy
49:17 candies contain artificial food coloring
49:20 such as red 40 yellow five and blue
49:23 one colorings are linked to hypera
49:26 activity in and attention issues in
49:28 children and apparently 58-year-old
49:31 adults who have haribu gummies right
49:33 next to their little
49:38 desk M
49:41 tasty chemicals good for
49:50 living
49:52 okay let me show you another thing here
49:55 um
49:57 I'm going to go
50:02 grab I'm going to go grab our
50:06 outline so this is the outline that chat
50:08 gpt3
50:12 wrote and I'm going to copy it and now
50:15 I'm going to go over I'm going to do a
50:16 new chat and I'm going to go to if you
50:18 haven't seen Dolly 3 within chat GPT do
50:22 you have any formal training in Ai No
50:25 Sir what are your qualifications yes
50:28 rather uh so uh by qualifications do you
50:31 mean training per se oh yes yes yes we
50:33 were expecting if you had something
50:34 called the AI learning lab first where's
50:36 your lab coat second where's the lab and
50:39 and third by Sir by chance uh what are
50:41 your qualifications sir good day sir uh
50:43 yes about that uh rather I uh I consider
50:47 myself rather an expert in in in uh
50:49 considering the exploration of
50:51 artificial intelligence or machine
50:53 learning it is more accurately known
50:56 I've got no qualifications sir please
50:58 please don't kick me out please oh God
51:01 no no I don't have any anyway
51:04 um no I don't but what I do have so
51:07 here's what I do have I I'll stop being
51:08 a smartass for a
51:10 second
51:12 um I've been an entrepreneur most of my
51:15 life and I've started 12 or 15 companies
51:18 I started one of the first digital
51:19 agencies in the mid 90s when the
51:21 worldwide web came out and so I think
51:24 that one of the skills that that
51:27 I've developed in my life
51:32 is I love technology but I like
51:37 technology when it hits that inflection
51:39 point where the rest of us get access to
51:42 it like my company Story vine I had an
51:45 epiphany about storytelling video
51:48 storytelling when YouTube first came out
51:50 that a lot of the content on YouTube
51:53 sucked because it lacked basic story
51:55 structure and I thought what if there
51:57 were a system what if there were a way
51:59 to provide like like story structure
52:04 story guidance for someone to tell their
52:07 story in a really easy way that would
52:09 create a high quality video without them
52:11 having to learn to be a [ __ ] film
52:13 editor and a director and a writer and a
52:16 cinematographer and a sound designer and
52:18 a graphic designer like it was like it
52:20 was like in order to do like a a solid
52:23 two-minute video you had to you had to
52:25 kind of Master seven different
52:28 professions and it just seemed [ __ ]
52:30 ridiculous to me right and so I'm I'm
52:34 fascinated by technology that that puts
52:37 a s Simplicity layer on top of
52:41 incredibly sophisticated stuff below it
52:43 it's why I'm a fan of apple right Steve
52:45 Jobs was
52:47 maniacal about making technology perform
52:51 as
52:52 magic right the worldwide web prior to
52:57 the worldwide web let me let me flip
52:58 this for a second prior to the worldwide
53:00 web the internet was command line
53:03 interface and it was it was the purview
53:06 of scientists and mathematicians and
53:08 researchers and Librarians and then Tim
53:10 burner Lee said hey we could make a
53:14 hyperlink from this word to a whole
53:16 other
53:17 document and then I forget who did the
53:19 first browser but whoever created the
53:21 first browser made it so that you could
53:23 click on it and so you got five or six
53:28 or seven decades of Internet development
53:32 and all of a sudden the rest of us can
53:33 do this click click click and it changed
53:37 the
53:37 world so my expertise is
53:42 in jumping
53:45 in the moment that [ __ ] shows up so I've
53:49 been [ __ ] around with AI for 5 years
53:51 and like every time Google would say
53:53 there's a new version of tensor flow and
53:54 it's it's much more more intuitive and
53:56 I'd go in there and be like awesome
53:58 intuitive Ai and I'd go in and I'm
54:01 like I it's clear I don't have a
54:03 mathematics degree from Stanford this
54:05 hurts my head and I would
54:08 leave and then I would go in again and I
54:10 would try and I and I did this for
54:12 years and then all of a sudden when I
54:14 started playing around with blockchains
54:16 and nfts I got involved in a project
54:18 called pixel mind which was an nft
54:21 project built on top of stable diffusion
54:23 and it was kind of like Mad Libs where
54:24 you just fill in a few words and it
54:26 would create art for you and I was like
54:28 oh that's kind of magical I started
54:29 playing with that and I started learning
54:31 more about
54:32 it and then I and then I got into stable
54:35 diffusion and dream booth and I created
54:37 an art project I'm like oh this is
54:39 really cool and then chat gbt came out
54:42 and I was like holy [ __ ] this is one of
54:46 those chat
54:48 PT does for the Decades of machine
54:51 learning and artificial intelligence
54:53 technical infrastructure develop
54:55 velopment what the worldwide web did for
54:57 the internet and so what this Channel's
54:59 about is well what the [ __ ] do we do
55:00 with it
55:03 now we've we've all just been given this
55:08 profoundly powerful technology that's
55:10 sort of like as simple as clicking on a
55:13 hyperlink it's like the Google box but
55:16 it's way [ __ ] deeper than
55:18 Google what do we do with it so this
55:22 channel is about exploring that and and
55:25 just like demystifying it and saying
55:28 these tools right now are just
55:30 horrible
55:32 unsophisticated
55:34 janky nasty pieces of software not
55:38 because the companies that develop them
55:40 are bad but just because we're that
55:42 early we're that early that there hasn't
55:44 been time to figure out what these tools
55:47 even do much less put good interface in
55:50 front of
55:52 them
55:54 so um what can I do to use AI to create
55:57 video ads for Facebook here I'll show
55:59 you in a second here or uh to create
56:02 video ads so I so actually the answer to
56:04 that right now is just go to canva and
56:07 sign up for their AI tools canva right
56:09 now is is [ __ ] killing it they've got
56:12 AI scattered throughout canva you can go
56:15 in and just say give me copy for these
56:18 things and it turns it into Instagram
56:19 post it turns it into video they've got
56:21 Runway ml integrated in there just go
56:23 play with canva right now if you just
56:25 want to do video ads um but I want to I
56:28 want to show you something here so for
56:30 the for the people or for the paper we
56:32 were just talking
56:34 about we did our gummy bear paper and I
56:36 just want to I just want to
56:38 demonstrate um image generation with
56:41 Deli in the context of a
56:44 conversation um believe it or not I
56:46 couldn't pin my last comment well this
56:49 one's pinned that's
56:51 weird H I don't know Tik tok's being an
56:54 [ __ ]
56:55 all right so
57:00 I'm let's see writing a paper on the
57:03 history of gummy bears and here is the
57:06 outline so I'm going to hit uh shift
57:10 return I'm going to paste in my
57:13 outline I'm going to throw a couple of
57:16 uh whatever you call it
57:18 uh pound
57:21 signs
57:23 after the outline just kind of indicate
57:26 that it ended that seems to help a
57:27 little bit and then I'm going to
57:30 say I'd like to
57:33 create an image for each of the first
57:36 four
57:38 sections
57:40 that's in a watercolor style and a
57:44 horizontal aspect ratio that brings that
57:48 section to
57:51 life and
57:54 so
57:56 it will now it's it what what it's doing
57:58 right now is it's writing the prompts I
58:01 don't need to write the prompts it it
58:02 has read our outline and I said the
58:04 first four sections so the first one
58:07 is
58:10 um section one um the history of dancing
58:14 bears right section two the inspiration
58:17 behind gummy bears section three Hans
58:19 rle and the birth of the gummy bear
58:21 section four gummy bears go global and
58:26 so it's reading all that and I said I
58:30 wanted watercolors I said I wanted them
58:31 to be
58:33 horizontal and it's off writing prompts
58:35 and now creating
58:37 images and let's see how they are oh
58:41 they're kind of
58:42 interesting so there's dancing
58:45 bears there's dancing bear candy
58:48 watercolor painting of a vibrant fair in
58:50 the early early 20th century dancing
58:53 bear candy
58:57 there's Hans ryel he he he was making
59:00 some mighty giant sized gummy bears
59:03 there and there's gummy bears go Global
59:07 so now I could spend time you know
59:09 tweaking out the uh the style and the
59:12 details and things like that but Better
59:14 Living Through Chemistry remember
59:15 remember those commercials Dragon I
59:19 do
59:21 um this is revelatory to me because if
59:25 if you've not seen how mid Journey works
59:27 let me go show you how mid Journey works
59:30 so I'm now over in Discord so I talked
59:33 earlier about the the uh AI Salon
59:36 Discord so we're in the AI Salon Discord
59:39 one of the channels we have is called
59:40 mid Journey
59:41 play and if if you're want if you're
59:44 wondering what Discord is Discord is
59:47 like a community software essentially
59:50 for gamers and then the nft Bros kind of
59:53 took it over and and almost all nfts
59:57 have Discord communities and servers and
1:00:00 and it's it's not great
1:00:02 software but it's but it's been around
1:00:05 for a long time and it's functional and
1:00:06 you can program within it and so one of
1:00:09 the things you can do is you can do Bots
1:00:10 and one of the the only way you can make
1:00:13 images with mid Journey right now is to
1:00:15 get in Discord and learn all the
1:00:17 commands so if I want to make images in
1:00:19 mid Journey I'll scroll to the bottom
1:00:22 here here's some images other people are
1:00:24 making and notice this got like these
1:00:25 weird codes like u12 34 V12 34 whatever
1:00:29 that symbol is okay what's that I don't
1:00:32 know so you have to know this slash
1:00:34 imagine is what you type in and then you
1:00:37 see this imagine prompt create images
1:00:40 with mid Journey so I click on that or I
1:00:43 just finish typing it and now I type in
1:00:45 a prompt and and so now I've got to
1:00:47 think about let's do the
1:00:50 um an image of Hans ryel inventing the
1:00:54 gummy bear
1:00:57 in his
1:00:58 kitchen with some other
1:01:01 workers in a watercolor
1:01:05 style and then I'm going to go dash dash
1:01:08 AR space 16 colon 9 well Kyle but what
1:01:14 is D- AR 16 colon9 obviously if if you
1:01:21 jeez we we're not it's so obvious isn't
1:01:24 it it's aspect ratio-- AR is aspect
1:01:28 ratio how do you know that I don't know
1:01:30 you just have to [ __ ] learn it right
1:01:31 like this is the problem so you're using
1:01:34 a command line interface to do visual
1:01:37 content creation Now why would anyone do
