aiLL 8/7/23 - The Evolution of AI Tools and What's Coming Next

Description

Chapters

Transcript

0:11 foreign
0:26 [Music]
0:37 welcome to the AI learning lab happy
0:40 Monday
0:41 everyone excited it's Monday no no
0:44 really
0:46 I had a crappy day today you know even
0:48 those days where you just [ __ ] hiding
0:50 from the world you had one of those that
0:52 happens what are you gonna do
0:55 um welcome everybody this is the AI
0:57 learning lab we'll wait for some folks
0:59 to get in here
1:01 and uh we'll get this party started look
1:05 at this pot I started yeah yeah yeah
1:08 hello Nicholas what's happening how you
1:11 doing
1:15 everyone if you're new here we talk
1:16 about all stuff generative Ai and uh
1:21 I answer as many questions as I can so
1:23 if you have any questions about AI pop
1:24 them in the comments below I'll talk
1:26 about these here URLs beside my head in
1:29 a moment
1:31 and uh
1:34 yeah talk about that I'm trying to think
1:36 if there's anything
1:37 big going on
1:41 nothing nothing too dramatic nothing too
1:44 dramatic although I did spend
1:47 I spent enough time on Claude trying to
1:50 get a prompt working that I got rate
1:51 Limited
1:54 they were like you can't use it anymore
1:56 for three more hours
2:00 I just could not get this prompt to do
2:02 what I wanted it to do I tried out
2:04 Claude over the weekend it was great
2:05 yeah Heather it's really good hey Tobias
2:07 what's happening good evening Lauren but
2:11 shaking what's going down
2:13 looks like the hairs got a mind of its
2:15 own tonight as it's want to do on these
2:18 things about this time of night
2:20 I I knew I shouldn't have worn three two
2:23 pays at the same time damn it
2:28 uh uh good evening good evening good
2:31 evening Kelly from Tampa I got family in
2:34 Tampa
2:35 Florida
2:38 all right so let's uh let's get this
2:41 thing started oh we got looks like we
2:42 got a few few more folks piling on in
2:46 I was so frustrated it's been hours
2:49 trying to convert photo to Cartoon
2:51 Avatar jeez yeah I know you know that's
2:54 that's the thing about these tools the
2:57 image tools and these word tools is
3:00 like getting the prompts you can get the
3:03 prompts quickly to like 90 of what you
3:06 want but then if you wanna consistent
3:08 and replicable
3:10 it takes for [ __ ] ever because they
3:12 just decide to just change how they do
3:14 [ __ ] or or I don't know just the way
3:17 they work they're they're uh
3:20 they're uh they're a little frustrating
3:23 on on that way uh Positive Vibes what
3:25 are we learning here today anything you
3:26 want to learn
3:28 um I'm talking about generative AI that
3:30 that's that's what I know but I'm also
3:32 an entrepreneur of many many years I've
3:34 started I don't know 12 or 15 companies
3:36 I was an entrepreneur in the early days
3:38 of the World Wide Web so I was kind of
3:41 there at the birth of that thing and uh
3:44 started one of the first digital
3:45 agencies one of the first online
3:47 magazines and and you know kind of rode
3:50 that wave up to the.com Bubble through
3:52 the.com bubble down the other side
3:55 and so when I you know
3:58 started playing around with generative
4:00 AI about a year and a half ago I thought
4:02 oh this is another one of those and when
4:03 chat GPT came out it was like all right
4:06 everything games game over games changed
4:08 everything's changing this is definitely
4:10 like the early days of the web but way
4:13 more profound so that's kind of stuff I
4:15 like to talk about I can talk about
4:16 starting businesses I can talk about
4:18 marketing
4:20 um I can talk a little bit about how
4:21 this [ __ ] works but I'm not a
4:22 technologist I'm not a mathematician or
4:24 an engineer or researcher I'm more of a
4:28 product guy and a and a Founder so
4:30 that's what we can talk about if you've
4:32 got questions just pop them in the
4:34 comments I slept during the rise of
4:36 the.com but I'm not missing this but
4:38 yeah Heather yeah I know it's it it was
4:41 one of those things the the.com stuff
4:43 you had you had a year or two where you
4:46 could you know they have probably more
4:48 than that probably more like three three
4:50 years where you could you could kind of
4:51 jump in on that thing and then it kind
4:53 of took off and got ridiculous this
4:55 stuff
4:56 we we probably have like the end of this
4:58 year first half of next year to to you
5:02 know get an early foothold and it's just
5:04 gonna it's gonna take off so
5:07 um so we're still in those early days so
5:09 lots and lots of stuff to talk about
5:11 um let me show you what I got got here
5:13 okay if you're new to this AI stuff if
5:18 you if you're just scrolling through
5:20 Tick Tock and thought is that Tom Arnold
5:23 or fat Alec Baldwin
5:25 um I'm neither but I've been accused of
5:27 both
5:29 um
5:30 these URLs here are all different
5:32 flavors of chat GPT so the official PT
5:35 websites right here chat.openai.com so
5:38 if you haven't played with it yet
5:39 strongly struggling strongly encourage
5:42 you to go there right now and just start
5:44 playing while you listen to me ramble
5:45 all these other things are just
5:47 different things to explore so if you
5:49 have played with this one and you want
5:51 to see you know how some of these other
5:52 things do it
5:54 um Bing is connected to the internet
5:56 that's from Microsoft Claude can take
5:59 really massive prompts like you could
6:01 paste an entire novel into claude's
6:03 prompt window and then just start asking
6:06 questions about it Bard is from Google
6:09 it's connected to the internet
6:11 labs.perplexity that's metas llama 2
6:14 models that just came out and perplexity
6:17 is kind enough to host them for
6:18 everybody so you don't have to install
6:19 anything you can just go play with them
6:21 there
6:22 and then Pi super conversational and Poe
6:25 is you can play with lots of different
6:27 models and uh and you can make Bots
6:30 there so all of these things are worth
6:33 playing with feel free to take a
6:34 screenshot and go go grab that stuff the
6:38 other thing
6:39 that I put out here is
6:42 um
6:42 on the bottom those are those are a
6:45 couple of the different image generation
6:47 tools mid journey in my opinion is the
6:48 best Adobe Firefly is more commercially
6:51 safe and Leonardo's sitting on top of
6:54 stable diffusion so Leonardo's just a
6:56 really nice web front end a stable
6:57 diffusion but that basically represents
7:00 stable diffusion and then these things
7:02 here prompts.chat will teach you how to
7:04 prompt the large language models and the
7:06 prompt book here at openart will teach
7:08 you how to prompt those image things and
7:10 then future pedia is just a site with
7:12 lots of lots and lots of stuff
7:15 so with that
7:17 questions down below I will start
7:21 answering them I do my best to answer
7:24 all the questions
7:25 uh I like to have fun I make voices I
7:29 cuss occasionally
7:32 what else the hair is always randomly
7:36 terrifying or not
7:37 [Laughter]
7:40 all right
7:42 let's get gouted
7:44 let's get doing this thing you know what
7:46 I mean Kyle Kyle you know I got a
7:49 question for you what are your
7:50 qualifications hmm
7:53 what are your qualifications you have
7:54 qualifications
7:56 all right let's say let's see if I've
7:59 got any qualifications here
8:05 Aloha from Hawaii hello hello
8:10 Tantra nice nice name nice emojis you
8:14 got it working
8:16 welcome welcome welcome Dr J hello
8:19 I slept during the rise of the.com we
8:21 saw that one
8:23 can you put that last slide on again yes
8:26 sure sorry about that
8:28 there you go prompts.chat
8:32 openart dot a I slash prompt book that's
8:37 a really cool resource and then
8:39 futurepedia.io
8:42 and then these ones down below mid
8:44 Journey Adobe Firefly or generative fill
8:47 within Photoshop is also really good
8:49 and then leonardo.ai those are all ones
8:52 that I think are worth playing with
8:56 and then if you want to join a community
8:58 I created a community called the AI
9:00 salon so if you go to the salon.ai
9:02 there's links there to
9:06 our meetup group we meet every other
9:08 week and we have an online Discord
9:10 server we have a LinkedIn group
9:12 um so yeah I'm gonna do that there's no
9:14 cost for any of that and I don't get
9:15 paid for advertising all this [ __ ] this
9:19 is all so the reason I have this here
9:20 these are the things that I think if
9:22 you're just getting started with this
9:24 are worth
9:26 at least understanding are out there and
9:28 that they they kind of all do some
9:30 different things but if you're just
9:32 going to play with one play with the top
9:33 one the official chat GPT website
9:38 all right
9:41 um I think you're absolutely adorable
9:43 thank you woodsy I appreciate that hey
9:46 Kyle nice to see you who's that Austin W
9:48 Tick Tock hadn't told me
9:51 when you go online yeah I know it seems
9:53 like it's doing that weird I don't know
9:54 maybe I'm in some sort of ghost band
9:56 thing because I was I was a bad boy and
9:58 I got I've gotten booted three times but
10:01 whatever I don't know
10:04 but welcome welcome welcome
10:07 um what am I experimenting with I am
10:11 well I've got a major Pharma company
10:16 um who we do a lot of work with has has
10:18 said that they want to do three Pilots
10:20 with us
10:21 in
10:24 um three different areas two of which
10:27 I don't know how we would pull it off
10:29 but I'm gonna say yes to it because I
10:31 think we can pull it off
10:34 um but we got some stuff to figure out
10:36 so so I'm kind of looking at
10:39 um
10:40 what's what's the what are the
10:41 infrastructure up options right is do we
10:44 do Microsoft Azure do we do AWS Bedrock
10:48 I think it's called Bedrock is there is
10:50 there machine learning platform
10:52 um so I'm looking at infrastructure and
10:55 then I'm looking at prompting
10:57 um
10:58 I spent about three hours today trying
11:01 to dial in a prompt
11:02 um
11:03 that I just couldn't get it to I just
11:06 couldn't get it to behave it I would get
11:09 it right and then I would
11:11 try a new one and it would just be
11:12 completely wrong and then I would get
11:14 that one close and it was just you know
11:17 if you've played with these things at
11:18 all and tried to get them to return
11:21 consistent results that that is
11:23 sometimes the challenge
11:25 um ah thank you very much for the Roses
11:28 valve van nice I did press yeah the
11:32 Roses I like a massive Rose they smell
11:34 very sweet
11:36 bonjour crypto Joe what's happening
11:39 Geniuses have messy hair well then I
11:42 must be [ __ ] brilliant
11:44 [Laughter]
11:50 I don't know I actually did run a brush
11:53 through it earlier and it just decided
11:55 to do that
12:00 what are you gonna do
12:03 um what about privacy info we put in
12:05 well Mez Merlin so a couple of things
12:08 you should if you're if you're dealing
12:09 with chat GPT
12:11 let me show you something
12:19 okay so if you're a chat GPT go to your
12:22 settings
12:25 and there's a thing called Data controls
12:27 and if you go to data controls
12:30 you can
12:32 um turn off your chat history and
12:35 training so you're basically saying to
12:37 open AI don't train on this data and
12:40 turning off your chat history that's
12:42 when there there was I think it was
12:44 Samsung
12:46 some folks from Samsung you know left
12:48 things in their chat history and they
12:50 got hacked because of that
12:52 um so so this is kind of a safety
12:54 feature here so so you can turn that off
12:57 if you turn that off you lose your chat
12:59 history so there's also an export data
13:01 feature here so export the data before
13:04 you turn off your chat history
13:07 um
13:08 but what I would also say is if it's
13:11 particularly sensitive information that
13:13 can help
13:16 um but
13:19 if it's very sensitive stuff you know
13:21 don't do it here maybe do it on the open
13:25 AI playground so if you go to platform
13:34 platform.openai.com playground
13:37 um this is this your sort of
13:39 uh using chat CPT directly from the uh
13:43 the base model so this is kind of the
13:44 the API access this this is probably a
13:48 tad more secure
13:50 um I think it's the same basic terms of
13:52 service though but but
13:55 um this might be good and then the the
13:57 real safe thing to do if you if you're
13:59 really concerned about Safety and
14:00 Security is spin up your own server with
14:03 a large language model like llama2 that
14:06 just came out from meta that uh
14:10 put that stuff in open source with a
14:12 commercial license
14:14 um so so spin up your own kind of server
14:17 um and then and then that way you're not
14:19 sending anything anywhere you're you're
14:21 doing everything locally and if you want
14:23 to train
14:24 um these large language models on your
14:26 own data anyway then you'd probably be
14:28 doing something like Lang chain where
14:31 you're going to want to spin up your own
14:32 instance of a server
14:34 anyway so
14:44 I tell you that every time I join I know
14:47 it's very very sweet thank you
14:50 um
14:52 all right let's see let's see what we
14:56 got here
15:00 evening Mr Shannon Mr Decker good day to
15:03 you sir good day to you sir good day
15:07 is Apple ready to move back up well Paul
15:13 um they're apparently working on Apple
15:15 GPT I think they've created a new
15:19 I don't know if it's a coding language
15:20 scripting language a new framework
15:23 called Ajax for machine learning and and
15:26 AI stuff and and apparently they built a
15:30 thing called Apple GPT with it I think
15:32 Apple's going to be a major player
15:34 um
15:35 I doubt that Apple stuff is going to
15:37 look like chat GPT I have a feeling it's
15:39 going to be more like an autonomous
15:41 agent that you just ask it [ __ ]
15:44 like imagine if Siri didn't suck that
15:47 might be what we see from from Apple and
15:50 hopefully they Rebrand it because
15:51 because I think it's time to put her to
15:54 bed I think she needs a nap it's it's
15:57 been what 15 years or something like
16:00 that something like 14 years of her
16:02 trying to figure it out
16:08 uh can you make wait can you take terms
16:11 and conditions and have chat GPT
16:12 summarize them so you know you're
16:14 agreeing to I've done that yeah yes you
16:16 can and if if the if the TNC are too
16:20 long for Chachi BT go to go to uh Claude
16:23 you can do it there
16:25 um it's brilliant you just say based on
16:28 these terms and conditions tell me what
16:30 I'm agreeing to
16:31 and and you know put put the put the
16:33 most important things I'm agreeing to in
16:36 a short bulleted list and it will it
16:39 will pull the most important stuff out
16:40 it's really good for that stuff
16:42 these these tools
16:45 um
16:45 they can be hit or miss when you're
16:47 generating original stuff but when it's
16:49 analyzing a document and and you're
16:52 summarizing and and doing things like
16:54 that they're really really good
16:56 and it still blows my mind how fast it
16:59 does it right you give it this big long
17:00 legal document and say hey what's in
17:03 there and it sort of spins for 20
17:05 seconds and then goes quit
17:07 so yeah it's a really good use case
17:09 crypto Joe
17:11 uh
17:13 um
17:15 the company I work at doesn't have
17:18 chat history feature on our internal gen
17:21 AI instance oh that's interesting
17:24 yeah now I see why yeah exactly yeah
17:27 yeah exactly sure he needs to retire
17:29 Siri does need to retire thank you very
17:32 much uh I appreciate that woodsy
17:37 oh
17:39 all right love it when you go into
17:41 character I it's it's uh
17:44 it's more like the spirits just take
17:47 over the spirits take over they just
17:50 take over
17:52 um
17:53 it's never really planned I often think
17:55 about oh what was that character I did
17:57 and then I can never remember it so it's
17:59 all just temporary insanity
18:02 all right
18:05 um
18:05 what do you think of the chat GPT free
18:08 app