1:01:39 this because mid Journey's quality is so
1:01:42 good that it's worth it right so I'll
1:01:45 just hit that and there's all sorts of
1:01:46 ways that you can get good at this but
1:01:47 you have to learn it but the difference
1:01:49 between this and and if if you think
1:01:52 about this so I'm over working on my
1:01:54 paper
1:01:55 in chat GPT and it's just in text mode
1:01:59 and then at some point I'm like oh it
1:02:00 would be cool to have representative
1:02:03 images in here for these different
1:02:04 sections what what what you have to do
1:02:07 historically is you have to kind of
1:02:08 shift gears you have to go okay now I'm
1:02:10 going to get into image prompting mode
1:02:12 and I'm going to go to Mid journey and
1:02:14 then I'll look at my paper and I'll come
1:02:16 up with huh I got to come up with
1:02:19 specific things I want in the image
1:02:21 right as opposed to what we just did in
1:02:23 chat GPT where I'm just like here's the
1:02:25 outline go make me some images and it
1:02:28 did it and so
1:02:30 here's obviously it doesn't understand
1:02:33 who Hans rle is because these are all
1:02:35 modern modern images but like notice the
1:02:39 quality of these like mid jour's quality
1:02:41 is really good like these are these look
1:02:43 more waterc colory than the do stuff did
1:02:46 and I can continue to refine this um and
1:02:49 then if you're wondering what these
1:02:50 letters are U stands for upscale v
1:02:53 stands for variation so let's
1:02:55 say I I like the one where there's
1:02:58 actually a bear in the kitchen with
1:03:01 him so I'm going to ask for variations
1:03:04 of this one so that would be V1 so I
1:03:06 click on
1:03:07 V1 and so now it's going to take that
1:03:10 basic whatever it did there use that as
1:03:13 a seed and give me some variations of
1:03:16 that and then I can pick one of these
1:03:18 and I can upscale it and then mid
1:03:20 Journey just added um 2X and 4x
1:03:23 upscaling where you can now make print
1:03:26 quality upscales of your mid Journey
1:03:28 images which is really cool but the
1:03:32 downside here is it is a this is very
1:03:35 much a technologist
1:03:38 approach to a Humanity's
1:03:41 tool
1:03:43 right this hasn't been Apple fied right
1:03:47 this isn't Magic right now like what it
1:03:49 does is Magic how you do it is a [ __ ]
1:03:53 nightmare um
1:03:55 let's see all right here I like this one
1:03:58 he's kind of working with these are the
1:04:00 Goldilocks and the Three Bears so that's
1:04:03 so we'll upscale number
1:04:09 four and there we go there's our image
1:04:11 right so that's mid Journey but I mean
1:04:14 that it it is just a compared
1:04:17 to you know what we just did here we was
1:04:20 like read my thing and I'm like um now I
1:04:23 can just say
1:04:25 those are okay but there's too many
1:04:27 people in the
1:04:28 kitchen comma the gummy bears are too
1:04:31 big comma and the style doesn't look
1:04:34 like watercolors period I'd like them to
1:04:37 look like hand-done watercolor paintings
1:04:41 rather than historical
1:04:45 illustrations so I can just give
1:04:48 it feedback like a creative director
1:04:51 would or just like if you're a client
1:04:54 and you hired someone to make you images
1:04:56 for your thing if they gave you shitty
1:04:59 images you'd go you you'd either what
1:05:02 you would likely do is if you're not a
1:05:04 fan of Confrontation you might go okay
1:05:06 these are great and then not use
1:05:11 them I've done that many many times in
1:05:14 my life or you would go oh well these
1:05:16 weren't what I had in mind can you go
1:05:18 back and do this and then another week
1:05:20 goes by or two weeks go by and you're
1:05:21 spending a lot of money
1:05:23 and
1:05:25 and now we've got
1:05:28 it right here in this
1:05:33 tool and so there's Hans rle in his
1:05:36 kitchen is that does that what Hans rle
1:05:37 looks like I don't fliping
1:05:39 know doesn't matter we can go find
1:05:42 images of him and if he's got a beard we
1:05:44 can say oh he's got a beard and he's got
1:05:46 dark
1:05:47 hair but there's his little Gummy Bears
1:05:49 on his little
1:05:52 plate and it kind of looks like a
1:05:53 watercolor looks like an Andrew WTH
1:06:04 watercolor you know there's our history
1:06:06 of dancing
1:06:07 bears there's our little dancing bear
1:06:09 candy stand there's Hans there's gummies
1:06:13 Go Global like perfect great love
1:06:17 it crazy
1:06:20 right [ __ ] nuts
1:06:23 man what world have we entered you want
1:06:26 to see something crazy too yes Kyle
1:06:28 that's why we're
1:06:29 [Laughter]
1:06:32 here um where is the
1:06:36 image I wonder if it showed up hang on
1:06:39 let me check something let but check
1:06:41 something oh okay yes it's here is this
1:06:47 it yes okay
1:06:52 so um I'm I'm going to export this image
1:06:55 hold
1:06:59 please
1:07:04 whiteboard
1:07:08 export where am I going to go desktops
1:07:11 fine anyone else have
1:07:15 13,743 items on their
1:07:19 desktop you know Kyle that'll make your
1:07:22 computer run less efficiently yeah I
1:07:23 know
1:07:25 what I wait for is technology companies
1:07:27 to make the processors fast enough to
1:07:29 keep up with my lack of
1:07:31 organization okay so we're going to try
1:07:33 something here um another mode in chat
1:07:37 gp4 is just the default mode and you're
1:07:40 like that's boring it doesn't have all
1:07:42 this fancy [ __ ] oh yes it does so the
1:07:44 default mode in chat gp4 now has Vision
1:07:49 they call it
1:07:50 gptv it's it's buried underneath this
1:07:53 little icon that says attach image and
1:07:55 it it's not clear what you do with it so
1:07:57 I'm going to attach an image I'm going
1:07:58 to go find that whiteboard that I just
1:08:01 so we had a meeting today with the AI
1:08:03 Salon organizers and Vicky was there I
1:08:06 don't know if Vicki's in here tonight
1:08:07 Vicky was
1:08:08 there and we wrote down some ideas on a
1:08:12 whiteboard so I'm going to
1:08:15 say here's some notes from a meeting we
1:08:18 had today can
1:08:21 you tell me what we talked about and and
1:08:25 come up with a summarization of the
1:08:31 meeting and so there's there's our
1:08:33 little whiteboard right so like any
1:08:36 whiteboard you've done if you're like me
1:08:38 you probably have I don't know uh 672
1:08:43 pictures of whiteboards in your Vicki
1:08:45 she's here awesome uh you got pictures
1:08:47 of whiteboards going back to
1:08:50 1982 um and and you keep meaning to go
1:08:52 in there and take notes from what was
1:08:54 written on the Whiteboard right well
1:08:57 look at this certainly based on this
1:08:59 meeting summary the end goal discussions
1:09:02 what's the end goal um who are the
1:09:04 audience and participants what were the
1:09:06 main topics the required expertise
1:09:09 engaging high value and daily active
1:09:11 users High development through Community
1:09:14 um platform for talent it appears the
1:09:16 meeting focused on strategies to engage
1:09:18 with Talent develop their skills and
1:09:20 provide a platform for them to Showcase
1:09:22 their work the end goals seem to be more
1:09:24 about informing supporting and educating
1:09:26 potential Talent about
1:09:28 AI um and the importance of recognizing
1:09:32 and nurturing all levels of talent was
1:09:35 emphasized lastly there's a mention of
1:09:37 some kind of board or platform requiring
1:09:39 certain expertise and and an apparent
1:09:42 login credential was noted so so in the
1:09:46 picture in the upper corner here I had
1:09:49 the Wi-Fi the Wi-Fi password for the
1:09:51 building we were in so it even like and
1:09:54 it didn't say user or password it was
1:09:56 just un P but it recognized that as a
1:09:58 login credential and then it took just
1:10:01 these kind of you know unorganized notes
1:10:04 took them I I
1:10:06 mean we we're we're in a different mode
1:10:10 cat GPT is not the same tool it was a
1:10:13 month ago a month ago it was like oh we
1:10:15 got to learn about prompting and we got
1:10:17 to do prompt engineering and we got and
1:10:20 there's a perfect way to prompt now that
1:10:22 you can use an image as a prompt
1:10:25 is prompt engineering as critical um for
1:10:29 a bit but I certainly don't need to
1:10:33 understand how to do image prompting
1:10:35 with the do stuff with the with the
1:10:37 gummy bear pictures I can just ask it
1:10:38 for what I want it's figuring out the
1:10:40 prompting so I think we are quickly
1:10:43 already we're not even a year into chat
1:10:47 GPT and we're already evolving from what
1:10:51 skill was required 6 months ago is is
1:10:54 fundamentally different than what's
1:10:55 required today and it's going to be
1:10:57 fundamentally different 6 months from
1:10:59 now and you're like but Kyle so what how
1:11:01 what the how what am I supposed to do
1:11:04 with
1:11:04 that the only the only strategy that I
1:11:09 know is get curious and adventurous and
1:11:13 think of it like a
1:11:14 game think of it like a game but Kyle
1:11:17 this is my job and this is important y
1:11:19 same same my comp I have an 11 and a
1:11:22 halfy old company that a lot of what we
1:11:25 do is going to get automated out the
1:11:26 value of what we
1:11:28 do is going to be threatened by this
1:11:30 stuff so this stuff's existential for me
1:11:33 too as excited as I am about it like why
1:11:37 I'm so committed to it why I do these
1:11:41 three-hour sessions seven nights a week
1:11:44 for the past four months is I do not
1:11:47 want to lose touch with what's
1:11:51 happening and I'm
1:11:54 trying my damnedest to create as big a
1:11:56 community as possible that's doing the
1:11:59 same thing that's what this is
1:12:03 about so so you know questions like what
1:12:06 are your qualifications
1:12:08 sir there are no experts right now I'm
1:12:10 not an expert in this [ __ ] I don't
1:12:12 [ __ ] know like I don't I don't know
1:12:14 what it means that I
1:12:17 can in five minutes have it create a
1:12:20 very detailed outline of a paper just
1:12:24 flip into a different mode and say go
1:12:25 make me pictures for that at some point
1:12:27 we're going to be able to say go write
1:12:29 me a song for that go make me a video
1:12:30 for that if you if you hang out in the
1:12:33 AI Salon Herod who who hangs out in here
1:12:36 he's one of the Irregulars he's doing
1:12:38 just some amazing [ __ ] with
1:12:41 video like all of all of the stuff that
1:12:45 that right now are these disperate tools
1:12:47 out in the corners of whatever this
1:12:50 stuff
1:12:51 is are very very rapidly going to roll
1:12:55 into a single
1:12:57 tool that is kind of the everything
1:13:00 app what are you supposed to do with
1:13:03 that as a professional I don't
1:13:08 know are there any AI that can do
1:13:10 animation shorts
1:13:12 absolutely um I but it depends on what
1:13:14 you mean animation shorts there there
1:13:17 are like Runway ML and P labs and what's
1:13:21 the other one something Mountain uh hang
1:13:23 on is
1:13:25 Moon Moon Valley something