18:09 is it the best one is an AI tool well
18:13 um I don't know which chat GPT free app
18:16 you're talking about if it's the
18:17 official chat GPT app it's pretty good
18:22 um
18:22 it doesn't quite have all the same stuff
18:25 well last time I used it I haven't used
18:27 it in a while it doesn't have all the
18:29 same stuff as the website so I would
18:30 just check some things the website's
18:32 better for some things the app's good
18:34 for one thing that the app has that not
18:37 a lot of people are talking about is
18:39 it's got whisper voice to text in it
18:42 which is way better than Apple's voice
18:44 to text so if you want to do
18:46 transcriptions and things like that and
18:48 just sort of talk to chat GPT
18:51 um the app has that and it's good
18:54 so that's worth doing but make sure you
18:56 have the official open AI version of the
19:00 app because there's like half a dozen
19:02 Faker apps out there that are trying to
19:05 scam you so just go to openai.com and
19:09 find the app
19:10 link from the official site they have it
19:13 for iOS and for Android now
19:16 I use cloud today to rewrite a GitHub
19:18 readme and then ran it through gpt4 yeah
19:21 that that's a really good use case if
19:23 you haven't done it Tom one of the other
19:24 things you can do with those GitHub
19:26 readmes is is you tell it to not only
19:28 summarize it we say read it and give me
19:31 step-by-step instructions for how to get
19:33 this thing up and running and then if
19:35 you say don't move on to the next
19:37 instruction until I've completed the
19:40 previous one and then it'll do it step
19:42 by step and if you get stuck on a step
19:44 you can have it walk you through how to
19:46 do that step so pretty pretty swell
19:51 all right analyze terms of service good
19:54 idea that's a great idea it's really
19:55 good at stuff like that I noticed Claude
19:57 did reduce the size down nicely awesome
20:00 oh I see you you used Claude to kind of
20:03 synthesize it and then put it into chat
20:06 gbt to actually do the work that's
20:08 pretty smart yeah yeah it can it can do
20:10 that stuff pretty well
20:13 the thing you need to be careful with
20:15 there is with any of these things when
20:19 you have it summarize it if you're gonna
20:21 then use that summary as a basis for
20:23 other stuff just check the summary and
20:26 make sure that it didn't add anything in
20:27 there and if it did pull it out because
20:29 otherwise it'll sort of build on stuff
20:32 that might not be in the original
20:33 document
20:37 um Chachi PT wrote a longer more funny
20:40 version like an ad oh that's funny yeah
20:43 and Pie is really pie is very very
20:46 conversational like annoyingly
20:48 conversational if you want to have a
20:50 chat play with the PI app Pi
20:54 um they have an app and you can put it
20:56 into voice mode and just talk to it it's
20:57 pretty fun
20:59 um but if you want to get something done
21:00 it's a pain in the ass so
21:04 oh man me and Alive what happened to the
21:09 link reader
21:10 um
21:11 I don't know let's go look so this is a
21:14 plug-in
21:19 [Music]
21:22 um
21:26 link
21:29 access wouldn't oh there's a bunch of
21:32 them now
21:34 let's look up web
21:40 web requests website assistant
21:44 web Pilot's the one I've been using link
21:46 reader I think is just gone
21:50 um unless it's back
21:52 reader
21:54 no
21:56 I think link reader just went away link
21:58 so the the problem with the with the
22:01 some of the chat GPT plugins is that
22:04 um
22:05 they might have just been like tossed
22:06 together over a weekend by someone that
22:08 didn't really have a company you can't
22:09 really you can't really tell you know
22:12 are these things sitting on top of
22:14 actual
22:15 companies that are you know solid or are
22:18 they just people that had a brain fart
22:20 and said oh let me whip up a
22:22 a plug-in
22:26 yeah I don't see I don't see anything uh
22:29 link reader went away but I've been
22:31 using
22:33 um
22:35 what's it called Web pilot web Pilot's
22:37 pretty good so let me get rid of that
22:39 one get rid of box script I'll turn on
22:41 web pilot
22:42 so I could say
22:44 um oh what was it what was a
22:47 there's a McKinsey
22:51 let me go to Twitter
22:53 I'm sorry X let's go to x
22:56 I'm gonna go
22:58 McKenzie
23:01 foreign
23:08 the economic potential
23:14 where is it
23:16 do you have a link here do you have a
23:17 link do you have a link
23:20 here we go
23:25 why is there no link
23:28 hang on
23:34 um
23:34 [Music]
23:51 there we go
23:56 okay so here's this McKinsey report on
23:59 the economic impact of generative AI
24:01 that came out August 1st so I'm going to
24:04 grab that URL
24:05 we're gonna go back to chat GPT and then
24:08 I'm going to say
24:09 um
24:10 please
24:12 create a 10 segment
24:16 tweet thread
24:18 of this article
24:22 and then I'm going to paste in that URL
24:28 and then this should fire off that
24:30 plug-in which it did
24:38 it's off reading it now
24:44 see how well it writes us a Twitter
24:46 thread an X thread
24:53 generative ai's economic potential broad
24:56 utility
25:03 so it's doing pretty good on deep
25:05 learning evolution
25:10 generative AI could add 2.6 to 4.4
25:14 trillion dollars annually
25:16 so let's let's go back to the article
25:21 and do a find for
25:24 what does say 4.4
25:27 yeah 2.6 trillion to 4.4 trillion so it
25:32 actually got that right
25:34 yay Chachi BT didn't just make up [ __ ]
25:39 so this went off and actually read the
25:41 article so so there's web pilot so web
25:44 pilot works pretty good for that
25:47 um
25:53 actually you know it might be
25:54 interesting is
25:57 let's try bard
26:00 let's see how Bard does with the same
26:03 exercise because it's connected to the
26:05 internet I don't know if it can see
26:06 hyperlinks
26:08 um right a
26:11 10 segment
26:14 tweet thread
26:17 of this article
26:22 let's see if Bard can do it because
26:24 Bard's connected to the internet and
26:26 Bard sucks a little less than it used to
26:28 I'm not programmed to assist with that
26:31 well you're a jerk
26:33 freaking Bard you dumb dumb
26:39 ha weird
26:43 can you read
26:47 web pages
26:54 yes I can read web pages
27:00 oh I'm not programmed to write tweets
27:04 all right so let's see
27:07 summarize
27:10 this article in 10 short
27:16 paragraphs
27:18 less this is funny less than 140
27:22 characters in length I'll see that see
27:26 if we can pull it let's see if we can
27:27 Jailbreak old Bard to write us a tweet
27:29 uh
27:32 yeah here we go
27:43 I'm just a large language model so I
27:45 can't help you do that so one of the
27:47 things so if you're new to this stuff
27:49 one of the things that you'll see
27:52 frustratingly more than you want to is
27:55 I'm sorry Dave I'm afraid I can't do
27:57 that from 2001. uh that's that those are
28:02 the safety guard rails I don't know what
28:03 they're protecting from I don't know why
28:05 Bard can't just write some [ __ ]
28:07 tweets but apparently that's gonna
28:09 unlock the nuclear codes
28:12 so
28:16 um so that's that's why you run into
28:18 that stuff if you want less of that
28:21 um you can start playing with the open
28:22 source large language models and those
28:25 don't have those
28:26 um those safety guard rails on them
28:29 so
28:31 all right The Bard experiment still a
28:35 piece of poop
28:42 you're on fire tonight my hair's on
28:44 [ __ ] fire my hair should be set on
28:46 fire
28:46 that's like a it's like a bad stack of
28:49 kindling getting ready to go into the
28:51 match
28:53 I'm a retired doctor I don't know too
28:55 much about AI here's the thing you don't
28:57 need to know much about AI
28:59 the the thing about these things is
29:04 we're we're at a new phase of
29:07 the evolution of machine learning and
29:10 artificial intelligence where Chachi PT
29:13 represents this this inflection point
29:16 where we're on the other side of it now
29:17 where
29:19 it's just like a Google search box right
29:21 the when when you use chat EBT it's just
29:23 like having the Google search box there
29:25 you can ask it anything
29:27 and it can answer
29:28 a surprising amount of things
29:31 and so
29:33 um
29:34 you know
29:36 whatever your hobbies might be start
29:38 interacting with it just start playing
29:39 with these things and and play with it
29:41 with your hobbies and and uh
29:44 you know do do things like parlor tricks
29:47 or or you know learn enough about this
29:50 stuff that you can help other people
29:52 um do do things they want to do you can
29:55 start out really simple with with what I
29:57 call parlor tricks have it write you a
29:59 song have it write you a poem have it
30:02 I don't know write you an itinerary for
30:04 a vacation
30:06 um just start to get your head around
30:10 using the English language almost like a
30:13 programming language where where what
30:15 it's actually closer to is it's almost
30:17 like you've been given a really smart no
30:20 a really well educated intern that's not
30:23 very smart
30:25 it's a bit of a dum-dum you got to tell
30:27 it everything you gotta say no don't do
30:29 that do it like this
30:31 act like this act like that
30:34 but then tell me about you know
30:37 the
30:38 the 10
30:40 most significant events that ended World
30:42 War II and it'll do that because it
30:45 knows it all it knows everything and so
30:47 as an ex doctor it might be fun for you
30:48 to throw doctory kind of stuff in there
30:51 and see how it does there's a famous I
30:53 was watching
30:54 a tick tock this was months ago
30:56 but this doctor
30:58 um
30:59 you know had heard about Chachi PT and
31:02 and said well let me go in and and he
31:04 gave it a case of patient you know
31:08 a diagnosis or you know a set of
31:10 symptoms and it got that right and he
31:12 said okay that's interesting and then he
31:14 gave it a really difficult case and it
31:17 got that right and then he gave it the
31:19 most difficult case he'd ever had the
31:21 most sort of complicated nuanced case
31:22 he'd ever had and it got it right and
31:25 he's like oh my God this is amazing and
31:28 he said you know give me this the the
31:30 sources that you know helped you get it
31:32 right and said sure things here's those
31:34 sources and they were all fake none of
31:36 none of them existed and so that was the
31:39 when when they talk about these things
31:41 hallucinating it's basically they just
31:43 make [ __ ] up
31:44 um but it is kind of remarkable like for
31:47 for that differential diagnosis it got
31:49 this really obscure case correct
31:52 um but was not able to you know cite the
31:55 the specific
31:57 um you know literature that it that it
32:00 got that information from it's because
32:01 it's just not how the thing works
32:03 um there are companies right now
32:05 Building Systems that do that are
32:08 getting better at providing those
32:09 citations and if you if you play with
32:11 these tools that are connected to the
32:13 internet like Bing and Bard are both
32:15 connected to the internet
32:18 um they do provide
32:19 um references for some of the things
32:21 that they provide to you and that you
32:22 can actually go check so
32:24 um some of these do some of them don't
32:26 but all of them will confidently go here
32:29 you go here's your here's your sources
32:31 so just make sure you check that stuff
32:35 um I'm using chat GPT to write songs
32:37 when will it create actual music
32:40 um
32:41 so there's a bunch of projects right now
32:44 um I don't remember what any of them are
32:46 called because I haven't played with
32:47 them let me see if I can find some of
32:48 those
32:50 there was a new one that just came out
32:52 that looked pretty good and then I think
32:53 meta just came out with one to text to
32:57 music
33:01 audio dm2 so here's
33:12 [Music]
33:22 interesting
33:24 [Music]
33:29 [Applause]
33:32 I wish you so this is the latest and
33:35 greatest this one came out
33:37 yesterday
33:40 audio dm2 text to audio music speech
33:44 generation it's astonishing
33:46 and then there's a the hugging face
33:48 models there
33:50 then you've got metas what's metas
33:53 called
33:54 um let's see what metas is called metas
33:56 is called
34:01 music chair is it music gen
34:05 yeah music gen
34:11 the simple framework that generates high
34:14 quality audio and music from text-based
34:16 user inputs yeah just type what you want
34:18 you know yeah just type what you want
34:21 and I I I I I I I
34:25 at some point the synthesized voices
34:27 aren't going to suck
34:29 but we ain't there yet
34:32 um
34:33 [Music]
34:38 August 6th meta also just unveiled audio
34:41 craft
34:46 not too bad
34:48 [Music]
34:54 Adagio for Strings
34:58 [Music]
35:18 audio craft all right meta's audio craft
35:21 and then
35:24 dm2
35:26 are the two are the two kind of most
35:28 recent ones there's a bunch out there go
35:31 to
35:32 um
35:35 go to here go to futurepedia.io
35:38 and then just look up the look up the
35:40 music tools here there's a ton of them
35:42 there's there's one out there called
35:44 sound draw which which is kind of a loop
35:47 composer it's not you know generating
35:50 waveforms like all the waveform stuff is
35:53 you know very artifacty and a little a
35:56 little janky right now and they they
35:58 tend to only do 30 seconds or 60 seconds
36:01 um sound draw does full length songs you
36:03 can specify the length of the song you
36:05 can specify moods things like that but
36:08 it's just assembling Loops so it's not
36:10 as sophisticated
36:12 um but but future pedia will have
36:15 um we'll we'll have a fair amount of
36:16 that stuff
36:24 um
36:24 [Music]
36:25 so many AIS I don't know what's worth
36:28 paying for advice yeah it's great great
36:31 point
36:33 I can I can tell you what I pay for and
36:36 again I'm not affiliated with any of
36:38 these so I don't I don't make any money
36:39 if you do this or not so I don't really
36:41 care from that perspective
36:44 um
36:45 foreign
36:46 I pay for chat GPT so chat.openai.com
36:50 [Music]
36:51 I don't you don't really need to do that
36:53 anymore you've got Bing that's free
36:55 you've got Claude that's free you've got
36:58 Bard that's free you've got Pi that's
37:00 free
37:02 um there's a fair amount of things that
37:04 PO that are free there's some that are
37:05 paid for some that are free so you
37:07 actually have a lot of free options when
37:09 it comes to the large language models
37:11 when it comes to image generation
37:15 I pay for mid-journey
37:18 um I pay for I pay for the lowest level
37:21 of mid Journey which is like 10 bucks a
37:23 month I pay for the lowest level of uh
37:26 Runway ml the video Generation stuff
37:30 and it's that one's it's cheap on a
37:33 monthly basis I think it's 10 bucks as
37:34 well but you don't you don't get very
37:36 much you can't make a lot
37:38 um so the video ones get pretty
37:40 expensive if you want to make a lot of
37:41 video If you really want to play with
37:43 that
37:45 um
37:46 so chat gbt I pay for it the reason I
37:49 pay for chat EBT gives you access to
37:51 code interpreter to plugins to when when
37:54 they come out with new technologies the
37:56 the paid members get access to them
37:58 first so
37:59 um that's a big deal
38:02 at some point claude's probably gonna go
38:06 subscription
38:08 and I don't know if I'll pay for both
38:10 chat GPT and Claude I will probably
38:13 choose one or the other there
38:15 so I don't know that I would ever pay
38:18 for bard
38:20 and then llama is free so llama's free
38:26 um and in terms of like which ones to
38:28 actually experiment with
38:31 right now the the things that are kind
38:34 of ready for prime time are images and
38:37 words right so gpt4 at openai.com is by
38:43 far the best of all of these they all
38:45 have some different things that are that
38:46 are good about them but if you just want
38:49 good solid consistent output gpt4 at
38:53 open AI is the best and then in terms of
38:56 images mid journey is consistently the
38:59 best although stable diffusion with
39:00 their new sdxl model that just came out
39:05 um it's catching up they're close
39:08 the thing about stable diffusion it
39:10 requires you to be a bit more geeky but
39:12 it's it's there's so many developers
39:15 working on it there's all sorts of
39:16 different variations and flavors and
39:19 models that you can play with and and
39:21 that's that's why I put up
39:24 um on this list the the bottom one there
39:26 is leonardo.ai Leonardo is a web front
39:30 end on top of stable diffusion so
39:33 Leonardo and Leonardo is cool they give
39:35 you like 150 credits a day for free so
39:38 you can have a free account at
39:40 leonardo.ai
39:41 you may have to get on the wait list but
39:44 I think they've opened that up now
39:46 um
39:48 yeah but so what did I say CIA
39:52 mid-journey or pay for
39:55 zapier I pay for which is not an AI but
39:58 it's an automation tool that you can
40:00 leverage all these AI things I pay for
40:02 that
40:05 um
40:06 I pay for Google collab when I want to
40:08 do stable diffusion and and do some
40:10 geekier things
40:13 I think that's about it right now I'm
40:15 I'm trying to keep it very slim because
40:17 here's the deal this this is another
40:19 piece of advice on what what to pay for
40:21 don't pay for a year on anything right
40:24 now this stuff is changing so fast that
40:26 if you get locked in for a year and then
40:29 you know three months later someone
40:30 comes out with something you're like oh
40:32 [ __ ] that one's really good I want that
40:34 one well you've already paid for the
40:36 year that's going to be a drag so I
40:38 would go month to month with all this
40:39 stuff right now
40:45 all right alternatives to link reader
40:48 web pilot
40:50 web pilot could sir is this 3.5 or 4
40:54 I was showing when I was showing
40:57 um
41:00 when I was showing web pilot that's gpt4
41:03 so GPT 3.5 is super fast generally how
41:07 you can tell is if it's ripping fast as
41:10 3.5 So like
41:12 um write me and iambic
41:17 pentameter
41:21 um epic poem
41:24 about
41:26 a chubby dog
41:31 so see how fast that's going generally
41:34 that's
41:36 gbt 3.5 super fast GPT 4 slower but more
41:40 powerful
41:42 and then if you're doing things like
41:43 code interpreter or plugins that has to
41:45 be gpt4
41:46 or it doesn't have to be but that's the
41:48 only place it's available
41:53 anyway to get it to post the tweets no
41:56 not right now not right now again all
42:00 these things understand that the current
42:03 state of the art
42:06 give it a little bit of Grace
42:09 so
42:10 these are the most powerful tools ever
42:13 put in the hands of mere mortals so
42:15 we're starting from the Baseline is very
42:18 high
42:19 in terms of what these things actually
42:21 do but in terms of the software that has
42:24 been wrapped around them it's all very
42:26 spit and duct tape right now none of
42:28 these things are very sophisticated
42:31 I would think that in futurepedia.io
42:34 there are probably some applications
42:36 that have been written that use a lot of
42:38 these tools that may have some more
42:39 sophisticated publishing capabilities
42:41 but from the from the core companies
42:43 that are doing this
42:45 it's it's pretty it's pretty weak right
42:47 now
42:49 you know a lot of these things are
42:50 really just
42:52 um
42:54 like like
42:55 I don't know Market
42:57 uh Market
43:00 validation right so so open AI is not
43:04 making money on chat GPT they're making
43:06 their 20 bucks a month but I'm sure it's
43:08 costing them way more than that for us
43:10 to use it
43:11 um they're gonna make their money off
43:12 the apis in the back end so a lot of
43:14 these front-end things these guys are
43:16 building are really just to get you hot
43:18 and bothered about oh I want to put that
43:20 in an application or I want to train it
43:21 on my own data or I want to you know all
43:23 that sort of stuff thank you Mike
43:25 thanks for the rose rose
43:28 arose by any other name would smell as
43:32 sweet
43:36 [Music]
43:39 what the hell
43:41 laughing my ass off I don't know what
43:43 you're talking about Heather what did I
43:44 show
43:45 wait I can go live
43:50 just sharing prompting and chatting wild
43:52 yes interesting
43:55 foreign
43:58 yes Jesse if you if your next question
44:00 is what are my qualifications I
44:02 absolutely have none and you know I
44:04 would encourage you to go find someone
44:06 that knows what they're talking about
44:08 you can go live doing whatever you want
44:10 there's the guys going live right now
44:12 with his he stuck an iPhone on his
44:14 robotic lawnmower and he's making like
44:17 1200 bucks a week from from his stupid
44:20 lawnmower
44:23 oh
44:24 oh my God Tick Tock is so insane ah
44:28 [Laughter]
44:30 I'm not making [ __ ] you money yet on
44:32 this one
44:33 I don't even think I've made enough
44:35 money yet there's a hundred dollar
44:37 minimum uh
44:39 you can't pull any money out of your
44:41 account I've been doing this for like
44:43 four months now I don't think I've hit
44:45 my hundred dollar minimum yet
44:48 but that ain't why I'm doing it so
44:50 for what that's worth
44:53 but yeah you can go live and talk about
44:54 anything doesn't [ __ ] matter
45:01 all right
45:06 tweets don't have a character limit
45:07 anymore really at all
45:11 is that true Crystal Archie or h e h e
45:16 achie
45:17 Crystal actually
45:21 feels like there's an R missing from
45:23 your name love it when it says
45:25 inappropriate messages
45:28 can't write and it's like recipes yeah
45:31 you can bypass safety guard rails by
45:34 using slang yeah you can you can get
45:36 around it sometimes there's a paper just
45:39 came out that there's a little chunk of
45:40 code that you can put at the end of a
45:42 prompt that will jailbreak any of these
45:44 things so jailbreaking just basically
45:46 turns off the safety guard rails so
45:49 apparently there's there's some some
45:51 code that they didn't put it in the
45:53 paper itself but they showed what the
45:54 syntax of it was
45:57 um when I asked Bard for references
46:00 Bard does that all the time does what
46:03 all the time does shitty fake references
46:06 yeah I'm sure all these things are just
46:10 we're not at the place yet where where
46:13 these things are checking their own work
46:15 so it's on us to check their work we'll
46:17 get there these things will eventually
46:19 have
46:20 um sort of validation layers you know to
46:23 validate that what they wrote is
46:24 actually useful uh they just don't have
46:26 it yet so give it a year I'm glad I
46:29 haven't wasted my time on Bard yeah Bard
46:31 right now is more like I have it on the
46:34 list as a thing to see like apparently
46:36 it's pretty good at coding
46:38 um it is going to be rolled into all of
46:41 the Google office you know Google Docs
46:44 and Google Sheets and things like that
46:46 it is going to be rolled into
46:49 um Google search they're experimenting
46:50 with that so I have it on here because
46:52 it's at least not
46:54 horrifically bad now and because it can
46:57 read images and it's connected to the
47:00 internet it's worth at least
47:01 experimenting with
47:03 um but I think Bing is still way
47:04 stronger than bard
47:07 um the the top three Chachi PT Claude
47:10 and Bing are the are the ones that are
47:12 probably the ones worth spending the
47:14 most time on
47:16 um
47:17 but pay attention to pie
47:20 that one
47:21 they just raised 1.3
47:24 billion dollars so so that little teeny
47:29 tiny URL and it's it's a it's a just a
47:32 very thin thin site right now it doesn't
47:34 it doesn't do much at all but it's I I
47:38 have a feeling they're gonna make some
47:39 noise
47:42 um what certification from udemy or
47:44 courses would be helpful
47:46 um
47:47 I don't know I am this this early in the
47:50 industry I'm not a fan of courses
47:52 although Google just came out with 10
47:56 um that include courses on generative AI
47:58 things like that if you just want to get
48:00 your head around it it might be worth
48:02 looking at the I think meta's got a set
48:04 of classes
48:06 um Google now has a set
48:08 um
48:11 if if if you want to get into the the
48:15 data science side of this the machine
48:17 learning side and the engineering side
48:19 there's probably lots and lots of
48:21 courses to take
48:23 but if you're trying to find courses on
48:25 how to use generative AI for marketing
48:27 or to run your business or this or that
48:30 all of that stuff has been created
48:33 within the last nine months and these
48:35 things are changing so so fast that
48:37 anything where you take a course by the
48:39 time you take the course it's it's going
48:42 to be it's going to be out of date
48:44 so what I strongly encourage right now
48:47 is just start getting curious and start
48:49 exploring with with the tools themselves
48:51 learn about prompting not necessarily
48:53 because prompting is going to be a big
48:55 important job in the future I don't
48:57 think it is I think these things are
48:59 going to start prompting themselves
49:00 we're just going to give them goals and
49:02 they're going to go off and do their own
49:03 [ __ ] and and come back they're going to
49:05 be like little golden retrievers they'll
49:07 run off and find a [ __ ] and bring it
49:09 back and can I get you some more can I
49:11 get you some more
49:13 um
49:13 but I think learning how to prompt right
49:17 now and learning what these things make
49:20 possible is really really important it's
49:23 a really important skill so
49:26 um
49:27 so I don't know so I don't have good
49:29 good advice on on any courses uh because
49:32 I I I don't know it's not so much that
49:35 it's a waste of time it's that the
49:38 amount of time that you're gonna put in
49:40 to have someone else teach you
49:43 what they learned three or six or eight
49:47 months ago
49:49 I think it would be better spent
49:50 teaching yourself how to do it just so
49:53 you can get the muscle memory of what
49:54 what these things are what they do well
49:56 what they don't do well if you're just
49:59 using someone else's
50:00 prompting and and philosophy
50:03 um you might discover better ways of
50:05 doing something than they did because
50:07 you just think differently right you're
50:08 coming at it with your world experience
50:10 and and you might use these things very
50:12 very differently so
50:14 I know that's kind of a shitty answer
50:17 but you know
50:19 yeah no what are you gonna do what are
50:22 you gonna do you know I'm just I'm just
50:24 an old guy
50:26 you know I just come on here and I I go
50:29 live and I just I just talk about
50:31 prompts prompts and you know I point to
50:35 sites not really much going on here you
50:38 know so
50:39 I wouldn't listen to me
50:42 I give advice on hair styling
50:46 would you like some of that I'm happy to
50:47 happy to give you advice on hair styling
50:55 all right I started with gpt3 in May and
50:59 then just late oh late last week
51:01 upgraded to three 3.5 or to GPT 4 for 20
51:05 bucks a month
51:07 I pasted my MRI result and asked it to
51:09 explain it to me like I'm five years old
51:11 it was fun yeah it's pretty it's pretty
51:13 uh
51:14 it's pretty good at that stuff it's
51:17 amazing dude I have that shirt I wore it
51:19 a lot before I figured out it's skulls I
51:21 know I like the skulls aren't they cool
51:25 why not wear skulls let people figure it
51:27 out
51:30 I've got Pharma clients I wear this to
51:32 Big Pharma meetings I had one today
51:36 [ __ ] them
51:40 farmer People Are People Too they're
51:42 just you know
51:44 a little more conservative than
51:46 some of the rest of us
51:49 I use GPT daily for a project for
51:52 recipes from recipes to VBA coding yeah
51:55 yeah Visual Basic stuff
51:57 um yeah it's
51:59 the the thing to do so so again if
52:02 you're new to this stuff
52:05 um
52:06 one of the things that you can run into
52:08 that that can be really frustrating is
52:11 after you get done doing your parlor
52:13 tricks and after you've had your Kevin
52:15 McAllister moment which is
52:17 remember home alone uh you have that
52:21 moment where you're like holy [ __ ] this
52:22 stuff's amazing
52:25 once you sort of crack that nut a lot of
52:28 times what I'll find is I'll sit looking
52:30 at that blank box and I get sort of the
52:33 the chat equivalent of white page
52:36 syndrome I don't know what to do I don't
52:38 know what to type in like if you've got
52:40 a box that can do anything it's like
52:41 well what do I do with it
52:44 one of the things that can help is
52:46 rather than think about what should I
52:48 use AI for flip it around upside down
52:51 and say what's something in my life that
52:55 I need to deal with or solve right so it
52:58 might be MRI data or it might be you
53:02 know you've got a bake sale coming up
53:04 for the school fundraiser and you want
53:06 to organize it or come up with recipes
53:07 or whatever it might be pick a project
53:09 either for yourself or for someone else
53:12 or for work
53:14 and then say okay
53:16 you know how you would solve that
53:18 problem see how you could use these
53:20 tools to solve that problem
53:22 and that's a way that you can very very
53:25 quickly figure out what it's good for
53:27 what it's not good for where you have to
53:29 work around it where you realize I'm
53:32 spending more time dialing in the prompt
53:35 than I could have just written this
53:36 thing like I've done that a couple of
53:38 times or I'm like an hour into getting
53:40 the prompt right and I'm like oh [ __ ] it
53:41 and I just go write the paragraph
53:44 you know
53:47 um so so that can help get you unstuck
53:50 if if you're trying to figure out what
53:52 to use it for
53:53 uh meta released one that creates music
53:55 yep they just released that I went over
53:57 that there's old Python scripts that
53:59 make beats and avai makes Beats also I
54:03 think yep
54:05 um what's it do if you type
54:08 the word doom I don't know
54:11 I do not know when will it be able to
54:14 pick a hit song
54:16 and a [ __ ] song
54:20 well
54:24 you could certainly train a large
54:26 language model here are hit songs
54:29 and here are not hit songs and then
54:33 you could have it analyze songs and say
54:36 these ones are more likely to be hits
54:39 than those ones
54:40 the problem with that is it doesn't
54:43 account for new [ __ ]
54:46 so I don't know I don't know that it
54:48 could be
54:49 predictive on kind of future stuff right
54:53 like you know I think about someone like
54:55 Billy eilish where
55:00 you know she sort of seemingly emerges
55:02 out of nowhere and she's got this really
55:04 unique style and if if you sort of
55:07 looked at
55:08 her work before she got famous and sort
55:11 of compared it with
55:12 what was popular at the time you'd go no
55:14 that's not the same like the chances of
55:16 that being popular are not good right
55:19 because it's so different and yet
55:22 you know the combination of her and the
55:25 music and the production and the story
55:26 and how she performed and her
55:29 relationship with her brother whatever
55:31 was the magic
55:33 that allowed her to you know sort of hit
55:37 escape velocity
55:39 um I don't know that I don't know that
55:40 AIS are going to be able to predict that
55:42 but
55:43 I suppose they could predict it as good
55:45 as you know probably mathematically
55:48 about as good as an A and our person
55:51 could
55:52 um
55:53 but but even those are different now
55:54 like the whole music industry is just
55:56 just different because it's not just
55:57 about the music it's about do you have a
55:59 following and and things like that so
56:02 but I would think you could train you
56:04 could train an AI to identify likely hit
56:07 songs and it would probably be in the
56:09 neighborhood but I just I don't think it
56:11 would it would get the outliers
56:13 the the unanticipated hits
56:18 because that's just human beings being