1:13:28 Mountain thanks add thanks thanks for
1:13:32 the attention to detail
1:13:33 brain it's something mountain or it's
1:13:36 Moon Valley so Moon Valley's one but
1:13:38 this this one you make videos in Discord
1:13:42 so this is like like mid journey and
1:13:44 then there's P Labs which might also be
1:13:48 um Discord and then there's Runway
1:13:52 ml which let's go to
1:14:00 runwayml
1:14:02 Runway
1:14:05 ml try it for free log
1:14:12 in so bring your image to life with Gen
1:14:15 2 so I can either start with a text
1:14:17 prompt or an image prompt so if I start
1:14:19 with a text prompt I can just
1:14:22 say
1:14:30 make me an image of or not an image hang
1:14:36 on a dancing gummy bear on a
1:14:39 confectioner's
1:14:43 table while the
1:14:46 confectioner dances
1:14:50 along and I can say generate a 4 second
1:14:53 clip
1:14:54 and I didn't put any style information
1:14:56 in there by the way do you like the um
1:14:58 the fingerprints I did that for you see
1:15:01 it's very it's very organic this channel
1:15:04 you know it's not very professional you
1:15:06 should probably clean your screen
1:15:08 probably I should probably have
1:15:09 qualifications too but I
1:15:12 don't I gotta I got to up the game game
1:15:15 for one of the AI learning Labs at the
1:15:17 Vegas sphere but you know that's that's
1:15:19 not till next summer um how to mimic
1:15:21 pictures you have for someone but
1:15:23 changing some elements um so so there's
1:15:28 there's a few different ways to do that
1:15:30 roro um within mid Journey you can
1:15:33 upload an image that's kind of your base
1:15:35 image and then describe a new image
1:15:36 after that if you really just want to
1:15:38 modify elements of an existing image I
1:15:41 would use something like photoshop's
1:15:43 generative fill where you can select
1:15:46 parts of an image and say if someone's
1:15:48 wearing a sweater and you want to change
1:15:49 it to a denim shirt you just kind of
1:15:51 highlight their shirt and say you know
1:15:54 change this sweater to a denim shirt um
1:15:57 so I would do something like that all
1:15:59 right let's see how bad this gummy bear
1:16:01 is okay that is a horrible gummy that is
1:16:02 a terrifying gummy
1:16:04 [Laughter]
1:16:17 bear I'm sorry I thought this was going
1:16:20 to be bad this is [ __ ] genius
1:16:27 I think we have a new Meme here people
1:16:30 um okay so that's pretty bad but that's
1:16:33 one of these so bad it's good things so
1:16:35 so now let's go TR try something else so
1:16:37 we'll go we'll go back to dolly or yeah
1:16:40 to to to uh J
1:16:44 J and let's say
1:16:48 um please create a 16x9 image of a
1:16:53 highly
1:16:56 photorealistic closeup of a gummy bear
1:17:00 on the confectioner's table that looks
1:17:03 as though it was shot in a high-end
1:17:05 photography
1:17:07 studio okay
1:17:09 so what I'm going to do now hopefully is
1:17:12 get an image of a gummy bear that
1:17:14 doesn't
1:17:16 suck and might be a good basis for an
1:17:20 animation like
1:17:22 that and now we'll use this image as the
1:17:26 basis for that
1:17:29 animation and and you know the caveat
1:17:32 with all of these tools that that you
1:17:34 know I say pretty regularly in here but
1:17:36 I you know it's oh that's great although
1:17:39 let's see that's it's not quiet so hang
1:17:42 on
1:17:44 um that's close but I want four
1:17:46 different variations and I
1:17:49 want the gummy bear um standing on two
1:17:53 feet
1:17:54 with its
1:17:57 arms extended as if it's dancing period
1:18:02 I also want it on a kitchen table with
1:18:07 sugar and a stainless steel
1:18:09 bowl and a little piece of paper that
1:18:12 says gummy bear
1:18:14 recipe
1:18:17 period the focus of the image should be
1:18:19 the gummy bear but I want it clear that
1:18:22 it's in the kitchen where it was
1:18:26 invented all
1:18:29 right um so all of these tools right now
1:18:32 are
1:18:36 um very
1:18:39 imperfect and and what I would strongly
1:18:43 encourage you not to do is don't dismiss
1:18:45 them because they
1:18:48 suck because 18 months ago image
1:18:52 generation tool were abysmal like they
1:18:56 were like like the the well so so let me
1:18:59 let me put it in context for what they
1:19:01 were doing they were absolutely [ __ ]
1:19:03 remarkable right from a mathematical
1:19:05 standpoint and and the fact that they
1:19:07 could generate images at all was
1:19:09 absolutely
1:19:10 miraculous um but now 18 months
1:19:14 later these tools are in seconds
1:19:17 creating worldclass images um but Kyle
1:19:20 are you demonstrating that chat GPT can
1:19:22 read handwritten notes down yes I am it
1:19:25 can read handwritten notes it can read
1:19:27 oh this is perfect look at this one
1:19:31 perfect gummy bear look there's the
1:19:34 gummy bear recipe there's our dancing
1:19:36 gummy bear this is perfect okay so watch
1:19:39 so now we're going to go save image
1:19:42 as and we'll just call it
1:19:45 gummy
1:19:48 dance so what' that take us 2 minutes to
1:19:52 get an image that's like okay that's not
1:19:54 bad so now we'll go back over to Runway
1:19:56 ml we're going to go to an image drag
1:19:59 and drop an image so let me go find my
1:20:03 image where did I put it gummy dance
1:20:05 here it
1:20:09 is well I'll upload a
1:20:12 file desktop is this it oh it saved it
1:20:16 as okay hang on it generated it as the
1:20:19 wrong file format it's a webp image and
1:20:23 now we've got to export it as a PNG this
1:20:25 is the kind of crap that that makes me
1:20:28 insane like the fact that there's all
1:20:30 these different file formats and all the
1:20:33 services don't support them is just
1:20:35 [ __ ] maddening but
1:20:38 whatever this is the kind of stuff that
1:20:40 slows you down you know when you're like
1:20:42 hey I have an idea for something and
1:20:44 then like two hours later you're still
1:20:46 [ __ ] Googling what's webp format and
1:20:49 is there a webp Transformer and you've
1:20:52 down downloaded six pieces of software
1:20:54 and they all install malware on your
1:20:56 computer and then they all charge you 20
1:20:58 bucks to trans
1:21:03 trans that needs to be fixed and that
1:21:05 will be fixed because um because AI is
1:21:09 going to fix that
1:21:12 [ __ ] all right boom there's our little
1:21:16 image oh wait I want image plus
1:21:19 description uh a dancing gummy
1:21:24 bear on a confectioner's table table
1:21:27 let's just we'll leave it like that okay
1:21:29 so we've got our uploaded image here
1:21:32 we've got a dancing gummy bear on a
1:21:34 confectioner's table and we're going to
1:21:35 say generate 4 seconds so while while
1:21:39 that's generating wait is it
1:21:44 generating yes oh wait it's generating
1:21:47 twice can I cancel one of
1:21:49 these no but while that's generating
1:21:52 let's let's watch our let's watch our
1:21:56 our
1:21:59 Masterpiece my name is I was the gummy
1:22:02 I'm a gummy gummy Cherry guy yeah the
1:22:06 dancer I'm awesome look at me go [ __ ]
1:22:12 awful all right so here's our our real
1:22:16 dancing gummy
1:22:19 bear
1:22:21 um I'm gonna be I'm going to be
1:22:23 disturbing dancing gummy bear for
1:22:25 Halloween exactly all right let's see
1:22:28 how it did please not suck please don't
1:22:31 suck it
1:22:34 sucks all right it's disturbing but you
1:22:38 know and again you could you could
1:22:40 likely muck with this and and get it to
1:22:43 do something predictable right I don't I
1:22:45 don't I haven't played around enough
1:22:47 with with Runway ml to know what are the
1:22:50 what are the secrets to get coherence
1:22:52 and and from what I understand P Labs is
1:22:54 better than
1:22:56 Runway but
1:22:58 again go back to this idea that these
1:23:02 tools are as bad as they're ever going
1:23:04 to
1:23:05 be right so every time it does
1:23:08 this they're looking at this and they're
1:23:10 going okay that's awful how can we
1:23:12 improve the coherence how can we improve
1:23:13 the coherence how can we improve the
1:23:16 coherence um last week I asked you about
1:23:19 using chat GPD to clean up a CSV file
1:23:21 for old leads of clients
1:23:23 then I heard elsewhere not to upload
1:23:25 personal
1:23:26 info why okay
1:23:30 so
1:23:32 um let me stop this let me flip this and
1:23:37 go
1:23:41 here um so part
1:23:45 of part of the tools right now is um you
1:23:51 should just assume that these tools are
1:23:55 not safe and secure until they prove
1:23:58 themselves
1:23:59 otherwise um
1:24:01 because um some of them aren't and and
1:24:03 in the case of chat GPT in
1:24:05 particular um if you leave on your
1:24:09 prompt
1:24:10 history um the prompts that you put in
1:24:12 there are susceptible to hacking
1:24:14 basically in in a in a pretty pretty
1:24:17 easy way so you should just assume that
1:24:20 that anything you put out there you know
1:24:22 can be gotten to
1:24:24 now open aai has created an Enterprise
1:24:28 version of their software which isn't
1:24:30 available to everyone yet you can also
1:24:34 if if you
1:24:36 um if you really want to be if you
1:24:39 really want to keep your stff your stuff
1:24:42 private and secure you can install a a
1:24:47 open-source large language model on your
1:24:49 local machine or on a virtual machine in
1:24:52 in the cloud that has its own boundaries
1:24:55 so you're not sending um data out to a
1:24:58 company like open Ai and then getting it
1:25:01 back um you're just doing it within your
1:25:04 own machine um so there are ways that
1:25:06 you can do things more securely but they
1:25:08 require a lot more technical
1:25:11 stuff so if it's the kind of thing where
1:25:15 um you can depersonalize the information
1:25:18 great if it's just you know you're
1:25:20 you're if there's nothing proprietary
1:25:22 about the information uh you can do it
1:25:25 if you're using something like code
1:25:28 interpreter within chat gbt which is
1:25:31 called Advanced Data analysis it's
1:25:34 actually running the the data analysis
1:25:37 it does in a container inside your chat
1:25:41 I don't think that that information is
1:25:44 going out to the world and back I think
1:25:46 it's doing all that analysis locally
1:25:48 there but even if that's true like I
1:25:52 wouldn't necessarily trust that so it's
1:25:54 just it's just the nature of these tools
1:25:56 right now we're very early so you should
1:25:59 just assume right now until proven
1:26:02 otherwise like included in your terms of
1:26:05 service is your data secure you should
1:26:08 assume that it's not and so don't put
1:26:10 anything
1:26:11 um
1:26:14 super private or sensitive out out in
1:26:17 these tools de depersonalize it
1:26:21 first
1:26:24 welcome everybody there's a bunch of new
1:26:26 people in here let me let me do I want
1:26:28 to do a little reset here quickly so um
1:26:32 my name is Kyle Shannon this the AI
1:26:33 learning lab uh I'm an entrepreneur and
1:26:36 uh this channel is about obviously AI
1:26:39 stuff and talking about AI stuff if you
1:26:41 have not tried chat gbt yet um
1:26:51 go go there chat.