56:20 weird like what what makes something
56:22 popular why does that movie a hit and
56:24 that one not a hit
56:26 if they're almost identical and they
56:28 have similar kinds of stars I don't know
56:30 just because we're humans are [ __ ]
56:33 weird
56:35 um Yani Yani would like the strings to
56:37 that when will it be able to okay
56:40 um I love mid Journey so much but until
56:42 I'm ready to actually do more with it I
56:45 paused my sub yeah and and that's I I
56:50 thought Heather is
56:53 I'm finding myself doing the same thing
56:55 where I'm kind of porpoising I'll like
56:56 come up I'll do a whole bunch of stuff
56:58 with mid journey and then I kind of go
56:59 underground and I don't do it for a
57:01 while
57:02 um I am using the large language model
57:04 stuff I am using these things pretty
57:06 consistently for me right now it's chat
57:09 GPT and Claude are the two biggies that
57:11 I'm using all the time
57:14 um the rest are just more kind of
57:15 interesting things for me to play with
57:18 um Claude because it's got the big
57:20 context window and it's pretty good and
57:22 then chat GPT because it's just the best
57:24 and it's got code interpreter I'm not
57:27 using Code interpreter a lot but when I
57:29 need it man is it [ __ ] brilliant it
57:31 really is amazing
57:34 um
57:35 I have a cool mid-journey designs but no
57:37 no idea what to do next
57:39 um you could ask Chach EVT what to do
57:41 with them you can start an Etsy store
57:44 you could
57:45 um I know I know someone who's making
57:48 um
57:49 uh what's it what's it called uh not
57:52 repeatable
57:54 seam seamless patterns in mid-journey
57:57 and then putting them on pattern sites
57:59 to sell patterns so you could do
58:02 patterns
58:03 um you could do images you could do
58:04 t-shirts I'm working with a company
58:07 right now that does uh put is putting
58:10 together
58:11 t-shirt Generation stuff pretty
58:13 interesting stuff
58:15 you could also just take the images if
58:18 you like them and go to a
58:19 print-on-demand company and just you
58:20 know get a mug
58:22 so
58:25 all right
58:26 are the image ones free though I thought
58:28 they were capped at like five to Fifteen
58:30 images a day yeah they are
58:32 um
58:33 what do I have up here
58:35 Leonardo's five to Fifteen a day
58:37 mid-journey you have to pay for
58:39 absolutely Firefly you have to have
58:43 um
58:44 a uh uh a subscription with
58:48 um what's that one called
58:52 there's one that is truly free that I
58:54 can't figure out how in the hell they're
58:55 doing it for free
58:57 Blue Willow
59:00 yeah Blue Willow there's there's you you
59:03 have to build stuff in Discord just like
59:06 mid-journey
59:08 um
59:10 but it's free so do you slash imagine I
59:14 think it is Imagine
59:22 um
59:23 [Music]
59:25 Eagles soaring
59:36 oh that's interesting now this now
59:38 they're saying so so blue willow
59:42 um is free but they're saying support us
59:44 with any amount you can just type slash
59:47 subscribe
59:48 so blue blue Willow's free there I just
59:51 type Eagle soaring
59:57 you know not not great you know this is
1:00:00 sort of
1:00:02 big big like in its quality maybe a
1:00:05 little bit better than Bing I think it's
1:00:06 built on top of stable diffusion
1:00:09 um
1:00:11 but there are probably some other free
1:00:13 ones out there I just I just don't know
1:00:14 them all
1:00:18 um
1:00:19 [Music]
1:00:25 Ashley Smith Babylon
1:00:29 has been fired the world looks in danger
1:00:31 87 seconds to Doomsday that's good
1:00:36 uh very good Leonardo yeah Leonardo is
1:00:38 quite good it's just good interface
1:00:41 it's I mean it's a little complicated
1:00:43 all these things are a little
1:00:44 complicated right now but if you wanna
1:00:46 if you want the power you need to you
1:00:48 need some interface in there
1:00:50 Firefly beta is part of Adobe Creative
1:00:53 Cloud sub yep that's correct
1:00:57 where to get web pilot
1:00:59 um
1:01:01 okay so that's it's a
1:01:03 so and I think you just said you just uh
1:01:05 upped your
1:01:08 um subscription to GPT plus so if you
1:01:11 want plugins and if you want code
1:01:12 interpreter and they they have this new
1:01:15 thing called custom instructions you
1:01:17 have to be a chat gbd plus member and
1:01:21 then under gpt4 you'll see those extra
1:01:23 things but notice they say beta you
1:01:26 actually have to turn these on so what
1:01:27 you do
1:01:28 is you go to your settings
1:01:32 and then within your settings there's
1:01:34 beta features and you actually have to
1:01:36 turn them on
1:01:38 um so custom instructions plugins and
1:01:40 code interpreter and then
1:01:43 um
1:01:44 then then you'll see them there and
1:01:46 custom instructions is kind of fun so
1:01:48 I'll show you here so once you enable
1:01:51 custom instructions
1:01:53 you have sort of
1:01:55 what would you like chat GPT to know
1:01:57 about you and then how would you like it
1:02:00 to behave so you could put in here you
1:02:03 know you're a marketing manager and I
1:02:05 want you to you know only write SEO
1:02:07 optimized copy or whatever it might be
1:02:09 but you can have some fun with it so
1:02:11 this is this is one I wrote I'm a master
1:02:13 of thespian and A pompous ass I have a
1:02:15 large vocabulary and I relish sounding
1:02:17 smarter and more dramatic than people
1:02:18 around me
1:02:20 and then no matter what anyone asks
1:02:21 start out with a pretentious statement
1:02:23 that draws from the plot of a classic
1:02:25 Broadway play follow up with a
1:02:28 declarative statement uh with an article
1:02:30 formatted using markdown complete with
1:02:32 title subtitles and bulleted lists no
1:02:35 matter how trivial the question is so I
1:02:37 can turn this on
1:02:39 and save it and now no matter what I
1:02:42 type in the chat EBT it's going to
1:02:44 respond as this pompous ass actor so
1:02:46 I'll say
1:02:48 um why
1:02:51 why are fire
1:02:54 hydrants red
1:02:57 I don't think they are anymore but they
1:03:00 were when I was a kid oh the Vivid Hue
1:03:03 of the fire hydrants reminiscent of the
1:03:05 sanguine curtains that of the
1:03:08 proscenium of Arthur Miller's Death of a
1:03:11 Salesman just as Willie loman's
1:03:14 Shattered Dreams are painted with
1:03:15 Strokes of tragedy so too is the design
1:03:17 of our Urban landscape marked with the
1:03:20 intention and function the red paint of
1:03:23 fire hydrants a Chronicle of necessity
1:03:25 and recognition
1:03:26 the birth of a color
1:03:31 so so that's kind of fun with custom
1:03:33 instructions and like no matter what I
1:03:36 ask in a fight you know however many
1:03:37 chats that I do it'll always you know
1:03:40 respond like that until I turn that [ __ ]
1:03:42 off which is fun until you forgot you
1:03:45 had it turned on and you try to do
1:03:47 something in front of a client
1:03:49 I've done that before that was pretty
1:03:53 funny
1:03:54 um
1:03:58 all right
1:04:00 where do you see AI taking us in 10
1:04:02 years any idea I don't so
1:04:07 foreign
1:04:14 I feel like
1:04:16 I have a good handle on the next two to
1:04:18 three years and I think what that looks
1:04:20 like is just a lot of disruption a lot
1:04:23 of
1:04:24 a lot of
1:04:26 disruption in business disruption in
1:04:29 jobs
1:04:30 a lot of displacement
1:04:34 I feel like
1:04:36 five years things start to normalize but
1:04:39 they normalize
1:04:41 in a way where we've got these tools
1:04:43 that do all of the mundane [ __ ] of today
1:04:46 for us
1:04:49 and then 10 years I don't have a [ __ ]
1:04:52 clue because it's like
1:04:55 what it what do we do
1:04:59 if if if these tools take all of the
1:05:02 complicated parts of society
1:05:06 from insurance claims to filing taxes to
1:05:09 building websites to coding games to
1:05:14 making movies
1:05:17 and they just get automated
1:05:20 what do those people do
1:05:23 and I and I think what it starts to look
1:05:26 like is people start to reconnect and
1:05:29 people start to be more self-expressed
1:05:31 and like I really do think these tools
1:05:33 give us the opportunity to raise
1:05:36 ourselves to another level right these
1:05:39 are the these tools are ability
1:05:40 amplifiers so if these tools take all of
1:05:44 the low-level soul-crushing [ __ ]
1:05:47 that most of us spend way too many hours
1:05:50 a week doing
1:05:52 if that were just done for us what would
1:05:55 you choose to do
1:05:58 that could look pretty compelling and
1:06:01 interesting right it could look
1:06:03 like the world's a lot less [ __ ]
1:06:06 stressed out
1:06:09 and and these things are also going to
1:06:10 dramatically increase prosperity in the
1:06:13 you know the cynical side of you
1:06:14 probably says oh if these tools
1:06:17 dramatically increase Prosperity then
1:06:19 the capitalisters is going to take all
1:06:21 that profit and you know [ __ ] The
1:06:23 Working Man
1:06:24 but I think it's going to be a profound
1:06:26 enough Gap
1:06:27 that you're gonna have to deal with the
1:06:29 people that are displaced and then so if
1:06:32 you redistribute some of the
1:06:34 the Delta of of that new Prosperity back
1:06:38 to the people
1:06:39 and these tools do all the main mundane
1:06:41 [ __ ]
1:06:43 then what do people do I don't know
1:06:44 people want to contribute they want to
1:06:46 connect they want to make a difference
1:06:47 so maybe they start connecting and
1:06:50 making a difference at a very different
1:06:52 level than they did before
1:06:54 so it could it could look like we're
1:06:57 starting to discover ReDiscover our
1:06:59 Humanity in 10 years
1:07:02 um
1:07:03 it could it could also be a lot darker I
1:07:06 don't know I like I don't I don't have a
1:07:08 good sense of 10 years out I feel like I
1:07:11 have a good sense of the next two to
1:07:12 three years
1:07:14 and that's not pretty like that just
1:07:16 feels like a lot of disruption a lot of
1:07:18 change it's exciting
1:07:20 like lots of excitement like oh that's
1:07:22 exciting if you're one of the ones who's
1:07:24 doing the cool [ __ ] but if you're one of
1:07:27 the ones that's been displaced that's
1:07:29 not exciting that's that's [ __ ] hard
1:07:33 laughs
1:07:35 and then five years I feel like we start
1:07:38 to see new kinds of jobs emerging at a
1:07:42 pretty good scale where
1:07:44 displaced people are stepping into new
1:07:46 kinds of
1:07:48 work
1:07:49 that I don't know what looks like what
1:07:51 what is the work in the future I don't
1:07:53 [ __ ] have a clue
1:07:55 we'll still work like we're people we
1:07:57 want to contribute we want to do [ __ ]
1:08:00 you know I don't I don't think that like
1:08:02 if these tools just start doing all the
1:08:03 mundane [ __ ] we all just stop
1:08:06 it's just not she's not in our nature
1:08:09 right we'll keep improving [ __ ]
1:08:13 in doing [ __ ] and contributing but we
1:08:15 just might contribute on a very
1:08:16 different level maybe there's a lot more
1:08:18 live music right maybe all the streaming
1:08:21 services and all the automated music
1:08:23 generation you know
1:08:26 we don't
1:08:27 need to listen to recorded music anymore
1:08:30 because everything's personalized maybe
1:08:32 that drives up demand for in-person
1:08:34 concerts
1:08:37 maybe that goes up I don't know I don't
1:08:39 know
1:08:40 it's a really good question
1:08:44 um
1:08:46 when the bod failed well Bard the board
1:08:49 of Avon stepped upon the circle in the
1:08:52 Square
1:08:55 attempted to make words and indeed what
1:08:58 emerged was word salad
1:09:02 it was not a tasty affair
1:09:07 you know you probably shouldn't call
1:09:09 your AI Bard if it can't write
1:09:17 just saying
1:09:24 uh uh
1:09:32 all right I lost my place in my
1:09:34 questions but my little questions
1:09:37 it's sheep mentality yes the Microsoft
1:09:40 Edge is clunky
1:09:46 um
1:09:48 wait the Microsoft Edge link was hokey
1:09:51 because it takes you directly to the
1:09:54 Discord AI what's is that for Mid
1:09:56 Journey
1:09:57 how do you style your hair
1:10:01 apparently Something About Mary uh kind
1:10:05 of Vibes going on there I I don't well I
1:10:08 normally just put some product in it and
1:10:10 like you know run a brush through it and
1:10:12 I'd leave for the day but by the time I
1:10:14 come to these lives all the glue has
1:10:16 broken or some of the glue has broken
1:10:18 and some of it hasn't so the uh the
1:10:22 stylins this time of night are pretty
1:10:24 cray-cray
1:10:29 oh
1:10:31 [Music]
1:10:33 those are cool
1:10:43 I got new new bands for my Apple watch
1:10:49 they're pretty cool
1:10:51 um
1:10:52 I went to get a new phone yesterday and
1:10:55 the the lovely young woman at the ATT
1:10:57 store was like do you have an AARP card
1:11:00 and I was like Madam how dare you how
1:11:05 dare you
1:11:06 and then but I ended I actually did have
1:11:09 one and I got 20 off and so she goes you
1:11:11 want a free Apple watch and pay 10 bucks
1:11:13 less a month than you were paying them
1:11:15 like yeah
1:11:16 so it got me an Apple Watch I've never
1:11:18 had one
1:11:20 um
1:11:22 by the way I played around with church
1:11:24 EBT a month ago and asked it
1:11:27 it got drug brand names it didn't go so
1:11:30 well yeah it's so the safety the safety
1:11:33 guard rails on these things like they
1:11:36 don't want you to hurt yourself or hurt
1:11:37 others and or or they don't want to get
1:11:40 sued so these things are a little
1:11:42 hobbled Bing is really hobble bing.com
1:11:44 is is just this side of useless because
1:11:47 it's so
1:11:48 done did you get mid-journey AI to work
1:11:51 on Microsoft Edge
1:11:52 now
1:11:54 I didn't
1:11:56 it's
1:11:58 um
1:11:59 I know you showed me that stuff
1:12:00 yesterday I never got it to work
1:12:04 um
1:12:08 sorry I'm trying to figure out where my
1:12:12 where my questions are where was I and
1:12:15 the questions people where was I
1:12:20 oh okay these I did
1:12:24 beta
1:12:26 I can't make 50 cents Skee-Ball tickets
1:12:29 Tick Tock is totally ski ball tickets
1:12:32 tick tock you know let's keep all things
1:12:35 those are the best this coupon tickets
1:12:37 were like you do really good and it
1:12:38 shoots out like you got thousands of
1:12:40 pounds of those tickets and you you take
1:12:42 them up to the desk and you're like I'm
1:12:44 gonna get a new car and you plop it on
1:12:46 there and they're like you can get half
1:12:48 a pack of Juicy Fruit gum
1:12:50 you're like hey I just spent 700 on this
1:12:57 oh
1:12:58 that is totally Tick-Tock right now can
1:13:01 you shout out my name vanisha yes I can
1:13:04 Venetia
1:13:06 got it out
1:13:10 I think she's just trying to figure out
1:13:11 wait do you know High Fly AI I don't
1:13:16 but I'll go look at it because maybe
1:13:18 there's something there
1:13:23 High Fly AI
1:13:27 Arrow
1:13:29 is a leading wetly specialist what what
1:13:32 is a wet lease specialist
1:13:34 Fleet Management I don't think that's
1:13:36 what you're talking about
1:13:38 I fly
1:13:41 I don't know you're gonna need to give
1:13:42 me a
1:13:44 um
1:13:45 URL because I'm not seeing High Fly AI
1:13:49 anywhere
1:13:51 so no I do not wow I just found the chat
1:13:53 gpt4 plugins wow yeah there's like 800
1:13:56 of them
1:13:58 um Firefly here's the thing about the
1:14:00 GPT uh for plugins most of them are crap
1:14:05 like web Pilots good there's one called
1:14:08 prompt perfect that's good
1:14:10 there's one called stories which isn't
1:14:12 bad that writes it creates a little kids
1:14:15 book
1:14:17 um but there's 800 to play with the
1:14:19 thing that that is really impressive is
1:14:22 code interpreter so code interpreter
1:14:24 when you go to that there's a little
1:14:26 plus button next to the prompt you can
1:14:28 upload a file so you can upload like a
1:14:30 CSV file
1:14:31 and then you can just start asking chat
1:14:34 to analyze it and tell you what trends
1:14:37 are in it and tell you what's
1:14:38 interesting about the data and it does
1:14:40 all this crazy stuff so play around with
1:14:42 code interpreter too
1:14:47 um 100 100 heck I'm learning stuff in
1:14:50 these Tick Tock lives you see I'm going
1:14:52 to tick tock University
1:14:54 Professor Shannon how can I help you
1:14:57 [Laughter]
1:15:02 micro great hair now I think you're just
1:15:05 trolling me Mike
1:15:05 [Laughter]
1:15:11 oh Lord by the way I played around with
1:15:12 chat EBT a month ago oh that was the
1:15:14 drug name ones Napster never think of of
1:15:18 a song to get yeah exactly resume
1:15:21 writing and rewriting is on point
1:15:23 haven't refreshed your resume in 20
1:15:25 years chat GPT result the bear giving me
1:15:29 the dangly likes the dangly lights make
1:15:31 me happy that's my favorite of those
1:15:33 little bits the NPC bits I like the
1:15:38 lights I like the lights I'm wearing a
1:15:40 hat I'm wearing a hat I like the light
1:15:44 we can turn this channel into that if
1:15:46 you want
1:15:47 should I give twerking lessons
1:15:50 well all right fine we'll keep doing AI
1:15:53 it's up to you
1:15:55 how do you style your hair I admit it
1:15:57 Kyle Microsoft Edge link was hokey yep
1:16:00 when do you think GPT will be able to
1:16:03 reference past conversations I think
1:16:05 we'll see that relatively soon
1:16:08 um
1:16:09 so
1:16:15 let's see
1:16:17 I'm trying to think when we'll see it
1:16:19 first
1:16:24 it sounds like
1:16:27 Microsoft
1:16:29 co-pilots not going to be
1:16:31 I don't think they're going to release
1:16:32 that until 2024
1:16:36 um
1:16:37 and if that's the case then I think
1:16:39 Bard's behind them so we won't see it
1:16:41 embarked I think when when co-pilot and
1:16:45 Bard both launch in in the office suites
1:16:49 um you're going to be able to reference
1:16:52 your own data and then I think
1:16:55 a feature on top of that will will
1:16:59 likely be that you'll have memory of
1:17:01 past conversations
1:17:09 see I I was more optimistic about chat
1:17:12 GPT innovating with that kind of stuff
1:17:14 until
1:17:16 they pulled access to the internet out
1:17:18 and haven't replaced it like it's been
1:17:20 out for like five weeks now
1:17:23 um I think I think open AI right now is
1:17:25 dealing with a lot of they're trying to
1:17:27 Stave off a bunch of lawsuits from from
1:17:30 Publishers that are that are that are
1:17:32 that are like cut us a deal where you
1:17:34 pay us a big chunk of your revenue or
1:17:37 you know face the wrath of our lawyers I
1:17:40 I I just get the sense that that open AI
1:17:43 is a little distracted right now
1:17:46 um
1:17:50 anthropic
1:17:52 so claude.ai
1:17:55 could be the one that comes up with
1:17:57 something like that first like the fact
1:17:59 that they just kind of dropped a hundred
1:18:01 thousand token context limit
1:18:04 um I would pay attention to cloud.ai
1:18:06 they might come up with that but but
1:18:09 certainly within within six months
1:18:13 and then you know depending on how geeky
1:18:16 geeky you are Mark you can do that right
1:18:18 now with Lang Ching so so Lang chain you
1:18:22 can embed
1:18:24 um
1:18:24 your existing conversations into
1:18:28 a vector database and create your own
1:18:30 memory system
1:18:32 um it requires you to be a bit geeky but
1:18:34 if you just go hang out on uh just go
1:18:36 hang out on Reddit and where people are
1:18:40 talking about Lang chain
1:18:42 and you could probably find a project
1:18:44 where 90 of the work's done it's
1:18:45 probably just sitting in a in a Google
1:18:49 um
1:18:50 collab notebook where you just hit the
1:18:53 play button so
1:18:56 um yeah yeah
1:19:00 how do I do this oh I see oh that's cool
1:19:03 I'm looking at my new little watch
1:19:05 straps
1:19:07 they're pretty sexy they're pretty slick
1:19:13 all right moving on this isn't about my
1:19:16 watch straps it is when you have ADD
1:19:19 foreign
1:19:21 squirrel
1:19:24 uh
1:19:26 uh what's the best AI for duplicating my
1:19:28 voice for videos
1:19:30 um
1:19:31 11 Labs right now is pretty good
1:19:34 although
1:19:37 I just got a new phone I just got an
1:19:39 iPhone 14. so in iOS 17's beta
1:19:45 inaccessibility there's a thing called
1:19:52 personal voice
1:19:54 so built into the iOS 17 operating
1:19:57 system for their modern phones I assume
1:20:00 it's
1:20:01 probably 12 up maybe 13 up
1:20:05 um
1:20:08 they give you like 150 phrases to repeat
1:20:11 and then they synthesize your voice and
1:20:14 you can start responding to people with
1:20:16 your voice but just with text so you
1:20:18 type in text but it'll convert it to
1:20:20 your voice and send it I don't know how
1:20:22 good it is I haven't seen a review of it
1:20:24 but that's one of my next projects is to
1:20:26 use Apple's voice synthesis to see how
1:20:29 it is I have a feeling Apple's going to
1:20:30 be really good at this [ __ ]
1:20:33 um but 11 Labs right now is probably the
1:20:36 the one that is
1:20:39 kind of the the best recognized for
1:20:43 doing the synthesized voice well
1:20:47 um
1:20:48 I don't know if you can do your own
1:20:50 voice on D Dash ID
1:20:54 I don't think you can you can Import in
1:20:56 11 Labs voice that you did
1:21:04 [Music]
1:21:15 [Music]
1:21:15 um
1:21:20 create video type your script
1:21:28 now
1:21:29 I don't think I don't think you can do
1:21:31 your own voice at did I don't know about
1:21:34 Synthesia probably 11 Labs
1:21:37 um
1:21:39 I would think that there's going to be a
1:21:40 lot of them
1:21:41 I would I would give the voice stuff for
1:21:44 me feels like it's a year out before
1:21:46 it's really good right now it's not bad
1:21:48 like I can go I can go show you one
1:21:53 let's go to 11 labs
1:22:01 so it's elevenlabs.io
1:22:05 and
1:22:08 you basically upload a recording of
1:22:10 yourself talking and I tried to have
1:22:13 chat GPT write me a script
1:22:15 of different
1:22:17 you know phonetic sentences that would
1:22:20 be good for training one of these I
1:22:22 don't think it did a great job at
1:22:23 writing that script so there's probably
1:22:26 some things that I did wrong that if I'd
1:22:28 done a little more research I could have
1:22:30 done better
1:22:32 but let's see this is you know in a
1:22:34 world where
1:22:36 brains were bites and wisdom was Wi-Fi
1:22:39 I'm the all-knowing simpleton who's
1:22:41 downloaded the entire universe into my
1:22:43 Noggin but uh where do I plug in the
1:22:46 toaster so this was a little
1:22:49 phrase that I had chat CPT right the
1:22:51 other night so I've got two different
1:22:52 versions of me I've got the the normal
1:22:57 version and the bright version
1:23:00 so the normal version was me just kind
1:23:02 of talking like I'm talking now so let's
1:23:04 generate it with that one
1:23:07 foreign
1:23:09 in a world where brains were bites and
1:23:11 wisdom was Wi-Fi I'm the all-knowing
1:23:13 simpleton who's downloaded the entire
1:23:16 universe into my Noggin
1:23:18 but uh where do I plug in the toaster
1:23:20 right and then I have the bright version
1:23:23 so that one kind of sounds like me as I
1:23:25 talk now a little more monotone and you
1:23:27 know doesn't have inflection and all
1:23:30 that's the stuff that needs to get
1:23:31 better
1:23:32 but let's do this is the the bright
1:23:34 version so I talked a little more like a
1:23:36 Pixar
1:23:39 I'm the all-knowing simpleton who's
1:23:41 downloaded the entire universe into my
1:23:44 Noggin but uh where do I plug in the
1:23:46 toaster you know so not bad little
1:23:49 monotone doesn't quite have the
1:23:51 emotional inflection or or you know
1:23:54 Cadence shifting where where you know
1:23:57 just just the pacing is is still weird
1:24:00 but I would think that within a year
1:24:03 this stuff's going to be pretty
1:24:05 indistinguishable you know whether it's
1:24:07 apple or it's 11 Labs or one of these
1:24:09 other companies but these guys are
1:24:11 probably the ones to play with right now
1:24:13 that that's the the other thing to
1:24:16 um
1:24:19 the the other thing about
1:24:22 anything you do right now is don't get
1:24:24 married to
1:24:27 any given technology this stuff's
1:24:29 changing so fast like think of
1:24:31 everything for the next year is just a
1:24:33 big [ __ ] experiment you can't [ __ ] it
1:24:35 up right it's just an experiment just
1:24:38 learn just get curious and explore and
1:24:42 play with this [ __ ] and play with
1:24:43 mid-journey and play with that thing and
1:24:45 find a free thing and then use another
1:24:47 thing and realize oh that thing's really
1:24:49 shitty but oh that thing's really good
1:24:50 but no one's talking about it why is no
1:24:52 one talking about it I don't know maybe
1:24:54 you discovered a gem like like just just
1:24:59 throw all the pasta at all the walls is
1:25:02 my is my strong strong advice and again
1:25:04 that goes back to that thing about you
1:25:05 know is there a good course for this
1:25:07 stuff don't worry about courses like if
1:25:09 you if you really need to take a course
1:25:11 to get up to speed do like the Google
1:25:14 the 10 courses that Google just put out
1:25:17 go go do those because they'll be decent
1:25:20 but the real learning is is in learning
1:25:23 how to apply this stuff and learning how
1:25:25 to think about it and learning how to
1:25:26 experiment with it because it's just
1:25:28 there's just uh it's just all Gonna
1:25:30 Change it's all going to change
1:25:33 like a year from now
1:25:36 it's it's gonna actually be fascinating
1:25:38 to see I went back so I've started
1:25:40 putting these the recordings of these
1:25:42 lives on YouTube which that's the the
1:25:44 URL for the YouTube channel is AI
1:25:46 learning lab Dash TT on YouTube Dash
1:25:49 Tick Tock
1:25:51 um
1:25:53 and I went back and I looked at some of
1:25:55 the the early recordings of these one I
1:25:58 was a bit tubbier and I know that
1:26:02 that's that's tough year well you're
1:26:05 tubby now that was tough here
1:26:08 um
1:26:09 so I was Tellier but but like this list
1:26:11 only had
1:26:13 chat.open AI on it they only it only had
1:26:15 one and then I added Bing and then I
1:26:18 added
1:26:20 Poe
1:26:21 so then for like two months I had three
1:26:23 of them and now I've got seven it's
1:26:25 gonna be fascinating a year from now
1:26:27 what's on this list
1:26:30 wild wild wild times
1:26:33 I have Oodles of original artworks can I
1:26:36 upload an image and have an AI tool
1:26:38 write a description yes you can Firefly
1:26:42 so a couple of ways you can do that
1:26:44 let's go play with that because that's
1:26:45 actually a pretty cool
1:26:47 um pretty cool project so you can do it
1:26:49 in mid Journey
1:26:52 so let me go
1:26:55 and someone's doing some cool artwork
1:26:58 this is in the AI salon so if you go to
1:27:01 the salon.ai
1:27:03 um we've got a meetup group and a
1:27:04 Discord group and then within our
1:27:06 Discord you can do
1:27:08 um
1:27:09 uh mid-journey images
1:27:12 so if if in Discord or yeah if in
1:27:16 Discord where you've got mid Journey you
1:27:18 type in slash describe
1:27:22 you can drag and drop an image and it
1:27:24 will write you a mid-journey prompt for
1:27:28 that image
1:27:30 so let me go grab
1:27:46 all right that's a cool one all right so
1:27:48 there's some sort of groovy groovy-esque
1:27:51 Cloud bursting through this
1:27:54 there's sunbursting through the clouds
1:27:56 kind of thing so I'm going to drag that
1:27:57 into mid Journey
1:28:00 and
1:28:01 have it describe it
1:28:10 and then the other two places you can do
1:28:11 this so so mid-journey describing images
1:28:15 is
1:28:17 um is going to give you prompts right so
1:28:21 it gives you four a huge cloud
1:28:25 lights and Dark Water Fantasy desktop
1:28:28 wallpaper a dark scene with dark clouds
1:28:31 fantasy painting surrounded by darkness
1:28:34 in the clouds so so it gave me four I
1:28:37 can I can now say imagine all so I can
1:28:41 so I can either pick any one of those
1:28:43 four prompts and have it recreate it or
1:28:45 I can say imagine all
1:28:48 and so it's now going to give me four
1:28:50 images that are new original mid Journey
1:28:53 images based on the description of that
1:28:55 image so this is an interesting way
1:28:57 where you can start really playing
1:28:58 around with
1:29:00 um not having to write prompts but using
1:29:03 images as sources
1:29:06 so let's we'll see what this does and if
1:29:08 you haven't played in mid-journey the
1:29:10 notice how it starts as like these sort
1:29:13 of pixely kind of blobs blurry kind of
1:29:15 blobs and then they get a little more
1:29:17 specific and then they'll get a little
1:29:18 more specific
1:29:21 thank you
1:29:22 so this is one of them
1:29:24 there's another one there's another one
1:29:26 so these are the four different prompts
1:29:32 and you get four Images each for each of
1:29:34 those prompts
1:29:45 oh these are cool
1:29:49 look at that this this is a cool one
1:29:51 with the lightning and the little city
1:29:53 down here
1:29:55 and then this one's super layered
1:30:01 that's cool these are actually all
1:30:03 pretty cool everything's over kind of
1:30:05 like fiery lava
1:30:10 um
1:30:12 there's another set of descriptions
1:30:14 actually the the the images it created
1:30:17 are very much in the neighborhood of
1:30:19 what it described
1:30:22 huh those are cool okay so anyway so
1:30:25 that's that so that's one way you can do
1:30:27 it is you can do it with the describe
1:30:29 function in
1:30:31 mid-journey the other way you can do it
1:30:33 is we can head on over to Bing
1:30:37 and when you get to Bing you go to chat
1:30:42 and then within chat you see this
1:30:44 there's a little microphone there but
1:30:46 you have this little picture icon so I
1:30:48 can now go grab an image and upload it
1:30:51 from this device
1:30:54 is that in on desktop is this it this is
1:30:57 it right okay so now we're going to take
1:30:59 the same image and I'm going to say
1:31:03 um
1:31:05 please write
1:31:08 five inspiring
1:31:11 add headlines
1:31:17 add
1:31:18 lines
1:31:21 that will be placed on this image
1:31:27 for
1:31:29 um
1:31:33 fancy sunglasses
1:31:38 um
1:31:40 reference
1:31:46 what
1:31:48 in the image in the headlines
1:31:52 that's a really shitty prompt but let's
1:31:54 see how good Bing does with it
1:32:02 so it should
1:32:04 Bing's a little weird sometimes Bing's
1:32:06 brilliant and then sometimes it's just
1:32:08 like nope
1:32:10 see the world like new light with our
1:32:12 sunglasses shine bright like the sun
1:32:15 with our sunglasses protect your eyes
1:32:17 and style with our sunglasses get ready
1:32:20 for summer experience a new world in a
1:32:22 new way
1:32:24 um and then it also gave us it just went
1:32:26 ahead and generated some sunglasses
1:32:28 images
1:32:29 um let's let's try it again let's let's
1:32:31 upload that same image
1:32:35 and then we'll say
1:32:37 um
1:32:38 please
1:32:41 write a poem about the imagery in this
1:32:47 image
1:32:51 and again it's a really crappy prompt
1:32:53 right I could write something like in
1:32:55 the style of a certain poet or whatever
1:32:56 but
1:33:01 the light shines bright through the
1:33:03 clouds a river runs through Rocky
1:33:05 shrouds the sky is dark the mood is
1:33:08 Moody the mood is Moody
1:33:10 the landscape Baron yet so Broody the
1:33:14 light reflects off a river a sight that
1:33:17 makes my heart quiver the clouds are
1:33:20 dark yet they shine a sight that's truly
1:33:23 Divine the landscape is Barren yet so
1:33:26 alive
1:33:28 sight that makes me feel so alive all
1:33:30 right so it is you know
1:33:33 Eighth Grade B minus poetry but it
1:33:36 understood what was in the image and
1:33:38 then for some reason this is also
1:33:40 generating new images
1:33:44 that's weird but whatever that's fine
1:33:48 um they're trying to figure some [ __ ]
1:33:49 out over there but let's go try Bard now
1:33:52 let me I'll go copy the uh
1:33:57 I'll go copy the exact prompt that I did
1:33:59 the the brilliant prompt write a poem
1:34:03 about the imagery in this image