1:26:56 open.com that's the official chat GPT
1:26:59 website it's a lot of the demos I've
1:27:01 been doing tonight like the dolly3 stuff
1:27:03 that's all there um the chat GPT plus
1:27:07 their 20 Buck a month like upgrade in my
1:27:11 opinion is worth it I'm not connect I'm
1:27:13 not affiliated with them in any way so
1:27:15 if you don't do it I don't I don't
1:27:16 really give a [ __ ] I don't make money
1:27:17 whether you do it or not but they have
1:27:20 added so many features in the past month
1:27:22 that I think this is now kind of the the
1:27:24 deal of the deal of the year or deal of
1:27:27 like I think it's well worth the 20
1:27:29 bucks um if if you're you know super
1:27:32 curious and want to dig in if you're not
1:27:34 if you just want to dick around with it
1:27:35 just go there play with play with the
1:27:37 free version um you can also play with
1:27:39 Bard is free Claude has a free version
1:27:43 you can upgrade to get you know more uh
1:27:45 and Bing is free so all four of these
1:27:48 have free options but if you're brand
1:27:50 new to this this uh generator AI stuff
1:27:53 just go to chat GPT right now while
1:27:55 while I'm rambling you're like it's
1:27:57 Sunday night Kyle I'm a little tired I
1:27:59 don't feel like working I'm not very
1:28:00 technical I don't want to learn doesn't
1:28:02 matter just go there just do it just do
1:28:05 it do it do it do it do
1:28:09 it just go there and just start playing
1:28:12 well what do I do I don't know what to
1:28:14 ask and I'm not very technical you don't
1:28:16 need to be
1:28:17 technical just go here go go tell it to
1:28:20 write you a country song about two gummy
1:28:23 bears that are falling in love and it
1:28:25 will do that and then have it write you
1:28:29 a marketing plan for your pizza shop and
1:28:32 then have it give you an exercise plan
1:28:35 for getting you off the couch after
1:28:37 being sedentary for three years thanks
1:28:39 to co and it just just go just go play
1:28:43 while while I ramble here so anyway um
1:28:46 and then if you have any questions about
1:28:47 AI pop them in the comments below I'll
1:28:49 do my best to answer them and uh and
1:28:52 then we'll have some fun we'll have some
1:28:54 fun and you know you can you can make
1:28:56 money
1:28:58 with we can talk about that uh I'm happy
1:29:02 to demo stuff um
1:29:05 anything all right let's see we are on
1:29:09 to something let's see it looks like
1:29:11 it's playing Beat
1:29:14 saber this is giving Pillsbury Doughboy
1:29:17 Vibes those yeah those videos were
1:29:19 pretty disturbing weren't they they were
1:29:21 pretty good
1:29:25 [Laughter]
1:29:31 all right let's leave that in the
1:29:32 background while I talk
1:29:34 [Laughter]
1:29:37 here what is that disturbing image oh my
1:29:41 it is that
1:29:44 [Music]
1:29:48 Satan it's perfect Kyle it's paranormal
1:29:50 gummy bear for Halloween it it these are
1:29:52 very Halloweeny kind of disturbing this
1:29:55 looks like some Herod if you put this to
1:29:57 your heavy metal rock music this would
1:30:00 this would play pretty
1:30:06 nice
1:30:12 Waka oh
1:30:17 man all right that's that's distracting
1:30:19 me now sorry I got to go back to static
1:30:21 images I can't take it I can't take it
1:30:26 people howy partner life's tough it's
1:30:29 tougher when you're
1:30:31 [Music]
1:30:34 stupid there's not enough AI learning in
1:30:37 this town for the both of us one of us
1:30:40 is going to have to go Skippy
1:30:45 Skippy
1:30:51 yeah I just got in a fight with
1:30:55 Bing Bing's an [ __ ] I just bing bing
1:30:59 I'm just Bing's on my [ __ ]
1:31:03 list you should be able to talk back to
1:31:05 the robots without them getting offended
1:31:08 being is just an over sensitive
1:31:10 narcissistic
1:31:15 nin I hope someone from Microsoft is in
1:31:18 here lighten up on your safety guard
1:31:20 rails fingerprints is t
1:31:32 y
1:31:36 oh see
1:31:38 listen on other channels trolls just
1:31:41 come in and they just are mean but here
1:31:44 at the learning lab trolling can be
1:31:49 fun it's like trolling and self
1:31:54 expression oh my God this is so good
1:31:58 wait this is from Ted what did you say
1:31:59 here we're all just finding our way this
1:32:01 is very right this is what's what's
1:32:04 going on right now like this is what
1:32:05 we're all trying to figure
1:32:07 out like what is this stuff how do we do
1:32:10 it I don't
1:32:14 know it's been 15 minutes since my last
1:32:17 Dolly 3 I've got such
1:32:21 shame
1:32:25 oh my God it's so
1:32:27 funny it's crazy I tell you it's crazy a
1:32:31 Pap all right so um do me a favor if
1:32:36 you're new here um if you're
1:32:38 likeing if you're new here and like what
1:32:41 the [ __ ] is this channel
1:32:42 welcome uh if you're new here and you're
1:32:45 enjoying this do me a favor share the
1:32:47 live with your friends even though
1:32:50 they're going to think you're weird for
1:32:51 sharing something like this um follow
1:32:54 the channel um so you get notifications
1:32:57 when we go live I've I try to go live
1:32:59 every night I've been I don't think I've
1:33:01 missed a night except when I've gotten
1:33:02 kicked off Tik Tok three
1:33:05 times I've gotten better I haven't done
1:33:07 it in a while I should I should knock on
1:33:09 wood to to not get kicked off
1:33:13 um but follow the
1:33:15 channel um and and uh and yeah do any do
1:33:18 anything you can to to engage here I'd
1:33:20 like to like to get as many people in
1:33:22 here as possible I've got Bing
1:33:25 trapped where it can't hang up on me oh
1:33:27 that's good you should let's see what I
1:33:31 had going on with it you should have you
1:33:34 should see what you had going on with it
1:33:36 this
1:33:36 morning well that's good that you've got
1:33:38 it to to uh to not hang up on you it
1:33:41 hangs up on me all the time it drives me
1:33:44 crazy Bing drives me
1:33:45 nuts how about a gummy bear jumping into
1:33:49 the eye of a carved pumpkin for
1:33:50 Halloween that would work what's good
1:33:52 about Pi okay here's what's good about
1:33:55 Pi
1:33:57 so let me tell you what's bad about
1:34:02 Pi so if you go to Pi it's just like
1:34:06 chat GPT all these are just prompts
1:34:09 basically like I need whatever just vent
1:34:12 like this just tells you what you can do
1:34:13 with it but it's just like chat GPT
1:34:16 where you just you know there's a little
1:34:17 prompt box and then you type [ __ ] in and
1:34:20 it types stuff back
1:34:22 if you try to actually get something
1:34:24 done with pie like give me a list of the
1:34:28 ingredients of gummy
1:34:31 bears it will be the most frustrating
1:34:33 experience of your life because it's
1:34:36 designed to be a conversational bot So
1:34:39 when you say give me the ingredients for
1:34:41 Gummy Bears it'll be like that's an
1:34:43 awesome request I'd love to give you the
1:34:45 ingredients for Gummy Bears can you tell
1:34:46 me why you want the ingredients for
1:34:48 Gummy Bears you like I just want them
1:34:51 could you just give them to to me in a
1:34:52 bulleted list a bulleted list is a great
1:34:54 way to do that so it can be very very
1:34:57 annoying if you just want [ __ ] done
1:34:59 but if you want to talk it's really good
1:35:03 at that it's so it's designed to be a
1:35:05 conversational AI so rather than type
1:35:08 just put it into conversation mode and
1:35:10 start talking to
1:35:13 it hey I really love gummy bears and I'd
1:35:16 love to know a little bit about the
1:35:18 history of them where were they invented
1:35:20 and how did they get
1:35:25 [Music]
1:35:29 popular gummy bears are delightful
1:35:32 little treats with a fascinating history
1:35:34 they were invented in Germany in the
1:35:36 1920s by Hans Regal who was looking for
1:35:39 a way to create a new type of candy that
1:35:42 was soft chewy and had a longer shelf
1:35:47 life than other candies at the time
1:35:48 she's so Saucy the original gummy bears
1:35:50 were fruit flavored and made with a
1:35:52 combination of gelatin sugar and fruit
1:35:55 juice they became popular in Germany and
1:35:58 spread throughout Europe and the United
1:35:59 States in the 1950s since then they've
1:36:02 become a staple of candy stores and kids
1:36:04 lunch boxes
1:36:07 everywhere and why
1:36:19 Bears ah I see where you're going with
1:36:21 this New York Bears huh it's true black
1:36:23 bears have been spotted in parts of New
1:36:25 York State particularly in the more
1:36:32 [Music]
1:36:35 rural no I mean why are they gummy bears
1:36:39 why are bears the
1:36:50 shape
1:36:53 that's an excellent question the reason
1:36:55 why gummy bears are shaped like Bears is
1:36:57 actually pretty cool the creator of
1:36:59 gummy bears Hance Regal was inspired by
1:37:02 the trained Bears he grew up watching at
1:37:05 street festivals in Europe so so that's
1:37:09 that's what it's good at and then you
1:37:11 know you can it can be educational
1:37:13 content you can talk about just anything
1:37:17 and and it's it's designed to do that
1:37:19 it's funny I just saw a notification pop
1:37:20 up on my phone someone made a comment on
1:37:22 one of my videos chat GPT is overrated
1:37:25 whenever I see something like chat GPT
1:37:27 is overrated it's just aign to me that
1:37:30 someone's sitting on the sidelines and
1:37:31 is not actually curious and exploring it
1:37:34 because it is both underrated and
1:37:38 overrated um like it definitely has
1:37:42 improvements but the deeper you dig with
1:37:45 it the more remarkable it is so anyway
1:37:50 um my hope is that if you're here and if
1:37:55 you like and if you're digging it is
1:37:57 that you just start getting curious and
1:37:58 start exploring just be in an be in an
1:38:02 open mind
1:38:04 like while I'm enthusiastic and and
1:38:07 optimistic about this stuff you don't
1:38:09 have to be but I think it's important
1:38:11 for everyone to at least get educated on
1:38:15 what it makes
1:38:17 possible where are the boundaries what's
1:38:19 it good at what's it not good at
1:38:24 what of your job can it do better than
1:38:27 you
1:38:29 do like go there you know that that
1:38:33 saying don't ask a question you don't
1:38:35 want to know the answer
1:38:37 to I think that's a lot of what's going
1:38:39 on with the fear of AI is that people
1:38:42 don't want to know the answer to that I
1:38:43 think it's important to know the answer
1:38:45 to that because the more you use this
1:38:48 stuff the more you realize this is an
1:38:50 imperfect tool
1:38:52 that is going to change everything but
1:38:54 it still requires lots of human
1:38:56 engagement and lots of human oversight
1:38:58 and lots of human
1:39:00 participation and the more of us that
1:39:02 are using it and understanding that the
1:39:05 better we're going to get at it the the
1:39:07 better we're going to be able to
1:39:10 influence what it is moving
1:39:13 forward Amelia's wife we're just doing
1:39:15 chitchat up till now okay Kyle we can
1:39:18 start oh Amelia's wife is here hang on
1:39:21 got to get the uh the appropriate image
1:39:24 uhuh oh wait I'm in the wrong section
1:39:26 hold
1:39:34 please am I
1:39:36 late the answer is blowing in the wind
1:39:40 and yes you're late you're very late oh
1:39:43 hi Mark hey oh
1:39:50 hi
1:39:55 does he have
1:39:58 [Laughter]
1:40:03 Tourette's sometimes I realize what a
1:40:06 what a n poop I am and it just cracks me
1:40:09 [Laughter]
1:40:15 up all right let's get serious this is
1:40:19 this is called the AI learning lab this
1:40:21 is about learning thank you for the uh
1:40:23 for the hot pepper there Mike I
1:40:25 appreciate that the uh the uh those
1:40:28 gifts mean a lot thank you um I want to
1:40:31 see what can be done with Hardware
1:40:32 what's the potential for llm
1:40:35 applications on Hardware I think it's
1:40:37 huge Ron so if I were you thank you
1:40:40 Robert Rosy I appreciate that um if I
1:40:42 were you I like go to Reddit and and go
1:40:45 start trolling the open-source um uh
1:40:50 like GitHub reposit atories and the
1:40:54 um hugging face um
1:40:59 projects and
1:41:01 um just start like like yes there's
1:41:05 there's all sorts of stuff you can do
1:41:06 with Hardware right now there's I showed