1:34:07 let's go to Bard
1:34:12 barred because Bard can also read images
1:34:15 Bard's also multimodal
1:34:19 all right so we're here so I'm going to
1:34:21 hit the plus icon and see you can upload
1:34:24 a JPEG PNG or webp so we're gonna go
1:34:27 grab that PNG
1:34:29 pop it in there we're going to paste in
1:34:32 our shitty prompt and see how Bard does
1:34:34 with it
1:34:42 okay
1:34:44 a cloudy Sky a light that shines a hope
1:34:48 that flickers through the rains a world
1:34:50 of gray a world of Gloom a world that
1:34:53 seems to have no room that's actually
1:34:55 pretty good for Joy or hope or laughter
1:34:58 bright for anything that's worth the
1:35:00 fight but then the light a ray of Hope
1:35:02 glimmer in the darkest coat goes on and
1:35:06 on so it it clearly saw it was in there
1:35:09 it wrote the poem you know again these
1:35:13 these poems are
1:35:15 um bad not so much because Bard's bad
1:35:17 because I gave it a really shitty a
1:35:20 shitty prompt um
1:35:22 let's see
1:35:24 um let's make that a sonnet
1:35:30 by Shakespeare
1:35:34 let's see if it let's see if it gets any
1:35:37 more word fruitier
1:35:40 when the skies are gray and filled with
1:35:42 rain a ray of Hope still doth remain
1:35:45 though all around is filled with Gloom a
1:35:47 light shine still shines a beacon true
1:35:50 in darkest days when all seems lost
1:35:53 there's a hope that is yet tossed upon
1:35:56 the waves of Time and Tide a hope that
1:35:59 will never reside it's not horrible but
1:36:02 it clearly sees it
1:36:04 um yeah and then you can do whatever you
1:36:06 want to do then so so there's there's
1:36:08 two main options
1:36:11 um mid-journey you can kind of Remix by
1:36:15 having it describe the image and then
1:36:17 with these things you can do kind of
1:36:19 chat gbt stuff with an image
1:36:21 so hope that helps Firefly
1:36:26 firefly firefly
1:36:31 all right
1:36:33 meant to say art piece what is Blue
1:36:35 Willow so blue Willows another
1:36:37 um it's like mid-journey it's just not
1:36:40 as good but it's free
1:36:42 so if you go to Blue Willow
1:36:45 AI
1:36:47 it's bluewillow.ai
1:36:50 it's a free image generator so you go to
1:36:53 the website and then it'll point you to
1:36:55 the Discord server so so you say try it
1:36:58 now for free and then it'll point you to
1:37:00 Discord and then that's confusing as
1:37:03 [ __ ] because if you're not expecting to
1:37:04 go to Discord
1:37:06 how would you know what the hell that is
1:37:08 but that's that's what's going on there
1:37:12 Tobias plus one watching
1:37:17 oh
1:37:18 Laura Petty your Tobias is plus one
1:37:21 watch it on my own welcome Laura I
1:37:23 understand you're coming to Denver
1:37:24 that's very cool yup found it web pilot
1:37:28 plugins bring out all the features in
1:37:30 gpt4 yeah so Firefly when when gpt4
1:37:36 um the the first feature that they
1:37:39 launched for it was
1:37:41 um browse with Bing it was connected to
1:37:43 the internet and then browse with Bing
1:37:45 was pulling content from behind Fortune
1:37:48 Magazine's paywall and apparently they
1:37:51 got pissed off about it and they haven't
1:37:53 brought it back since that was like more
1:37:54 than a month ago
1:37:55 so
1:37:57 um so web pilot definitely makes gpt4
1:38:00 better because
1:38:03 oh it's ridiculous but whatever it's
1:38:06 just we we got to give these
1:38:08 companies Grace right now because they
1:38:11 are transforming
1:38:13 all of what we do for all of the future
1:38:18 good
1:38:19 it's just a little shitty right now
1:38:23 okay
1:38:25 I need the pompous one for hosting a
1:38:27 meeting next month at work to keep
1:38:29 people awake yeah exactly I mean that's
1:38:31 the thing that's the other thing about
1:38:33 these tools is so so it's actually a
1:38:35 really good point
1:38:37 um Mez is
1:38:38 um
1:38:43 a lot of companies right now are very
1:38:45 skittery about this technology one
1:38:47 because it's changing things too because
1:38:49 they don't necessarily trust the privacy
1:38:51 and security which they shouldn't
1:38:54 um but also a lot of employees are
1:38:57 probably using these tools to kind of
1:39:00 enhance their personal productivity and
1:39:02 might not want work to know about it
1:39:04 there's there's a lot of sort of
1:39:08 there's a lot going on
1:39:11 so anything that you can do to demystify
1:39:13 this stuff and kind of bring it out in
1:39:15 the open and educate people and if you
1:39:18 use something entertaining like the the
1:39:19 pompous ass actor prompt
1:39:22 um
1:39:23 just to show people that you know you
1:39:26 can do creative interesting things with
1:39:28 this is is probably not a bad thing to
1:39:30 do just to get people to [ __ ] lighten
1:39:33 up let's lighten up people let's start
1:39:35 exploring this stuff because whether
1:39:38 you're an individual or whether you're a
1:39:40 business
1:39:41 my personal point of view is
1:39:45 disrupt yourself disrupt yourself before
1:39:48 you get disrupted right because this
1:39:50 shit's coming whether you want it or not
1:39:52 and if you sit on the sidelines it's
1:39:54 just gonna [ __ ] smack you upside the
1:39:56 head none of us want that for you
1:40:00 that's so cool beans it's so big like
1:40:03 it's yelling at you oh when I did the uh
1:40:05 the text with the headlines do it in
1:40:09 markdown no use for us
1:40:12 I'm a I'm a wait I'm in project
1:40:14 management all the processors
1:40:18 wait
1:40:20 all the pro all the processes are there
1:40:22 just no clear way to implement AI so so
1:40:27 Nick critic what I would say is if
1:40:29 you're still here
1:40:31 um
1:40:33 go check out
1:40:36 um
1:40:38 futurepedia.io
1:40:40 because there are probably some project
1:40:41 management Tools in there I would also
1:40:44 just pay attention to the project
1:40:45 management tools you're using now
1:40:47 whether it's Trello or jira or things
1:40:50 like that
1:40:51 um
1:40:56 the um
1:40:58 so so
1:41:00 almost all of those tools will likely be
1:41:02 rolling these kind of things into the
1:41:05 tools
1:41:06 so so just pay attention to that but the
1:41:08 other thing you can do with project
1:41:09 management is you can take little sub
1:41:12 processes and you can figure out how you
1:41:15 might be able to do them you know faster
1:41:17 or more efficiently so either analyzing
1:41:20 documents or
1:41:22 um doing call summaries or you know next
1:41:25 next actions based on a meeting
1:41:27 transcription
1:41:29 you can also use things like zapier to
1:41:32 do automations so if you've got the same
1:41:35 kind of thing you're doing over and over
1:41:36 again you might find specific things you
1:41:39 can start to explore manually with these
1:41:42 things and then if you're doing them
1:41:43 over and over again
1:41:45 um
1:41:46 you can uh
1:41:48 you can automate them and and kind of
1:41:50 make your own tools but I would also go
1:41:52 check out futurepedia there's probably
1:41:54 some project management things out there
1:41:56 that might be useful but I I like I
1:42:00 understand what you're saying I would
1:42:01 still explore how to do what you do
1:42:05 see if you can do it better faster or
1:42:07 cheaper with these tools and if you
1:42:09 can't then you can't but my strong
1:42:13 suspicion is that you probably can
1:42:19 all right ask chat GPT and see what it
1:42:21 says about 10 years from now that's a
1:42:23 good idea
1:42:25 okay oh this will be interesting so
1:42:27 let's let's say
1:42:31 um
1:42:37 act as a futurist who understands the
1:42:40 implications
1:42:42 of machine learning
1:42:45 Automation and advanced
1:42:49 AI Technologies period
1:43:02 um describe
1:43:07 foreign
1:43:21 your best understanding
1:43:25 of how things today are being impacted
1:43:29 by Machine learning and artificial
1:43:31 intelligence period
1:43:38 as of my last update in 2021 and I'll
1:43:41 I'll talk about that in a second so so
1:43:43 basically chat EBT not connected to the
1:43:46 internet
1:43:47 has data up to September 2021
1:43:51 so automated jobs financial markets
1:43:53 business decision making
1:43:56 in health care diagnosis and treatment
1:43:59 drug discovery
1:44:03 content recommendation creation okay so
1:44:06 so I'm just so one of the things I'm
1:44:08 trying to do here is is get it to get it
1:44:11 to level set itself to understand kind
1:44:13 of what's going on now
1:44:16 because I feel like that's going to help
1:44:17 with the
1:44:19 the future stuff
1:44:27 projecting forward is AI and ml continue
1:44:30 to evolve we can expect
1:44:32 deeper integration into our daily lives
1:44:36 making many processes more efficient
1:44:39 there will be a further blurring of the
1:44:41 lines between human and machine
1:44:42 collaboration
1:44:45 shift in the job market will likely
1:44:47 continue
1:44:49 ethical considerations will play a
1:44:51 larger role
1:44:53 and then symbolic relationship between
1:44:55 humans and AI will likely be a focal
1:44:57 area
1:44:59 um okay so let's say
1:45:01 um
1:45:05 describe
1:45:09 describe
1:45:15 what you feel the major phases will be
1:45:18 over the next 10 years
1:45:23 period
1:45:28 while I know you can't predict the
1:45:30 future
1:45:36 foreign
1:45:38 share the implications
1:45:42 assuming these Technologies get
1:45:45 exponentially more powerful
1:45:53 on a five-year curve
1:45:58 let's see if it comes up with
1:45:59 interesting phases
1:46:05 2023 to 2028 integration and
1:46:08 specialization deep integration
1:46:10 specialized AI systems
1:46:17 Healthcare Revolution personalized
1:46:19 medicine will become a norm
1:46:21 so this is five years out this is phase
1:46:23 one
1:46:25 robust natural interaction AI systems
1:46:28 will understand and generate human
1:46:30 language with exceptional proficiency
1:46:32 making interactions feel natural and
1:46:35 intuitive that feels right to me
1:46:37 entertainment Evolution virtual and
1:46:39 augmented reality experiences will be
1:46:41 powered by AI creating deeply immersive
1:46:43 interactive worlds that feels more like
1:46:45 10 years than five to me but whatever
1:46:48 the line between technology and daily
1:46:50 life will blur even further okay so so
1:46:53 phase one is integration and
1:46:55 specialization
1:46:57 phase two 2028 to 2033 so this is the
1:47:01 10-year time frame five to ten years
1:47:03 autonomy and enhancement
1:47:06 true autonomous vehicles cognitive
1:47:08 enhancement AI might be used to augment
1:47:11 human cognition either through external
1:47:13 devices or potentially even biologically
1:47:16 Integrated Systems hello neural link Mr
1:47:19 musk I'll take my implant now
1:47:22 AI in governance we can see AI playing a
1:47:26 role in City Planning resource
1:47:28 allocation maybe even policy simulations
1:47:30 to protect predict societal impacts
1:47:34 advice Environmental Management human AI
1:47:37 collaboration
1:47:39 throughout the phases continuous
1:47:41 learning and education yes I agree with
1:47:43 that
1:47:45 as as capabilities of the AI grow the
1:47:48 importance of humans learning and
1:47:50 adapting will be crucial new educational
1:47:53 paradigms focusing on flexibility
1:47:56 ethical and Regulatory Evolution
1:47:58 societal transformation just as the
1:48:01 Industrial Revolution transformed
1:48:02 Societies in unforeseen ways the AI
1:48:05 Revolution will lead to shifts in in
1:48:07 social structures values and Norms
1:48:10 um
1:48:11 let me try one other thing
1:48:15 if you were to imagine the new kinds of
1:48:19 jobs that will be created
1:48:22 that will replace
1:48:28 the ones for workers that were displaced
1:48:32 by AI comma
1:48:37 what are those jobs
1:48:42 it's kind of a half-assed finish to the
1:48:44 request but that's all right the
1:48:46 displacement of jobs
1:48:48 is a significant concern however history
1:48:51 has shown that technological revolutions
1:48:53 while displacing certain jobs often lead
1:48:56 to the creation of entirely new ones I
1:48:57 think that's right so Ai and Technology
1:49:00 focused roles an AI trainer so someone
1:49:03 who trains AI systems a maintenance
1:49:06 technician someone someone who keeps the
1:49:08 robots turned on when we can't live
1:49:10 without them a Quantum Computing
1:49:12 specialist that's interesting as Quantum
1:49:15 Computing growth grows specialist in
1:49:18 this field will be in high demand human
1:49:20 AI collaboration human AI teaming
1:49:22 manager individuals who understand both
1:49:25 human psychology and AI capabilities
1:49:28 that's a fascinating role AI interaction
1:49:31 designers
1:49:33 that's cool ethics and oversight AI
1:49:36 ethicist algorithm bias auditor that
1:49:40 sounds that's some real 1984 future
1:49:43 dystopian [ __ ] right there
1:49:45 Education and Training AI literacy
1:49:47 reskilling trainer Healthcare and
1:49:49 well-being
1:49:51 Telehealth technicians mental health
1:49:53 professionals
1:49:54 the mental health thing is going to be
1:49:56 big because so many people are gonna
1:49:57 there's gonna be so much change there's
1:49:58 gonna be a lot of anxiety
1:50:00 environment and sustainability
1:50:04 virtual reality designers digital
1:50:06 artists and animators security and
1:50:08 safety space exploration space tour
1:50:11 guide
1:50:12 as space tourism begins to take off
1:50:16 that's interesting
1:50:18 local and craft craft-based roles so
1:50:21 artisanal Craftsmen a Renaissance in
1:50:24 handmade local Goods might arise as we
1:50:27 go to fully automated everything content
1:50:32 everywhere that's what I was saying
1:50:33 about you know we might go back to more
1:50:35 in-person music artisanal Craftsmen
1:50:38 that's interesting
1:50:41 yeah so there you go oh it's still
1:50:43 writing we've choked it out it's going
1:50:46 really slow
1:50:47 it's like I'm tired
1:50:49 it's essential to understand that while
1:50:51 some jobs will undoubtedly be displaced
1:50:53 human creativity empathy and the ability
1:50:55 to think critically will remain
1:50:57 Irreplaceable for the foreseeable future
1:51:01 that's the big one for me and you know
1:51:02 people ask about education all the time
1:51:04 like what's the most important thing to
1:51:06 teach
1:51:06 for me it's critical thinking because I
1:51:08 think that's the one thing that remains
1:51:10 in the purview of humans where it's our
1:51:13 power center but it's also we get to
1:51:15 choose
1:51:17 how to deploy these machines and we get
1:51:19 to choose you know how we're going to
1:51:21 use them to augment what we do that
1:51:23 requires critical thinking so that that
1:51:25 feels like that gets way more important
1:51:28 over the next decade
1:51:32 all right what time is it
1:51:35 all right I have another 15 minutes or
1:51:38 so before I gotta go my voice is getting
1:51:40 all crunchy
1:51:43 oh crunchy that was a fun one thanks for
1:51:45 that crypto drill that was a good idea
1:51:47 all right more time for liquid lunches
1:51:51 absolutely
1:51:53 I got out of programming for years
1:51:55 barely did anything but all these things
1:51:58 got me going at it apart I know it's
1:52:01 like it's like I'm doing the same thing
1:52:03 I'm like diving back in I'm like oh crap
1:52:05 I gotta make websites again I gotta
1:52:07 figure this stuff out it's way easier to
1:52:09 go from an idea to working code yeah
1:52:11 these things are pretty remarkable with
1:52:13 coding and and not just working code Tom
1:52:16 like that's a that's a really good point
1:52:18 it that that it's it's easier to get to
1:52:20 work in code but it's really easy to get
1:52:22 to sort of a working level of all sorts
1:52:25 of things marketing copy strategies
1:52:29 um you know images
1:52:31 just all the outputs you can you can get
1:52:34 to some
1:52:35 like 80 percent interesting very very
1:52:38 quickly and and that changes everything
1:52:42 um I'll just be a dog walker there's
1:52:44 gonna be a need for dog walkers have you
1:52:46 heard about putting temperature of
1:52:48 language 1.0 to 2.0
1:52:51 from basic to elaborate yeah and I think