1:41:08 I showed a thing last night that's
1:41:12 um this is some
1:41:14 Nvidia a team from Nvidia created this
1:41:17 thing yesterday this is [ __ ]
1:41:19 insane let me see where I have it do I
1:41:22 have it anywhere no let me go
1:41:29 x.com and go
1:41:32 to
1:41:34 pen
1:41:39 spinning look at this
1:41:47 so so this group at Nvidia use chat GPT
1:41:52 T4 with this this new approach they
1:41:55 called
1:41:56 Eureka to train a virtual robot hand to
1:42:00 spin a pen they trained it in 2 minutes
1:42:04 this is apparently a notoriously
1:42:07 difficult exercise for even 3D animators
1:42:11 to get right for like like like just
1:42:14 doing the animation and they actually
1:42:16 trained the robot to do this in two
1:42:17 minutes using chat
1:42:19 GPT um
1:42:21 so that same technique I mean that's a
1:42:23 virtual robot right but you could send
1:42:26 that code to a real robot hand right and
1:42:28 then maybe it doesn't do it in the real
1:42:30 world but give it another two minutes so
1:42:33 so there's all sorts of projects right
1:42:35 now um go look at the Boston Dynamics
1:42:38 stuff where they put chat GPT inside one
1:42:40 of their little robot dogs that was that
1:42:42 was like months ago three or four months
1:42:44 ago so I assume at this point it's it's
1:42:46 much further but they're doing all sorts
1:42:48 of cool stuff where they're um they're
1:42:51 having chat GPT output like using
1:42:54 special instructions basically they're
1:42:56 having g chat GPT output Json so it can
1:43:00 read Json and then it can output Json
1:43:02 and then they they basically just hook
1:43:03 the apii up to the to the uh to the
1:43:08 robots instructions and you can talk to
1:43:11 the robots and control them and do [ __ ]
1:43:13 so um lots and lots of stuff going on
1:43:15 there I'm not in that world at all so I
1:43:17 don't know anything about it I'm new to
1:43:18 Mid journey and Discord do you have an
1:43:21 AI learning lab Channel kind of we do so
1:43:24 we've got a
1:43:26 um a thing called the AI Salon which is
1:43:29 the a community that I started about a
1:43:31 year ago and the AI Salon the salon. a
1:43:35 is the
1:43:36 URL and that uh we have a Meetup so
1:43:41 every two weeks there's an inperson
1:43:43 Meetup uh it's iners here in Denver but
1:43:45 then it's virtual for everyone
1:43:47 else and we meet every two weeks so this
1:43:50 coming Tuesday is the next meeting and
1:43:51 then there's a salon Discord so a lot of
1:43:54 the people that come to this channel
1:43:56 regularly um we're using the salon
1:43:59 Discord is the place to kind of extend
1:44:01 this community and extend the
1:44:02 conversation and and talk and
1:44:04 collaborate in between you know these
1:44:06 meetings as well so yeah there
1:44:09 is and looks like there's a lot of
1:44:12 people joining and introducing
1:44:14 themselves which is awesome so if you go
1:44:17 over there if you if you go join the
1:44:19 salon um please introduce yourself
1:44:22 there's there's an introduction Channel
1:44:24 an introduce yourself Channel let us
1:44:25 know who you are what your background is
1:44:27 where are you on the AI Journey like if
1:44:29 you're just starting cool if you're like
1:44:31 you know I've been doing machine
1:44:33 learning for 30 years cool if you're
1:44:35 somewhere in between like most of us
1:44:36 probably are cool um everyone's trying
1:44:39 to figure this stuff out right now
1:44:41 everyone's trying to figure this stuff
1:44:43 out no one is an expert zero people zero
1:44:46 of the people zero zero
1:44:50 people know what's going on right now
1:44:54 what a time to be alive isn't it I mean
1:44:57 that thing two minutes to train a robot
1:45:00 to do that so so think about
1:45:05 that I didn't get that Siri Siri
1:45:10 Siri Siri so jealous Siri hey go
1:45:19 away
1:45:24 it's oh you can't see it wa oh you can't
1:45:27 see it it said it said Siri is so
1:45:30 jealous Siri go away and then it just
1:45:35 faded
1:45:39 um we're in this remarkable transition
1:45:42 right
1:45:44 now
1:45:47 where we're moving from a Time time
1:45:51 where if you wanted to program a
1:45:53 computer you had to think
1:45:56 about the
1:45:58 logic like part of what makes
1:46:00 programming fun is doing the problem
1:46:03 solving and the
1:46:05 logic navigation to say okay if I want
1:46:08 this thing to do that how do I do that
1:46:10 and so if you wanted to get a robot hand
1:46:12 to twirl a pen you'd have to think about
1:46:15 okay how do I you know okay I'm going to
1:46:18 move this finger down like that and I
1:46:20 got to figure out the timing and there's
1:46:22 going to be milliseconds and I got to
1:46:24 and then there's the inertia of the
1:46:26 thing spinning and there's gravity and
1:46:28 and like you have to think about all
1:46:29 that logic
1:46:30 right and so that takes a lot of time
1:46:33 and it's a lot of trial and error it's a
1:46:35 lot of back and forth and what they did
1:46:37 in this case is they had chat GPT just
1:46:40 do the sort of problem solving and and
1:46:43 they did the kind of feedback like
1:46:45 that's good that's not good and in two
1:46:47 minutes it just trained it to do
1:46:49 it
1:46:51 wait like
1:46:55 what think about all the problem solving
1:46:58 that is complicated problem solving that
1:47:02 we're going to be able to just toss at
1:47:04 these tools and let it figure it out
1:47:08 that's like that we're in the middle of
1:47:09 that transition right now and I think
1:47:11 that's that's one of the things that
1:47:13 people are kind of missing with this
1:47:16 stuff is like if they go in and they
1:47:17 like ask it to to write something and it
1:47:20 writes something they're like oh well
1:47:21 it's it's it's inaccurate so therefore
1:47:23 it's
1:47:24 useless but they haven't sort of dug
1:47:27 down to say well what what can it
1:47:28 actually do like I understand that it
1:47:30 might make hallucinations and it might
1:47:32 make mistakes but what can it actually
1:47:34 do and the people that are taking these
1:47:37 tools and and running down the rabbit
1:47:39 holes are inventing [ __ ] like
1:47:42 this and so now once you get it so that
1:47:45 you can train a robot to be that
1:47:46 dextrous in 2 minutes with chat
1:47:49 GPT make money with what the [ __ ] else
1:47:53 can you do with
1:47:55 it I don't know someone in one of these
1:47:59 calls is going to go oh you know what I
1:48:02 could probably do with it and they'll oh
1:48:05 it can do that and they're going to
1:48:06 invent some [ __ ] new way of doing
1:48:08 something
1:48:10 because they weren't closed-minded to
1:48:13 the fact that it's not perfect right now
1:48:14 sorry I had to log out what's the robot
1:48:17 what wait what's with the robots dancing
1:48:19 where's that
1:48:21 oh wait the the gummy bears
1:48:25 dancing
1:48:28 this
1:48:30 that that's a gummy bear
1:48:34 dancing you're like why why does it have
1:48:37 things coming out of its head because
1:48:39 because that's where we are that is
1:48:41 Runway ml so I took an
1:48:45 image
1:48:47 um where's my image
1:48:51 I took that image and put it into Runway
1:48:54 ML and made it dance if you're asking
1:48:57 about the robot hand spinning the pen
1:49:01 that
1:49:05 thing that is some researchers at Nvidia
1:49:09 use chat GPT to train a virtual robot
1:49:12 hand how to do pen spinning and they
1:49:15 they got it so that it could train it in
1:49:17 two minutes or something like that some
1:49:19 crazy ass thing and it's doing all sorts
1:49:22 of dexterity stuff using chat
1:49:26 GPT I was also playing with trivia with
1:49:29 Sydney it refreshed and it was Bing's
1:49:32 trivia was 100% different oh that's
1:49:36 interesting
1:49:38 huh yeah the these things they really
1:49:40 are weird like like here here's the here
1:49:45 here's my CH my challenge with these
1:49:47 tools right now is they demo really well
1:49:51 and then when you try to get them to do
1:49:53 something for a particular project
1:49:55 you're like uh they're not quite there
1:49:59 like dang
1:50:01 it so they're frustrating in that way
1:50:04 but like like I've gotten I've I've
1:50:07 created four or five different tools for
1:50:09 my company that we use we use regularly
1:50:12 now and and like one of the tools that
1:50:14 we use I think it was just written on um
1:50:17 it's not even using uh GPT 4 it's just
1:50:19 using GPT 3.5
1:50:21 so anyway any
1:50:27 any let's go to oh you know which one I
1:50:29 want to go
1:50:34 [Music]
1:50:37 to all these images by the way were all
1:50:40 made in Del this one this is AI
1:50:43 Anonymous hi my name is Kyle it's been 2
1:50:46 minutes since I've made something in DOI
1:50:49 I'm making something right now now Kyle
1:50:52 thanks that's awesome you're the
1:50:58 best could could you make a picture of
1:51:00 an obscure history book inside a UFO
1:51:04 sure here you go Kyle what did you
1:51:08 create the dancing things in I missed
1:51:09 the beginning um Runway
1:51:12 ml so I did I just typed in text to
1:51:17 video you'll appre appreciate this over
1:51:20 Fox so did
1:51:22 um I did text to video so this was
1:51:27 um this this one was um gummy bear
1:51:32 dancing on a on a confectioner's table
1:51:35 and so this is what it created so so
1:51:37 this you can have nightmares about but I
1:51:39 think that's kind of awesomely
1:51:43 bad and
1:51:45 then we went over to
1:51:49 um chat
1:51:52 GPT
1:51:56 and created this image with Dolly 3 and
1:52:00 I'm like oh that's going to be a great
1:52:03 video and then so that was the start
1:52:05 image for this thing which you're like
1:52:08 okay this is going to be awesome and
1:52:10 then it similarly kind of goes kind of
1:52:13 goes all devil on
1:52:17 you so so that's what that is
1:52:22 um could pen spinning become surgical
1:52:24 assistants yes yes there there one of
1:52:28 the one of the other things there there
1:52:30 was like one of the other things that
1:52:32 happened this week was was someone
1:52:34 figured out a way to get chat GPT to
1:52:38 understand like the folds of the brain
1:52:40 or to visualize the folds of the brain
1:52:42 based on something like like yes like
1:52:45 that's the thing
1:52:47 is for any of these tools for any of
1:52:50 these demos that you see right now
1:52:52 extend them out say okay that kind of
1:52:54 dexterity if you can get a robot to
1:52:57 train a robot to do that then we we we
1:52:59 will start to be able to train these
1:53:01 robots to do incredibly sophisticated
1:53:04 things without someone having to program
1:53:09 it
1:53:10 it's it's C it's cuckoo so yes yes that
1:53:15 is that is the implication there it's
1:53:18 like holy cra
1:53:21 crap just the speed of innovation I mean
1:53:24 think about that if if if you go from
1:53:26 having to figure out the physics of how
1:53:29 a robot with whatever its you know
1:53:32 dexterity and speed of actuators and you
1:53:36 know stickiness of rubber substance and
1:53:39 physics and gravity If instead of having
1:53:41 to think about all that and program in
1:53:44 all those those things manually as
1:53:48 humans the thing can just just figure it
1:53:52 out think about every complicated system
1:53:55 every complicated system ask the gummy
1:53:58 to do a Latin dance at a competition
1:54:00 level with another gummy bear oh my
1:54:03 God it's going to be too bad I'm not
1:54:06 going to do it by the way if I I assume
1:54:09 Corey you've seen the Tik toks all the
1:54:11 West Coast Swing all those dancers that
1:54:13 do the the uh the the uh the dances that
1:54:17 they kind of make up in real time those
1:54:19 those people people are
1:54:22 amazing all right what are you doing
1:54:24 time wise getting there getting there
1:54:26 happy happy Sunday night people hope
1:54:28 you're having a good time tonight watch
1:54:30 two Shapiro YouTubes today and the guy
1:54:33 from Google also I was Kink yeah so so
1:54:38 what Silver Fox is talking about is
1:54:40 David Shapiro so one of the guys that I
1:54:42 recommend you watching is um David
1:54:46 Shapiro on YouTube his channel is here
1:54:49 hang on hang on
1:54:59 it's come
1:55:02 on right right that one at 4ir which
1:55:08 stands for fourth Industrial Revolution
1:55:11 at
1:55:12 4. david.
1:55:15 Shapiro go watch his three most recent
1:55:19 videos and and then run down the rabbit
1:55:22 hole he's wild so previously inoperable
1:55:26 conditions will no longer necessarily be
1:55:29 so yeah and and you know things like
1:55:32 things like paralysis like they they've
1:55:34 already used um large language models to
1:55:39 this is just insane they did this with a
1:55:41 guy in Italy about I don't know four or
1:55:43 five months ago so I'm sure it's
1:55:45 Advanced since then they basically took
1:55:47 MRI scans of his brain
1:55:50 they said think about moving your legs
1:55:52 he was in a motorcycle accident so
1:55:54 severed spine can't move his legs
1:56:00 um so they used large language models to
1:56:03 read the MRIs when he thought about
1:56:05 moving his legs they then sent those
1:56:08 signals via a Bluetooth jumper they so
1:56:11 they bluetoothed around his injured
1:56:13 spinal cord with electrodes you know
1:56:15 going down to his
1:56:17 legs and and basically sent the large
1:56:21 language model
1:56:23 learned or the Transformer learned
1:56:26 patterns of thinking of moving his legs
1:56:29 down to
1:56:30 electrodes in in his lower spine and now
1:56:33 he's
1:56:34 walking so so it's like neurological
1:56:37 disorders injuries
1:56:42 um yeah like just the yes
1:56:48 yes like when I I don't know like how I
1:56:52 don't
1:56:53 know but it's but it's all happening
1:56:55 like it's all happening in parallel
1:56:57 right now like that that's the thing
1:56:59 what what what's fascinating about these
1:57:01 tools is
1:57:04 that the core technology for how chat
1:57:07 GPT does what it does and how the image
1:57:12 generation tools do what they do and how
1:57:14 the music tools do what they do and how
1:57:16 these brain reading things do what they
1:57:18 do it's all using the same [ __ ] it's all
1:57:20 just different data as the input so they
1:57:23 they basically take massive amounts of
1:57:25 data they run it through this thing that
1:57:27 Google invented called the Transformer
1:57:29 that embeds this data into some
1:57:32 mathematical Vector
1:57:35 space and then they take some other
1:57:37 input in the case of chat GPT the input
1:57:39 is your
1:57:43 prompt and then that prompt turns into
1:57:45 this mathematical probability thing and
1:57:47 it [ __ ] just goes into this Vector
1:57:49 space and pulls out the relevant data
1:57:51 and executes it and so like in the case
1:57:54 of chat GPT it'll write you your paper
1:57:56 or it'll program your code in the case
1:57:59 of controlling robots it's it's it's
1:58:02 been trained up on lots of robot code
1:58:05 and then you tell it what you want it to
1:58:07 do and it sort of goes into this Vector
1:58:09 space pulls some probability [ __ ] out
1:58:11 like how it does that I don't [ __ ]
1:58:13 know but it's all using the same core
1:58:17 technology so it's just about what data
1:58:20 do you train it on is what it can do
1:58:23 like when you go to the haen thing like
1:58:26 the the the
1:58:28 haen where do I have this open hang
1:58:45 on oh that's not what I wanted hold
1:58:48 please hold
1:59:00 please where am I missing something
1:59:04 [Music]
1:59:07 here I want to show you
1:59:14 something here it
1:59:18 is
1:59:23 so this is this thing where I'm speaking
1:59:26 Mandarin
1:59:29 [Music]
1:59:42 Chinese so how that translates me into
1:59:47 Mandarin is it's the same tool that you
1:59:52 know that translates me into Italian so
1:59:55 we've heard from a number of clients
1:59:57 that modular content is important to
1:59:59 them getting more efficient with how we
2:00:00 create content and how we leverage
2:00:05 content so it's not like it's not any
2:00:08 more complicated for this tool to learn
2:00:11 Mandarin Chinese than it is for it to
2:00:13 learn Italian it's just you take the
2:00:15 training data of
2:00:18 Mandarin and it learns Mandarin you take
2:00:21 the training data of Italian and it
2:00:22 learns Italian you take the training
2:00:24 data of brain waves and it does brain
2:00:26 waves you take the training data of
2:00:27 images and it does images now it's
2:00:30 listen it's way more complicated and
2:00:31 sophisticated than that but this is a
2:00:34 situation where data is data is data
2:00:37 like like once you get it into a data
2:00:39 form that these machines can understand
2:00:41 using this Transformer technology you
2:00:44 can stitch together multiple of these
2:00:47 models and get it to do [ __ ] like you
2:00:50 input text so it's using a large
2:00:51 language model for text understanding
2:00:54 and then it's got a thing called a
2:00:55 diffusion model for how it creates
2:00:57 images and this has got some voice
2:01:00 synthesis model combined with some
2:01:02 translation model combined with some
2:01:04 large language model and it's it's
2:01:06 combined with some animation model to
2:01:08 change the lips while I talk so it's
2:01:12 combining multiple models models
2:01:14 together but the same
2:01:16 core exercise of how you train the data
2:01:19 and how it does what it does is all
2:01:20 essentially the same [ __ ] so this is all
2:01:22 just about what we train the data on and
2:01:24 and what we have it
2:01:26 do and it's and that's why we're just
2:01:28 going to continue to see over the next
2:01:30 three years like
2:01:35 miraculous capability after miraculous
2:01:38 capability after miraculous
2:01:41 capability I I still find it miraculous
2:01:45 I can go to chat GPT and say write me a
2:01:47 paragraph and it just goes
2:01:51 make me an
2:01:53 image write a
2:01:56 song is that not blowing anyone's mind
2:01:58 other than
2:01:59 me oh chat GPT is overrated really have
2:02:03 you actually used
2:02:04 it it's not perfect okay all right check
2:02:09 back in a year um Kyle have you seen the
2:02:11 latest image generation using brain
2:02:13 waves I have not is there is there a
2:02:15 place to go to look at
2:02:17 them but I'm sure it's
2:02:20 face
2:02:21 melting hi my name's
2:02:24 Kyle hi Kyle they pay you for this
2:02:27 advertisement nope they don't they
2:02:30 should morals morals do you have morals
2:02:35 are you using this stuff morals or you
2:02:37 just in here to troll cuz I'm fine
2:02:39 either way just be creative in you're
2:02:42 trolling but I'm just curious if you
2:02:44 actually use this stuff or if you're
2:02:46 just like coming in and yelling people
2:02:48 that don't have qualific ifications
2:02:50 life's tough it's tougher when you're
2:02:53 stupid you know what I
2:02:55 mean you know what I'm saying
2:03:00 anyway um will you be getting elon's
2:03:03 brain chip no but I did
2:03:05 get um Sam alman's
2:03:09 um um world ID he scan Sam Alman scan my
2:03:14 retina that should piss you off want to
2:03:17 see want to see my world ID
2:03:28 L Lites like moral morals here is going
2:03:30 to be pissed off at
2:03:31 [Laughter]
2:03:34 this I think this is really cool but you
2:03:38 know as you said this is just an
2:03:42 advertisement so that
2:03:44 thing World ID verified at orb Anonymous
2:03:48 so this is my
2:03:49 blockchain
2:03:52 verified proof of human uniqueness based
2:03:56 on my retina that I scanned at a silver
2:04:00 ball in a clothing store in
2:04:04 Manhattan
2:04:06 so so
2:04:09 yeah
2:04:11 um speaking of the the Elon Musk
2:04:16 um uh what's it called neuralink um I
2:04:19 don't think I'll be doing neuralink
2:04:22 because I don't think we're going to
2:04:24 need to do surgically implanted stuff um
2:04:27 for things like that if you paid
2:04:29 attention to
2:04:33 um The Vision Pro announcement from
2:04:38 Apple you'll love this morals okay check
2:04:42 this
2:04:43 out this is this is exciting okay so so
2:04:46 neuralink what what what they're doing
2:04:48 with neuralink if you don't know what it
2:04:49 is they drill a hole in your brain or
2:04:51 multiple holes they put in an electrode
2:04:53 that's got 60 or 120 electrodes they
2:04:57 they've now built a robot that that can
2:04:59 insert these electrodes into your brain
2:05:02 tissue and avoid hitting any um
2:05:05 capillaries or or veins or you know any
2:05:08 blood flow so they basically steer the
2:05:10 these electrodes in between your blood
2:05:13 vessel so that's insane but it's
2:05:15 basically like instant connection to you
2:05:18 know to certain parts of your brain and
2:05:20 then they can you can it's a computer
2:05:22 interface that's instant but check this
2:05:24 out in the Vision Pro
2:05:28 goggles one of the ways they work
2:05:30 there's four cameras pointing at your
2:05:32 eyes and then in the in the goggles
2:05:35 themselves there's this chip called the
2:05:36 R1 chip that's doing realtime machine
2:05:39 learning artificial intelligence kind of
2:05:42 machine learning analysis of all these
2:05:44 sensors
2:05:46 right so the way the interface works for
2:05:49 Vision Pro is it's it's based on where
2:05:51 you look so so it'll pop up some icons
2:05:54 in front of you and then you look to an
2:05:57 icon and when you look to that icon
2:05:59 it'll highlight and then I think you do
2:06:02 this with your hands and there's some
2:06:03 other camera on the outside that when
2:06:05 you do this you select that icon but
2:06:08 here's something that they're
2:06:13 doing you you emotionally
2:06:17 know which icon you're going to click
2:06:20 before you look over and click on
2:06:22 it and so they're measuring micro
2:06:25 movements in your irises dilations in
2:06:28 your
2:06:29 irises to know which icon you're going
2:06:32 to look at before you click on it and
2:06:34 they start pre-loading that application
2:06:36 before you actually click on it so
2:06:39 Invision proos Hardware today that's you
2:06:44 know probably going to come out to
2:06:45 developers Before
2:06:48 Christmas is
2:06:50 essentially real time like like brain
2:06:55 analysis of what's going on in your
2:06:57 brain before you even know it so you
2:07:00 don't even necessarily
2:07:02 need um brain implants for that level of
2:07:08 response that's freaky
2:07:13 huh China prioritizing AI with the goal
2:07:16 of something ever since you start these
2:07:18 lives I use chat GPT every day from
2:07:21 cooking to coding
2:07:22 perfect again my goal here is not that
2:07:26 you love AI I don't give a [ __ ] if you
2:07:27 love AI or hate it but don't be ignorant
2:07:31 of it don't assume that because it isn't
2:07:36 perfect it's a waste of time I think I I
2:07:39 think I think that's the biggest mistake
2:07:42 that's what this Channel's
2:07:44 about China prioritizing AI with the
2:07:46 goal of intelligence AG
2:07:49 intelligentized Warfare Pentagon
2:07:52 warrants of
2:07:54 course go you want to be scared go look
2:07:58 up a company called
2:07:59 paler so paler is a military company
2:08:03 that builds War Logistics
2:08:06 software and they announced with pretty
2:08:09 decent Fanfare about six months ago
2:08:11 they've put a chat GPT like interface on
2:08:14 top of their War Logistics machine with
2:08:18 with their Vision [ __ ] so they can
2:08:20 basically take um imagery of a
2:08:23 battlefield and if they notice in the
2:08:25 battlefield that the bad guy moves from
2:08:27 that hill to that
2:08:29 Hill the system will now come up with
2:08:32 response scenarios in real time and send
2:08:35 those to generals so so yeah uh the
2:08:40 warship so like here's the deal kids
2:08:45 like all this chat GPT [ __ ] is new and
2:08:49 shiny Ai and machine learning and expert
2:08:52 systems have been around for [ __ ]
2:08:54 decades you don't think the military's
2:08:56 got crazy scary ass [ __ ] I was talking
2:09:02 to I was talking to someone at DARPA we
2:09:05 we did we did a proposal to DARPA for
2:09:08 for our storyvine technology and it was
2:09:11 I I came up with a use case for being
2:09:14 able to use our our video storytelling
2:09:17 thing to have
2:09:19 soldiers on the field um kind of they
2:09:23 could they could watch questions from a
2:09:26 psychologist and then they could answer
2:09:28 those questions and they could do it in
2:09:30 a in a non-real Time way right so they
2:09:32 could do it in a Time shifted
2:09:34 way and so I'm presenting this idea to
2:09:37 this guy from DARPA and I said and and
2:09:39 you know for example you could even have
2:09:41 like the the
2:09:43 um the
2:09:45 doctor if if the question is you know
2:09:48 are are you okay and and and the soldier
2:09:51 says yes I'm okay the doctor could look
2:09:53 at them and and know that they're not