1:52:53 it's actually Chris isn't it from zero
1:52:56 to two
1:52:57 because I know in in the playground it's
1:52:59 zero to two
1:53:01 um so what what Chris is talking about
1:53:03 there caruno
1:53:05 is talking about
1:53:07 um I haven't actually played with it but
1:53:08 let's try it
1:53:10 um
1:53:11 so I'm going to say
1:53:13 um
1:53:15 let's see
1:53:17 explain
1:53:20 quantum mechanics
1:53:27 um
1:53:29 as if I were
1:53:32 a 10-year old
1:53:36 um
1:53:37 I forget how you do it is it set
1:53:41 set
1:53:43 temperature
1:53:48 to zero
1:53:52 so this should be I'll put 0.0 this
1:53:56 should be
1:53:57 um very dry and
1:54:00 not creative
1:54:05 all right kiddo imagine you're playing
1:54:07 with a Lego set when you build things
1:54:09 with Legos you have individual pieces
1:54:11 that make up the world
1:54:12 and on your Lego world a block can be
1:54:14 only in one place at a time right but
1:54:18 what if I told you in the world of tiny
1:54:20 particles things get a little dot dot
1:54:22 dot wacky sometimes these particles are
1:54:25 a little more blah blah blah blah here's
1:54:27 the craziest part this doesn't seem like
1:54:29 it's temperature zero to me but let's
1:54:30 try it
1:54:32 I'll try to set it to one
1:54:34 and that might not be the way you
1:54:36 actually set it here
1:54:39 set temperature to 1.0
1:54:48 all right buddy let's imagine we're a
1:54:50 super cool video game world
1:54:57 oh
1:55:03 sorry
1:55:04 it was pretty straightforward now it's
1:55:07 doing a video game world and it's
1:55:08 creating
1:55:10 like a little list of items magic dice
1:55:12 teleporting coins Twin Magic shy
1:55:15 particles
1:55:17 it also added some some little doodads
1:55:20 here so let me try now
1:55:23 set it to 2.0
1:55:26 so what temperature is for for those
1:55:29 that are wondering is it's just
1:55:31 basically
1:55:33 um
1:55:35 for lack of a better term less creative
1:55:37 to more creative
1:55:39 okay kiddo buckle up because quantum
1:55:41 mechanics is like the wackiest zanious
1:55:44 cartoon cartoon show you've ever seen
1:55:46 But for tiny tiny things in our world
1:55:48 jelly bean jumble dance party twin
1:55:51 telepathy so yeah it's got it's gotten
1:55:53 more creative here so so yeah it looks
1:55:56 like adding the temperature is doing
1:55:58 something
1:55:59 I just put set temperature to 0.0 to 1.0
1:56:03 and 2.0 yeah that's interesting
1:56:06 I haven't really played with that but
1:56:07 but I do when I'm when I'm actually
1:56:09 building prompts in the GPT playground
1:56:13 um where you can actually change the
1:56:14 temperature slider I play with it a lot
1:56:17 depending on the the type of prompt that
1:56:19 I'm using so it's I I haven't thought
1:56:21 about using it in chat GPT that's a
1:56:23 really good idea actually
1:56:27 um
1:56:29 all right maybe I should get an annuity
1:56:32 and wait on universal basic income well
1:56:34 actually crypto Joe
1:56:36 um I'm going to New York I leave Sunday
1:56:40 for New York
1:56:41 um and I'm going to get
1:56:43 um
1:56:44 my world ID so I think I talked about
1:56:48 this when you're on last night Joe
1:56:50 but um
1:56:52 Sam Altman's new company is called World
1:56:55 coin and his whole idea is to have that
1:56:58 be a universal basic income platform
1:57:01 um to redistribute some of the
1:57:03 um wealth generated by AI
1:57:06 um back to the people so I'm gonna go
1:57:08 get my Iris scanned
1:57:11 against the better judgment of everyone
1:57:13 I know
1:57:17 and if I weren't in America I'd actually
1:57:19 get some free crypto out of it but but I
1:57:22 am in America so I don't get [ __ ] thanks
1:57:25 Gary Gensler you're a douche
1:57:28 um I think I'll be in amalgamation of
1:57:31 wait oh I think it will be an
1:57:33 amalgamation of state corporatism melded
1:57:36 with AI controls it might be Nick critic
1:57:39 but
1:57:41 meta giving away their
1:57:45 large language model with commercial
1:57:47 rights I think throws a
1:57:50 a wrench in the works of the of the
1:57:52 government
1:57:54 take over in control of all this [ __ ]
1:57:57 unless they do something
1:58:01 crazy like you know
1:58:03 take over the internet
1:58:06 access to it like China does Thelonious
1:58:09 I I just give backpacks and vests to
1:58:11 kids for back to school at the shelter
1:58:13 remain hopeful oh that's sweet that's
1:58:15 awesome
1:58:16 people will yearn for the Nostalgia I
1:58:18 agree
1:58:21 um wait Thelonious just said something
1:58:23 about world coin will worldcoin issued a
1:58:26 privacy statement that they don't retain
1:58:28 the scan they just turn it over to you
1:58:30 yeah the the the
1:58:32 um and and that makes sense to me like
1:58:35 there's no reason for them to
1:58:38 to retain
1:58:40 um
1:58:42 that biometric data I mean quite frankly
1:58:46 um
1:58:46 ancestry.com with with all their DNA
1:58:49 scans like they've got way more than
1:58:51 just getting a retina scan so the only
1:58:53 thing World coins
1:58:56 looking to do here is they want to
1:58:58 create a a network of verified human
1:59:01 beings
1:59:02 and then they can incent them by saying
1:59:05 hey if you're part of this network
1:59:06 you're going to be part of the universal
1:59:08 basic income program so you can you can
1:59:10 make money here you can spend that money
1:59:12 and then what they get is they get a
1:59:15 network of verified humans that's going
1:59:17 to be incredibly valuable
1:59:20 um once we realize that you can't trust
1:59:23 that anyone online is human anymore
1:59:25 because there's going to be so many deep
1:59:27 fakes and all that sort of [ __ ] so
1:59:29 having a trusted verifiable you know
1:59:32 tokenized identity
1:59:35 I think becomes really really important
1:59:37 so
1:59:38 I'll go get one so you guys don't have
1:59:40 to but but if you're right now you can
1:59:43 get them in New York San Francisco and
1:59:45 Miami
1:59:46 um and you basically just look into this
1:59:48 thing they call the orb and they scan
1:59:50 your retina and give you an ID
1:59:53 so it's just cuckoo I tell you it's
1:59:56 cuckoo
1:59:59 Kyle I don't like to chat Bard Claude
2:00:02 don't do anything now wow analyze
2:00:05 controversial news dots oh won't analyze
2:00:08 controversial news talks oh that's
2:00:10 interesting yeah Claude and Bard the the
2:00:12 safety guard rails on these these tools
2:00:15 right now are are a real pain in the ass
2:00:19 they really are I I lost my I lost where
2:00:22 I was in the question so I'm just going
2:00:24 to skip down toward the bottom here
2:00:25 because I don't have a ton of time left
2:00:27 all right what if I ask AI to print out
2:00:29 money will alarm Bells go off it
2:00:32 probably it probably just won't do it or
2:00:36 or if it does do it it'll it'll just
2:00:37 look like [ __ ] for now
2:00:40 um but again if if the if the safety
2:00:44 guard rails are frustrating you go start
2:00:46 exploring
2:00:47 um
2:00:52 go start exploring labs.perplexity
2:00:55 that's the the Llama models and and
2:00:58 just go go start playing with the open
2:01:00 source large language models because the
2:01:02 there you don't have the guard rails
2:01:05 my friend owns an AI Fleet Management
2:01:07 Company oh that's fascinating actually
2:01:09 like AI Fleet Management I think
2:01:11 anywhere where you've got incredibly
2:01:13 complicated Logistics the AI stuff's
2:01:16 just gonna be [ __ ] magic and it's
2:01:18 it's they're gonna make a lot of money
2:01:20 you can make money with chat gbt by by
2:01:24 automated complicated Logistics
2:01:27 yeah I wonder what AI will do to online
2:01:30 dating
2:01:31 um
2:01:32 hopefully it makes it better and and
2:01:35 helps people communicate I mean I you
2:01:38 might have a new kind of catfishing
2:01:40 where the AI
2:01:43 sort of coaches you to be more
2:01:45 interesting than you might actually be
2:01:48 um and you say all the right things and
2:01:51 then when you get together in person
2:01:52 you're like
2:01:53 the other person's like
2:01:56 is this the same person I was talking
2:01:58 with online it's probably not
2:02:04 funny
2:02:07 um but the other thing with the online
2:02:08 dating I think that I think that um
2:02:10 virtual companionship is going to be a
2:02:12 thing I think it's going to actually be
2:02:14 an important thing
2:02:15 um and I don't think it's necessarily
2:02:16 all bad I know it's a little creepy and
2:02:18 weird because we're not used to it but
2:02:20 there's a lot of lonely people and I I
2:02:22 think that as these as these tools get
2:02:24 more and more sophisticated and and the
2:02:28 the boundary between
2:02:31 you know talking to a person and talking
2:02:34 to a bot gets blurrier
2:02:37 um
2:02:38 like these tools as they are right now
2:02:40 behave with a lot of empathy and a lot
2:02:43 of nuance they're actually good at
2:02:44 having conversations Pi in particular is
2:02:48 is very much designed to be a
2:02:49 conversation bot so I think that's going
2:02:52 to be interesting
2:02:53 it was a stylized B for Bing chat GPT
2:03:00 then type take me to mid-journey oh
2:03:03 that's interesting okay
2:03:06 um
2:03:07 thought you about to rap I I will not
2:03:10 rap I will not twerk and I will not rap
2:03:12 I will spare you all both of those a
2:03:15 project manager can literally be
2:03:17 replaced by a very effective AI the
2:03:19 voice stuff is scary
2:03:21 I warmed my
2:03:23 or warned my 87 year old mother if I
2:03:26 call her to ask for money hang up and
2:03:28 call me yep
2:03:30 yeah Thelonious that's that's a really
2:03:32 good thing I think that um
2:03:34 I think it's not a bad thing for
2:03:36 families to it's like it's like uh I I
2:03:41 feel like families need to now have safe
2:03:43 words right you know the safe word is
2:03:45 Windex or whatever like if if any of us
2:03:48 call anyone else
2:03:49 and ask for something weird like you
2:03:53 need to send me two thousand dollars
2:03:55 right now I'm in trouble that
2:03:58 we all need to say what's the code word
2:04:00 and if we don't know it because that's
2:04:02 happening right now where people are
2:04:03 voice synthesizing you know daughters
2:04:06 and sending it to the mother and saying
2:04:07 I need money
2:04:09 um so yeah that's a that's a really good
2:04:11 idea Thelonious
2:04:12 um my wife says I can watch you
2:04:17 Nate thanks Nate is 11 Labs free no 11
2:04:21 Labs is not free and it's not super
2:04:23 cheap but
2:04:24 it's it synthesizes your voice if like
2:04:28 if you want to experiment with it the
2:04:30 other thing you can do if you've got an
2:04:32 iPhone is go get ios 17 beta and then
2:04:35 inaccessibility there's a thing called
2:04:37 personal voice you can just play with
2:04:39 that for free
2:04:42 um
2:04:43 I don't know Thelonious Nate filipowsky
2:04:46 you're lucky my girlfriend's jealous of
2:04:48 Kyle oh no I bring him to the dinner
2:04:52 table I'm so sorry that's bad for
2:04:55 everyone involved
2:04:58 oh my God uh Claude reminds me of the
2:05:02 book lion taming 101 by Claude
2:05:05 Bales balls that's funny
2:05:08 um Kyle I don't like chat Bard uh don't
2:05:11 do anything now one analyze yeah
2:05:13 uh controversial news
2:05:16 very handsome in your shirt today it's a
2:05:19 fancy shirt and I got I got you know
2:05:22 some sort of hair thing going on there
2:05:25 uh I just got it his his shirt is a
2:05:28 bunch of toilet papers it is my kids
2:05:31 love Teletubbies they love you I'm sure
2:05:34 now you guys are just trolling me do you
2:05:37 have a background in AI or just sharing
2:05:39 some of the things you've looked at
2:05:41 um I don't I I've got a background in
2:05:43 entrepreneurship and
2:05:46 kind of
2:05:48 pouncing on technology as soon as it
2:05:51 you know sort of breaks
2:05:54 breaks free of the the Geeks and
2:05:56 scientists and and is ready for prime
2:05:58 time so I I started one of the early
2:06:01 digital agencies in the mid 90s so so
2:06:03 when I stumbled upon this thing called
2:06:05 the World Wide Web
2:06:07 um
2:06:08 I got my head around it I'm like holy
2:06:10 [ __ ] this is going to change everything
2:06:11 and I started this company called
2:06:12 agency.com and ended up building a lot
2:06:15 of the first websites for the Fortune
2:06:16 500 so I'm I'm constantly fascinated by
2:06:21 technology and and for the most part
2:06:24 dramatically disappointed in it you know
2:06:27 like apple since Steve Jobs died is just
2:06:29 a steaming pile of [ __ ]
2:06:33 um
2:06:34 but then I look at you know this AI
2:06:37 stuff and I look at the Vision Pro that
2:06:39 Apple just put out and it's like okay
2:06:41 they're actually redefining the category
2:06:43 and this this generative AI stuff all of
2:06:46 a sudden starts to look like the early
2:06:48 days of the World Wide Web so I feel
2:06:50 like we're at the beginning of a new
2:06:51 Renaissance and this stuff is raw and
2:06:55 janky duct tape pieces of [ __ ]
2:06:59 but it's
2:07:01 but it's profoundly powerful and like
2:07:04 the the implications of this stuff are
2:07:06 insane so this stuff isn't going to look
2:07:08 magic to the masses for probably two or
2:07:12 three years
2:07:13 but if you if you're willing to put up
2:07:16 with a little bit of the limitations of
2:07:18 it right now it's pretty [ __ ] magic
2:07:20 right now and and the other thing is
2:07:22 chat GPT when it launched it got to 100
2:07:25 million users in six weeks
2:07:28 the the World Wide Web when when it was
2:07:31 launched it took us six years to get to
2:07:33 100 million users so the speed that this
2:07:36 stuff is coming is is absolutely [ __ ]
2:07:39 insane
2:07:40 um so yeah so that's my background is in
2:07:43 you know I I guess applied use of
2:07:45 Technology rather than developing the
2:07:47 technology I just
2:07:49 there's something about me like once I
2:07:51 get passionate about something I got to
2:07:52 figure out how it works so like I've
2:07:54 I've learned enough about how it works
2:07:56 what it does how it does it why it does
2:07:58 it why it hallucinates
2:08:01 um so that I can start to think about
2:08:04 how might we use it how might it be used
2:08:07 how how is it going to transform things
2:08:10 um and boy it's it's this is nuts like I
2:08:15 felt like with the early days of the web
2:08:16 that was pretty easy to see what it was
2:08:18 gonna do and what the future looked like
2:08:20 this stuff is much harder because it's
2:08:22 so much more sophisticated
2:08:26 um hello from Majorca Spain
2:08:28 [Laughter]
2:08:32 oh it looks like I might have hung
2:08:35 um it just said I lost connection
2:08:38 um double tap the screen make the hearts
2:08:39 go
2:08:40 if you can still hear me
2:08:42 let me see if I'm I'm actually hung I
2:08:45 might be hung
2:08:49 oh my please okay great
2:08:52 um Tobias oh my please quickly repeat
2:08:54 repeat what you just did with the mid I
2:08:56 wouldn't
2:08:57 um
2:08:58 go back and show something like this but
2:09:00 you're pretty cool so
2:09:03 I'll do it for you so basically when
2:09:05 you're in Discord thank you for the rose
2:09:08 I like a rose a rose by any other name
2:09:11 would smell it sweet
2:09:13 uh
2:09:18 um
2:09:19 well that was weird
2:09:22 can you still hear me
2:09:25 hit the hearts again my my Tick Tock app
2:09:27 just shut down okay I'm whatever okay
2:09:31 Tobias so so if you're in mid-journey
2:09:33 you type subscribe I had a paper cut out
2:09:36 image of a a cruise liner
2:09:40 on the Open Seas
2:09:42 and one of the descriptions yielded this
2:09:44 image and I thought oh my God that's
2:09:46 beautiful and then I extended it and
2:09:48 stuff like that so let's grab that one
2:09:51 and we'll drag it in here
2:09:58 wait did it
2:10:01 brought the image in hang on I think I
2:10:04 messed something up there
2:10:05 oh yeah I did
2:10:06 image too large yeah because it wasn't
2:10:08 an image you dumb dumb okay hang on
2:10:11 describe
2:10:19 wait where did it go
2:10:27 oh
2:10:28 [Music]
2:10:36 I lost my VW bus
2:10:41 foreign