2:09:55 okay based on and the guy from DARPA was
2:09:58 like oh no we can do a real-time
2:10:01 emotional analysis like we the doctor
2:10:03 doesn't need to to actually analyze that
2:10:06 we know if they're lying like this was
2:10:08 like 10 years ago so like the military's
2:10:12 been doing this stuff for ages I you
2:10:14 know I think I think what's different
2:10:16 here potentially is the speed with which
2:10:19 the advancements are going to come and
2:10:22 the ease with which so so historically
2:10:25 with machine learning you had to do
2:10:26 these kind of
2:10:27 specialized um expert systems so you
2:10:30 train it on very very specific data
2:10:32 what's happening with with chat gbt is
2:10:34 it's a generalized intelligence so it
2:10:37 knows lots of [ __ ] and so so now they're
2:10:40 taking these generalized tools and
2:10:42 applying them to expert systems and it's
2:10:44 it's accelerating things so so so yeah
2:10:48 yeah I and and I heard um in
2:10:51 um in Ukraine um I I think there's been
2:10:55 the first use of an
2:10:58 autonomous AI controlled drone that took
2:11:01 out an enemy right so like that like I
2:11:04 think already on the battlefield in in
2:11:06 Ukraine are
2:11:09 AI um
2:11:11 controlled autonomous war machines like
2:11:15 like it's happening right now like yes
2:11:17 so yes um should that be something we're
2:11:20 concerned about yes I mean the military
2:11:22 hopefully is on it and this is why it
2:11:24 gets weird about oh you know the
2:11:26 government should regulate this and shut
2:11:28 it down and da well the other countries
2:11:31 aren't going to stop this like we the
2:11:33 the the horse the
2:11:36 horses have long left the barn and and I
2:11:39 think you know meta open sourcing their
2:11:41 large language model um you know makes
2:11:44 this even less controllable than it was
2:11:46 before that happened so so yeah there's
2:11:49 some there's some scary [ __ ] out there
2:11:52 but anyway um it doesn't mean you
2:11:54 shouldn't play with this stuff even if
2:11:55 it's [ __ ]
2:11:58 scary how do I get this back wait I go
2:12:03 uh go here and I go here ah I go there
2:12:06 all
2:12:07 right investors bet Samsung smaller
2:12:10 memory chip Rivals SK hinx will be a
2:12:13 winner okay I don't
2:12:15 know SK hinx I don't know what that is
2:12:18 their smaller memory
2:12:20 chip I don't know what the implication
2:12:22 of that is lonus sounds cool I don't
2:12:25 know what it's got to do with AI didn't
2:12:27 Tom Cruz do a movie about that yeah
2:12:29 there's been a lot of movies about about
2:12:30 this
2:12:31 stuff uh uh uh uh uh would you like to
2:12:35 play a game Reddit can survive without
2:12:38 search company wait without search
2:12:41 company reportedly threatens to block
2:12:44 Google oh
2:12:47 interesting
2:12:49 I the the whole thing of like the future
2:12:52 of search is fascinating I mean I mean
2:12:55 you got to ask yourself if you're if
2:12:57 you're a thinking
2:12:59 person why did the company Google that
2:13:03 invented the Transformer that's making
2:13:05 all this possible why did they sit on it
2:13:07 for so
2:13:09 long cuz I think they knew if we make a
2:13:13 machine that just answers what people
2:13:14 wants
2:13:16 want our search engine business is
2:13:21 over
2:13:24 now their search their their market
2:13:27 share hasn't really gone down since all
2:13:29 this came out so I don't know that it's
2:13:31 playing out right now and Google it
2:13:34 sounds like what they're coming out with
2:13:36 with Gemini is going to be pretty
2:13:37 [ __ ] powerful so I wouldn't count
2:13:39 Google out right now but I think I think
2:13:42 search just has to
2:13:45 change I barely go to Google anymore
2:13:48 ever
2:13:56 AI military drone the valkyrie xq58a is
2:14:01 designed to deliver weapons or assist
2:14:04 fighter jets oh yeah I saw that thing
2:14:05 that was pretty cool doesn't it Fly
2:14:07 doesn't that fly in conjunction with a
2:14:09 piloted jet it's like a wingman for a
2:14:12 piloted piloted
2:14:15 plane someone said something about when
2:14:17 I was talking about David Shapiro
2:14:19 they're like oh no I thought you were
2:14:20 talking about Ben Shapiro
2:14:23 yuck that's
2:14:25 good yeah face to face is the only way
2:14:28 when troops come home I got a couple of
2:14:31 stories there oh interesting yeah yeah
2:14:34 exactly that's that's what we were
2:14:35 trying to do was can can we do something
2:14:38 like where where soldiers were on the
2:14:40 battlefield it never ended up going
2:14:41 anywhere so that was kind of a drag but
2:14:43 whatever it was
2:14:47 interesting
2:14:49 yes yes and yes too late ET is talking
2:14:51 about too late I assume about the
2:14:52 military
2:14:54 stuff what's the worldcoin
2:14:59 message I don't know what you
2:15:04 mean I don't know what you mean Silver
2:15:10 Fox the reason why you should play with
2:15:12 it Sherry D the reason why you should
2:15:14 play with it is because it's scary
2:15:17 exactly this [ __ ] this [ __ ] is way
2:15:19 scarier when you don't know what it is
2:15:23 cuz you can only imagine and you hear
2:15:24 stuff like oh the war machines and the
2:15:26 this and it's going to do that and then
2:15:29 you use it and you realize okay it's a
2:15:31 tool now if you sort of play out
2:15:34 scenarios and say wait if it's really
2:15:36 good at this kind of thing I wonder if
2:15:38 it could do this kind of evil thing over
2:15:40 here the answer is
2:15:42 yes like like right now these things
2:15:45 have safety guard rails and things like
2:15:46 that but the The Meta llama models that
2:15:49 they put out there they don't have any
2:15:51 safety guard rails on them so people are
2:15:53 out there right now building all sorts
2:15:54 of crazy ass nasty [ __ ] does that mean
2:15:58 you shouldn't learn how to use it and
2:15:59 learn where the boundaries are and learn
2:16:00 how to use it for good absolutely not
2:16:03 the more of us that understand how to
2:16:05 use this stuff and bend it even a little
2:16:08 bit toward the
2:16:10 good is is going to have an effect Sher
2:16:13 D you were a medic you were a medic in I
2:16:15 forget I didn't see what it was in in
2:16:17 Afghanistan in Afghanistan or one of the
2:16:19 wars face to face is the only way when
2:16:22 troops come home got a couple of stories
2:16:24 there fascinating oh medic during Desert
2:16:27 Storm wow wow amazing well thank you for
2:16:31 your service
2:16:38 Sherry you know I again back to the
2:16:41 virtual companionship thing I think that
2:16:45 there there are going to be countless
2:16:48 stories of things that
2:16:50 are difficult for people to talk
2:16:54 about that for whatever reason are not
2:16:56 comfortable seeking help and doing that
2:16:59 with people where they can do it with
2:17:02 something like pie or like a you know a
2:17:04 medically upgraded version of
2:17:07 Pi where they are going to be
2:17:09 comfortable with that conversation and
2:17:11 it's going to help a lot of people like
2:17:12 I really believe that YouTube meta AI
2:17:15 using brain
2:17:17 waves YouTube okay meta AI using brain
2:17:21 waves 240
2:17:23 timestamp let's go look at that
2:17:30 YouTube oh look it's me again I can't
2:17:33 get enough of me I'm
2:17:36 everywhere wait um stop shut up um meta
2:17:43 AI using brain
2:17:47 waves
2:18:00 okay So Meta
2:18:15 240 it's as if these AI systems are are
2:18:18 mirroring the neural intricacies of the
2:18:20 human brain and while other tools like
2:18:22 functional magnetic resonance imaging
2:18:24 can yield more detailed images the
2:18:26 superpower of meta's new AI decoder lies
2:18:30 in its speed being able to generate
2:18:32 images from brain activity in mere
2:18:35 milliseconds notably these visuals are
2:18:38 still somewhat lacking in fine details
2:18:40 but they can still accurately capture
2:18:42 the broader essence of what the human
2:18:44 eye is perceiving now for the bonus look
2:18:47 into what this techn ology is leading up
2:18:48 to by just connecting the dots of
2:18:51 progress made meta's future research
2:18:53 trajectory is becoming clearer for one
2:18:56 last year meta introduced a novel AI
2:18:58 architecture aiming to break existing
2:19:01 limitations in order to think and act I
2:19:03 don't know if that's the one that that
2:19:05 you were talking about or if that was
2:19:06 the time stamp but yeah that's that's
2:19:08 pretty amazing the fact that they're
2:19:09 able to do this in milliseconds I mean
2:19:11 you know again think about this so if
2:19:14 we're now at the point we can we can you
2:19:17 know read brain waves we can also write
2:19:19 likely write to brain waves so someone
2:19:22 who's blind you could you know this
2:19:25 stuff can see and you could potentially
2:19:29 allow them to see or at a minimum you
2:19:31 know in real time describe what they're
2:19:35 seeing and yeah I yeah so
2:19:41 so I
2:19:43 yeah it's it's uh
2:19:49 we're in weirdness people we are in
2:19:54 weirdness we're in post SciFi this is
2:19:58 we're in a post sci-fi situation here
2:20:00 people hi my name is Kyle my mind was
2:20:03 blown again today so with ours Kyle
2:20:06 [Laughter]
2:20:14 welcome
2:20:16 yeah the images are what I wanted you to
2:20:18 see the one on the left was the pick the
2:20:20 right is generated yeah and and
2:20:24 like those images that they just showed
2:20:26 like I've seen similar Tech to
2:20:30 that I've like the white papers for that
2:20:33 Tech I saw like 6 months ago and the the
2:20:36 generated images wasn't even close like
2:20:38 these are a hell of a lot closer like
2:20:40 again I'll just go back
2:20:43 to go look at what stable diffusion and
2:20:46 mid Journey images looked like like 18
2:20:48 months ago and look at what they look
2:20:50 like
2:20:51 today and take any technology where you
2:20:54 look at it today and you're like well
2:20:55 that's just crap that's horrible and
2:20:59 just play it forward 18 months to make
2:21:02 the same technological leap that images
2:21:04 did from when if you typed in gummy bear
2:21:08 it was you know kind of as bad as that
2:21:09 video as I did or Worse where today it
2:21:13 can create a
2:21:15 photorealistic gummy bear like that
2:21:17 level of regression is going to happen
2:21:19 with all of this
2:21:23 Tech the other thing that people are not
2:21:25 really talking about right now with the
2:21:26 exception of David Shapiro or just in in
2:21:29 general press like people like people
2:21:31 aren't even talking about chat GPT right
2:21:34 now but what's coming and what's going
2:21:36 to what's going to
2:21:39 what what you can already see glimpses
2:21:42 of in in chat
2:21:44 GPT is autonomous
2:21:49 action
2:21:51 meaning right now with chat jpt you type
2:21:53 in a prompt and you get back a response
2:21:56 and then you type in a response and you
2:21:57 get back a response and it's this very
2:21:59 manual kind of back and forth except
2:22:03 when you use the do
2:22:05 thing and you type in hey give me an
2:22:09 image and then it says generating images
2:22:11 what it's actually doing is it's writing
2:22:13 four distinct image prompts that then
2:22:15 generate those images you're not having
2:22:18 to write that if you use Advanced Data
2:22:20 analysis within chat GPT and you upload
2:22:23 a CSV and you say hey go um tell me
2:22:26 what's interesting about this data it
2:22:28 writes and executes its own python code
2:22:31 and then analyzes the answers and promps
2:22:33 itself to figure out how to give you the
2:22:36 answer that you
2:22:38 want so so these tools are starting to
2:22:40 take their own actions you give it an
2:22:42 imperative you give it a goal and then
2:22:45 it starts doing a series of things that
2:22:48 you don't have to control you don't have
2:22:49 to think about how to do it just does it
2:22:53 that's going to get more and more and
2:22:55 more and more
2:22:56 sophisticated and it'll be able to take
2:22:58 more and more kind of actions on your
2:23:01 behalf and that's where this [ __ ] can
2:23:04 get
2:23:05 scary
2:23:07 right there's there's there's an app out
2:23:10 right now that's making a lot of rounds
2:23:12 where it's like you can um take a
2:23:15 picture of someone and say go find all
2:23:17 the pictures of that person on the
2:23:19 internet and it will find them in crowds
2:23:21 like like again the surveillance uh the
2:23:24 surveillance industry's been doing this
2:23:26 for years but now all of these tools are
2:23:29 being revealed to us right because this
2:23:32 stuff's being democratized right and
2:23:34 it's it's being put
2:23:36 out largely because of chat GPT having
2:23:40 such massive adoption in such a short
2:23:42 amount of time all the other companies
2:23:45 are going we got to get our version of
2:23:47 that out out there cuz we've got to
2:23:48 compete right so business interests are
2:23:51 driving this but what it means is you've
2:23:54 got the entire investment world and the
2:23:56 entire development Community are all
2:23:59 pushing in the same direction so things
2:24:01 are going to start accelerating further
2:24:02 and further and
2:24:05 further is protein Source a legit resume
2:24:08 bullet I don't know what that means but
2:24:11 it's
2:24:13 awesome the images are what I wanted to
2:24:16 see oh the one on the right was the pick
2:24:18 the or the left was the pick the right
2:24:20 was generated then we'll be able to
2:24:22 communicate with
2:24:23 aliens they're already talking right now
2:24:25 about I think they're they're already
2:24:27 starting to do um communication with
2:24:29 animals with with whales Dolphins um I
2:24:32 think turkeys some someone did something
2:24:33 with turkeys or maybe I saw a guy that a
2:24:36 guy spent enough time listening to
2:24:38 turkeys that he he understands turkey
2:24:40 language but yeah you could just like
2:24:42 you translate you know English into
2:24:44 Mandarin Chinese if you have the right
2:24:46 input data for from dolphins or
2:24:49 whales you could create a translator
2:24:51 where I speak in English and it
2:24:53 translates it into whale and then the
2:24:54 whale responds and it translates it back
2:24:56 to English like that is not out of the
2:24:58 realm of possibility I would be very
2:25:00 surprised if the researchers that have
2:25:03 been trying to figure out um animal
2:25:07 language for for decades if they're not
2:25:10 all over
2:25:12 this what AI is suggested for getting
2:25:15 into scra the internet and use things
2:25:18 like
2:25:19 research you don't really need to scrape
2:25:23 the internet I would use perplexity doai
2:25:26 um it it searches across the internet
2:25:28 and it provides um citations and things
2:25:32 like that there's also some plugins in
2:25:34 chat GPT there's one called Web pilot
2:25:36 there's there's certain ones but also
2:25:38 chat GPT now is connected to the
2:25:40 internet so is Bard so's Bing now
2:25:45 because the commercial of those have
2:25:48 safety guard rails on them there might
2:25:50 be things you want to search for like
2:25:52 personal information about people it's
2:25:54 probably going to tell you to [ __ ]
2:25:57 off um but but the tools are out there
2:25:59 right now like I I don't know that I'd
2:26:01 recommend one there might be specialized
2:26:02 tool tools to do that but there's just
2:26:05 all the stuff connected to the internet
2:26:07 right now Pim eyes is the the photo tool
2:26:11 I used it to get Intel on people I'm
2:26:14 dating found one was married with five
2:26:17 kids there you go that that that pre-ms
2:26:20 the the harder conversation 3 months
2:26:22 down the line right amazing Pim
2:26:27 eyes my brother's new book is out just
2:26:29 got it it's incredible awesome what's it
2:26:31 about
2:26:34 Thelonius uh uh uh uh uh uh thing is on
2:26:37 a certain subject there may be 10 groups
2:26:39 researching yes they
2:26:41 collaborate but with this I don't quite
2:26:44 know what that was in reference to
2:26:48 the free version of Charity is not
2:26:50 connected to the internet they just told
2:26:52 me of Charity the free version of
2:26:55 Charity what do you mean charity I don't
2:26:56 know what that is all the data is there
2:26:59 and can be filtered in seconds is chat
2:27:01 GPT Pro worth the money perspective
2:27:04 matters I historically have said it's
2:27:07 really up to you I now think chat gp4
2:27:11 the 20 bucks a month is is the best deal
2:27:14 like as whoever said that was in here
2:27:16 earlier kind of trolling me do you get
2:27:18 paid for this
2:27:19 advertisement I think the 20 bucks the
2:27:21 chat GPT 20 bucks is the best deal going
2:27:24 right now I think it's absolutely worth
2:27:26 it I think I oh charity equals chat GPT
2:27:30 what what was what did he say
2:27:32 then um the free version of chat GPT is
2:27:36 not connected to the internet that's
2:27:38 correct yeah again I think it's I think
2:27:41 I think it's absolutely worth paying the
2:27:44 20 bucks for chat GPT Pro you get GPT 4
2:27:47 which which is dramatically better than
2:27:48 3.5 it's connected to the internet it
2:27:51 has Vision it has Advanced Data analysis
2:27:55 it it is uh it is connected to the
2:27:57 internet I think I said that it's got
2:28:00 plugins um it can generate images and
2:28:03 then in the app version of it it has um
2:28:08 it can talk and it can listen so it's
2:28:11 got like eight major modes right now
2:28:15 that are all at some point going to kind
2:28:17 of merge together into the same app but
2:28:19 all those different Mo modes in and of
2:28:21 themselves are like staggeringly good
2:28:24 values Advanced Data analysis on its own
2:28:27 is worth I don't
2:28:29 know 50 bucks a month a 100 bucks a
2:28:32 month dolly is at least worth 10 bucks a
2:28:35 month the image generation its ability
2:28:37 to read and understand images like I
2:28:39 don't even know the value of that yet
2:28:41 but it's
2:28:42 insane like just the ability to take my
2:28:45 whiteboard and instantly give me the
2:28:47 notes on it and and summarize
2:28:51 them that's
2:28:53 Insanity
2:28:55 um yeah the fact that it's connected to
2:28:57 the internet oh it's got custom
2:29:00 instructions so you can find tuna I like
2:29:02 I just I I think the 20 bucks a month is
2:29:05 is is the best deal going right now for
2:29:07 GPT Plus Beauty and Mischief the design
2:29:11 Alchemy of Blackman Cruz oh nice that
2:29:13 sounds
2:29:15 cool nice
2:29:18 do you like ml net I don't know ml net
2:29:21 is that military net ml net let me go
2:29:24 look um mln
2:29:29 net machine learning made for net
2:29:38 oh it's a machine learning framework I
2:29:40 don't know that that's that this is
2:29:42 probably more on the infrastructure side
2:29:44 so I I tend not to be on the
2:29:45 infrastructure side as much as I just
2:29:47 want to use the tools um although I'm
2:29:51 I'm potentially building something for a
2:29:52 Pharma company where I'm going to have
2:29:53 to geek out on that stuff but I think
2:29:55 we're probably going to develop the
2:29:57 initial thing we build on AWS not on
2:29:59 Microsoft but it it looks fine it looks
2:30:03 cool um I can't even imagine the
2:30:06 conversations we'd all have if we were
2:30:08 in the same room yeah exactly well
2:30:09 that's that's that's what the um that's
2:30:11 what the salon is about so come to the
2:30:14 salon so how we doing time wise I got to
2:30:16 go
2:30:18 um if you have
2:30:21 not uh seen this yet tonight
2:30:25 so that URL right there the salon. a
2:30:29 that's a community that I started last
2:30:31 December um where we have um bi-weekly
2:30:34 meetings so we meet every other Tuesday
2:30:37 so this coming Tuesday is the next
2:30:38 meeting we meet in person here in Denver
2:30:40 and then remotely everywhere
2:30:42 else and we have a an online Discord
2:30:45 community so join the Discord Community
2:30:48 introduce yourself if you if you want to
2:30:50 basically if you want to hang out with a
2:30:51 bunch of Irregulars if you want to hang
2:30:53 out with a bunch of weirdos that are
2:30:54 like okay I got to figure this AI stuff
2:30:57 out that's what this community is about
2:31:00 it's not about loving it it's not about
2:31:01 evangelizing it it's about what do we do
2:31:04 what do we
2:31:05 do how does it work what do I do now I'm
2:31:08 a copywriter should I be scared oh [ __ ]
2:31:11 yes I should okay great what do I do
2:31:13 here's where to start here's right it's
2:31:14 it's all people just trying to figure
2:31:16 this out from all walks of life from all
2:31:19 different technical backgrounds creative
2:31:22 backgrounds there's Architects there's
2:31:24 writers there's artists there's you know
2:31:27 Pate who works at Google on on on their
2:31:31 uh their tpus he's in there he runs the
2:31:35 the um the AI mechanics Guild we have
2:31:39 different guilds there's a writer Guild
2:31:41 an Art Guild business guilds I mean it's
2:31:43 it's pretty
2:31:44 slick um so yeah so so come hang
2:31:50 out but you can't be a douche the first
2:31:53 the first link that's a link tree the
2:31:55 first link talks about what the salon's
2:31:58 about and what our values
2:32:00 are and the values are things like
2:32:03 curiosity
2:32:05 exploration um inclusion empathy
2:32:08 generosity like share what you learn
2:32:11 don't just come in there and like you
2:32:12 know sit in the corner
2:32:14 and do that like teach people what
2:32:16 you're knowing Herod who's who's this
2:32:19 amazing designer is in there teaching
2:32:22 people how he does these amazing
2:32:23 animations right like it's just it's a
2:32:25 it's a generous community so if you're
2:32:28 stingy it's not the group for you if
2:32:31 you're generous and curious and
2:32:33 adventurous that's what it's
2:32:38 about wow well that was very very
2:32:41 intense
2:32:43 well we're trying to change the world
2:32:45 here people don't you
2:32:48 understand don't you get it
2:32:51 yet I think Ted Ted made this one Ted
2:32:54 looks at life the varsity
2:32:56 jacket oh man retr punk thank you again
2:32:59 for all your help you've been an
2:33:02 inspiration yeah retr Punk's
2:33:08 amazing all right everybody I think it's
2:33:11 time to wind this puppy
2:33:15 down
2:33:18 IR regulars present for share
2:33:20 [Laughter]
2:33:23 bear thank you Tobias good to see you
2:33:27 hair growth geek what's happening good
2:33:29 to see
2:33:30 you um Sherry D thank you Kyle thank you
2:33:33 thanks for being in here you guys are
2:33:34 awesome jrc good to see
2:33:37 you Tobias Joker what I was trying to
2:33:40 say is Imagine oh wait could you repin
2:33:42 that again whoever just pinned that have
2:33:44 a good night Irregulars
2:33:47 who else we got in here will good to see
2:33:50 you I'm in there trying to be weird and
2:33:53 tongue and cheek semi offensive no
2:33:56 that's no you're I I don't ever take
2:33:58 your stuff as offensive Joker what I was
2:34:00 trying to say is imagine all the
2:34:01 research on cancer trained on AI and ask
2:34:05 it
2:34:06 questions yeah and not just be able to
2:34:09 ask it questions like have it come up
2:34:11 with correlations between you know
2:34:14 what's what's causing it or how to cure
2:34:17 it right that kind of stuff's going to
2:34:19 start
2:34:20 happening Source Camp thanks Kyle good
2:34:23 night good night have good having you in
2:34:24 here cuno good to see you forever hooked
2:34:26 steo peace buddy retr Punk great to see
2:34:29 you digital Gods good night Kyle you
2:34:33 should think about letting some people
2:34:35 join you on live for Halloween November
2:34:37 30th yes oh so that's another thing so
2:34:41 so
2:34:42 I'm strongly considering that on uh
2:34:45 November 30
2:34:48 um the the anniversary of the launch of
2:34:50 chat GPT doing like a
2:34:53 24hour AI learning laboon yeah I think I
2:34:56 think we're going to have to to
2:34:58 coordinate with the Irregulars so I can
2:35:00 take some breaks in there um thonus the
2:35:03 Tailwinds of AI are starting to kick in
2:35:04 for Microsoft says Jeffrey Brent thrill
2:35:08 oh that's interesting well did you see
2:35:11 Sam Alman just said um chat uh open AI
2:35:15 just hit 1.3 billion in
2:35:17 Revenue 1.3 billion in Revenue that [ __ ]
2:35:20 hasn't been out and they haven't even
2:35:22 they haven't even really kicked off
2:35:24 their Enterprise sales yet right
2:35:29 like it's it's [ __ ] crazy so remember
2:35:32 remember like two months ago in here
2:35:34 someone was like they're spending
2:35:35 $700,000 a day they're going to be out
2:35:37 of money by the end of 2024 it's like I
2:35:39 don't think so and now it's like they're
2:35:42 making some butt
2:35:45 Bank it's crazy all right everybody um
2:35:49 great seeing you all uh I appreciate
2:35:52 y'all a lot for being in here I'm sorry
2:35:54 about my hair but it happens what are
2:35:55 you going to
2:35:57 do
2:36:00 peace