2:10:48 I just lost connection again
2:10:52 uh please please hit the hearts
2:10:55 so that I'm not just talking to myself
2:10:58 if
2:10:59 if I'm not there am I hung
2:11:02 no I'm so good all right thank you
2:11:07 oh it made it a four-wheel drive tractor
2:11:10 look at that
2:11:11 what's this one gonna be
2:11:13 more four-wheel drive tractors
2:11:23 I mean again what's what's nice about
2:11:26 this approach where you you know have it
2:11:28 analyze an image and then give you a
2:11:30 prompt and generate an image out of that
2:11:32 it's a way that you can do creative
2:11:34 ideation like like you don't you don't
2:11:36 have to really think about it you can
2:11:37 you're just kind of remixing
2:11:40 you know from idea to idea to idea it's
2:11:43 like playing playing a game of telephone
2:11:44 with images
2:11:46 like those are pretty cool that one on
2:11:48 the bottom right is pretty
2:11:50 pretty bitching
2:11:54 so
2:11:57 [Music]
2:12:04 this is a very Hawaii 5-0 shirt isn't it
2:12:07 toilet paper you people are rude
2:12:10 all right
2:12:16 does Bard create images as well it does
2:12:19 not create images but Bing does
2:12:23 so if I go to Bing
2:12:28 Bing
2:12:32 and go
2:12:38 make me an image of a jet powered
2:12:45 tractor
2:12:49 then it will
2:12:52 so that's Bing Ding Ding Ding Ding Ding
2:13:03 there you go jet power tractors oh wait
2:13:17 that's actually pretty cool I like that
2:13:20 cab
2:13:26 go build one of these
2:13:28 oh that's one of the other things that
2:13:30 I'm actually really excited about so I
2:13:31 think I think we're gonna we're like
2:13:34 like right now all this stuff is just
2:13:36 images and things like that but if you
2:13:37 look at if you look at something like
2:13:39 that when when
2:13:43 you know builders start getting their
2:13:45 hands around
2:13:47 these tools they're going to generate
2:13:50 some crap in mid-journey or in bing
2:13:52 something like this they're gonna be
2:13:53 like you know what I'm gonna go build
2:13:55 that or someone's gonna do it with
2:13:57 Furniture Design or interior design or
2:13:59 architecture or product design I think
2:14:03 we're I think we're probably
2:14:05 in the next two to three years we're
2:14:08 going to start seeing just absolutely
2:14:10 wild design Trends because they're going
2:14:12 to be inspired by a lot of this stuff
2:14:15 that's actually really exciting to me
2:14:21 uh
2:14:23 when I'm going this I actually put in
2:14:27 oh on a bachelor degree level that's
2:14:30 interesting
2:14:33 Pablo the dog dancing I use it to write
2:14:36 emails it's good for writing emails a
2:14:39 lot of people that I know are writing
2:14:40 resignation letters I had someone sent
2:14:43 me an email and they're like how how do
2:14:46 I tell my current employer to go [ __ ]
2:14:48 themselves without burning the bridge
2:14:49 and I said just put that in there I said
2:14:53 describe what they did to you and put it
2:14:56 in there as nasty as you want it to be
2:14:58 and then just tell it to rewrite it as a
2:15:00 professional resignation letter and it
2:15:03 will do it and it'll do it really good
2:15:06 um I use it all the time and I've
2:15:07 replaced grammarly yep that's good
2:15:09 you're a poet laureate yeah I'm I'm
2:15:11 something
2:15:13 uh Lifesavers so mid-journey Bing chat
2:15:16 and Bard for image text descriptions yes
2:15:21 thank you you're welcome
2:15:23 um
2:15:24 this change is going to haunt everyone
2:15:27 who thinks this is changing the future
2:15:30 for good
2:15:32 well
2:15:34 it's just a start so there's more to
2:15:36 come
2:15:37 this change is going to haunt everyone
2:15:39 who thinks this is changing the future
2:15:41 for good well
2:15:43 like most things in life I don't think
2:15:45 this is going to be binary right I don't
2:15:48 think it's going to be all evil and I
2:15:49 don't think it's going to be all good
2:15:52 I think it's going to be profound
2:15:55 and I think there is a very real
2:15:58 possibility that this is dramatically
2:16:02 more positive than anyone is expecting
2:16:07 I think we're all expecting it to be
2:16:09 dramatically negative
2:16:11 I think that's actually the easier
2:16:13 position to take right we've got 50
2:16:15 years of Hollywood telling us the robots
2:16:17 are going to [ __ ] kill us
2:16:19 right so it's easy to go there we've
2:16:21 been trained to go there
2:16:22 what if
2:16:24 it got better
2:16:27 that's a possibility
2:16:29 now will it be 100 nope
2:16:32 are these things going to cause some
2:16:34 major [ __ ] Mayhem and destruction yep
2:16:38 but are they going to do some amazingly
2:16:40 good things probably
2:16:43 probably probably
2:16:52 I feel like it's pointless to speculate
2:16:57 on how good or evil these are right now
2:17:00 I feel like the only thing that's
2:17:02 relevant to do right now
2:17:04 is play
2:17:06 explore learn
2:17:09 what these things are good at learn what
2:17:11 they're bad at
2:17:12 figure out how to apply them for good
2:17:14 now
2:17:16 because if you apply them for good now
2:17:18 that may bend the trajectory
2:17:22 of the future of these slightly in a
2:17:24 positive direction
2:17:27 so
2:17:29 like I the the probably the thing that
2:17:32 frustrates me the most right now is
2:17:34 people
2:17:35 who are speculating about the future who
2:17:38 are not actually using the tools
2:17:41 that's just seems [ __ ] asinine to me
2:17:46 so I don't know if you're one of those
2:17:47 if you are then like I would say start
2:17:49 playing
2:17:50 like if you start using these tools
2:17:54 and then you decide here's how I think
2:17:56 it's going to destroy Humanity great
2:17:58 like yeah let's have that conversation
2:18:00 but if you're just sitting on a soapbox
2:18:02 going ah it's evil and they're not
2:18:04 perfect
2:18:07 not interesting
2:18:10 go play
2:18:11 go see if you can help someone get a job
2:18:14 or or Inspire them to do something
2:18:17 they've always wanted to do but didn't
2:18:18 quite have the skill for it and now they
2:18:20 do
2:18:23 isn't that energy better spent than
2:18:26 they're gonna kill us
2:18:28 I don't know
2:18:31 I might just be a a little ninny I might
2:18:34 just be overly optimistic I might just
2:18:36 be naive I might be very naive I just I
2:18:38 think everything's going to be
2:18:39 Shangri-La we're going to be singing and
2:18:42 dancing in the alleyways
2:18:45 why are you so happy because the AI
2:18:47 fixed everything
2:18:50 Maybe
2:18:52 you good enough you're smart enough and
2:18:54 doggone it people like you
2:18:58 hahaha
2:19:03 wow great prompts I think that was one
2:19:06 of my writing shitty prompts thank you
2:19:08 for trolling me it's nice I don't think
2:19:11 anyone in government is smart enough to
2:19:13 do that yeah I agree I think you know
2:19:15 the the government they don't understand
2:19:17 what Wi-Fi is how are they going to deal
2:19:19 with this [ __ ]
2:19:21 how did the meeting presentation go for
2:19:23 you I don't know which one it was I
2:19:26 don't I haven't I haven't gone to New
2:19:27 York I leave for New York next Sunday
2:19:29 that's that's the one where I've got
2:19:31 I'll find out if we got this job that
2:19:33 I'm looking to hopefully get three jobs
2:19:36 actually
2:19:38 that five-year future prompt was
2:19:40 powerful yeah that was pretty cool
2:19:41 wasn't it
2:19:43 DIY in many areas will expand from
2:19:46 accounting to tax filing to litigation
2:19:49 defense and education
2:19:51 yeah oh DIY in many areas yeah yeah
2:19:55 exactly I mean
2:19:58 if you think think about any profession
2:20:02 any profession
2:20:04 that is
2:20:08 that that leverages the complexity of
2:20:12 the world right so tax preparer
2:20:16 um accountant like all the ones that you
2:20:18 say here right litigation defense
2:20:22 um education a lot of the a lot of the
2:20:24 things where you've got someone who's
2:20:27 taken a lot of years to learn all the
2:20:29 noogly details of an overcomplicated
2:20:32 system and it's their job
2:20:35 to make sense of that complicated [ __ ]
2:20:38 for regular human beings this stuff can
2:20:41 now do that or will be able to right if
2:20:44 it's if it's not 100 there now it will
2:20:46 be able like that's that's where these
2:20:47 things are headed they can do the really
2:20:49 complicated [ __ ] like that
2:20:54 so think any job
2:20:57 where the purpose of the human being is
2:20:59 to
2:21:00 simplify the complicated
2:21:04 those positions likely go away
2:21:10 but you can't replace me I've been
2:21:12 studying tax for 20 years okay
2:21:16 this can read all of the tax code for
2:21:18 the last hundred years and know every
2:21:20 piece of it
2:21:23 here's a fascinating
2:21:25 thought thought experiment
2:21:29 so right now you got your Rich Fellers
2:21:32 right your billionaires and what do your
2:21:35 billionaires do your billionaires hire
2:21:37 these super smart people that know all
2:21:39 the tax code and hire all these minions
2:21:41 to go figure out all the possible
2:21:43 loopholes for how they can get their
2:21:46 money into foreign bank accounts and
2:21:48 this and that for the I wouldn't call it
2:21:50 tax evasion I would just call it
2:21:53 taking advantage of the tax code of the
2:21:55 United States of America well
2:21:58 what happens
2:22:00 when instead of
2:22:03 0.002 percent of the population being
2:22:06 able to take advantage of those
2:22:08 complicated loopholes
2:22:11 30 percent of us can take advantage of
2:22:13 that because we're now using these tools
2:22:15 to take advantage of that
2:22:17 what happens then
2:22:19 I think laws change I think laws will
2:22:22 change
2:22:25 when you know because that's going to
2:22:28 [ __ ] with some Revenue that's going to
2:22:30 [ __ ] with some taxes man shit's Gonna
2:22:32 Change shit's gonna go down
2:22:35 we can't have the people possibly
2:22:37 understanding what we've been doing for
2:22:39 a hundred years
2:22:41 that would be bad
2:22:45 those are my laws
2:22:47 where's my jet
2:22:52 what llm are you using tonight I think
2:22:54 mostly what I've been using tonight is
2:22:56 uh chat GPT
2:22:59 you can make money
2:23:02 sorry if this is a wait oh sorry if this
2:23:05 is moronic nothing's moronic I'm new I
2:23:07 didn't realize GPT remembers your
2:23:09 previous prompts for how long that's
2:23:11 actually a great question so what's your
2:23:13 name Pro uh
2:23:16 wait for forever
2:23:21 forever hooked I think forever forever
2:23:24 or hooked
2:23:27 you wrote that like a like a license
2:23:29 plate
2:23:31 um I think it's forever hooked I'm gonna
2:23:32 give you a 6.5 on the name I like the
2:23:34 name but it's too hard to read
2:23:36 I'm old
2:23:38 um
2:23:40 but it's a really good question so
2:23:43 chat gbt currently has what's called a
2:23:47 context window of 4 000 tokens you know
2:23:51 like Kyle but Kyle what's a token
2:23:54 doesn't [ __ ] matter it's what Geeks
2:23:56 call it's how they measure how many
2:23:59 elements they're processing so tokens
2:24:03 include punctuation in spaces
2:24:06 so 4 000 tokens equals about three
2:24:09 thousand words so think like
2:24:11 three quarters of the amount of tokens
2:24:14 is how many words you can have
2:24:16 so chat GPT right now has
2:24:19 a context window a memory in a single
2:24:23 chat session of about three thousand
2:24:25 words so what that means is when I start
2:24:27 typing and I say the the thing I was
2:24:29 doing about
2:24:31 um
2:24:32 about the future of AI
2:24:35 um you know do you understand this and
2:24:36 it said yes I understand this so as long
2:24:38 as that's all of that conversation is
2:24:41 within 3000 words it'll remember it and
2:24:43 and it'll be responding in context
2:24:46 a lot of times what happens is you keep
2:24:48 talking with it and like as the as the
2:24:51 the older stuff goes outside of that
2:24:53 context window it forgets it
2:24:55 so so what ends up happening is you're
2:24:57 having a really good conversation with
2:24:59 chat GPT and just all of a sudden it's
2:25:01 like I don't know what you're talking
2:25:02 about like it's just you're like wait a
2:25:04 minute which where was the Brilliance
2:25:07 from before it blasted out of the
2:25:09 context window so a couple of things
2:25:12 at some point it's going to go to 8 000
2:25:14 tokens and then ultimately it's going to
2:25:17 go to 32 000 tokens right so about 25
2:25:20 000 words so it's increasing they're
2:25:22 just rolling it out slowly because it's
2:25:23 expensive but
2:25:25 we have an ace up our sleeve
2:25:27 Claude dot AI has a hundred thousand
2:25:30 token context window
2:25:32 so Claude has a 75 000 word memory so
2:25:36 you can have a really long conversation
2:25:38 with Claude and it won't forget it it
2:25:40 won't like lose just lose its mind like
2:25:43 Chachi BT does
2:25:46 um
2:25:47 the other thing you can do with Claude
2:25:48 is you can upload up to five files so
2:25:50 you can upload five different PDFs of
2:25:53 whatever a screenplay you've written and
2:25:55 a I don't know a blog post and whatever
2:25:57 and just start interacting with it
2:26:00 um so that's that's actually a really
2:26:02 good question and then one of the other
2:26:05 questions someone asked earlier was you
2:26:07 know when is it going to have long-term
2:26:08 memory when is it going to be able to
2:26:09 remember what you talked about across
2:26:12 conversations that requires it to
2:26:15 actually embed all of your conversations
2:26:17 put them into a database some of these
2:26:19 tools at some point are gonna
2:26:22 release that functionality it's just not
2:26:24 there right now
2:26:26 AI will win a Grammy it AI will
2:26:29 absolutely win a Grammy well no
2:26:31 someone
2:26:33 who used AI will win a Grammy
2:26:41 you can lead a horse to water where is
2:26:44 the list I don't know which list you're
2:26:45 talking about critical thinking skills
2:26:47 will always be lacking that's
2:26:49 interesting I think they're increasingly
2:26:52 important I'd love a copy of that list
2:26:55 of jobs as seen by chat GPT
2:26:58 um I wonder where I can put that
2:27:06 where did I
2:27:09 AI impacts
2:27:13 [Music]
2:27:13 um
2:27:17 I'm trying to think where I could put it
2:27:20 because I can share prompts I could
2:27:21 share the prompt
2:27:24 uh
2:27:28 I'll think about that Silver Fox if you
2:27:30 think of actually silver effects if you
2:27:33 want DM me and I can I can I can send
2:27:36 you that list
2:27:37 um but I I should I should figure out a
2:27:39 way maybe I'll maybe I'll figure out a
2:27:42 way to put together a prompt Library if
2:27:44 if we do interesting stuff on these
2:27:45 things it might be an interesting
2:27:47 resource this is one of the most
2:27:49 fascinating conversations I've heard on
2:27:51 Tick Tock thank you oh thank you very
2:27:52 much Miz I appreciate that
2:27:55 um how will chat CPT cause and effect
2:27:58 the undisclosed Tech physics
2:28:01 understanding in the next 10 years
2:28:04 I will chat GPT
2:28:07 oh
2:28:09 that's an interesting one
2:28:11 hot on the heels of LK 99
2:28:15 and and we've got UFOs that can do 80
2:28:19 000 miles an hour and turn on a dime
2:28:21 without smashing the the people inside
2:28:25 um
2:28:27 let's I'm gonna do it in the same
2:28:29 conversation as that other future so
2:28:32 let's see
2:28:33 how
2:28:39 will llms
2:28:47 cause and effect I don't know if that's
2:28:50 quite the right way to say it but these
2:28:52 things are pretty smart it'll understand
2:28:53 it
2:28:56 um
2:28:58 the
2:29:01 undisclosed
2:29:06 physics understandings
2:29:12 and Technologies
2:29:16 over the next decade
2:29:20 let's see if it understands that I think
2:29:23 I understand what what you're asking
2:29:25 there it appears there may be a
2:29:27 misunderstanding
2:29:32 uh it didn't understand what an element
2:29:34 was
2:29:36 um
2:29:38 sorry
2:29:40 llm means
2:29:43 um
2:29:45 machine learning
2:29:49 Ai and tools like chat GPT
2:29:56 uh I see you're using llm to refer to
2:29:59 large language models like chat gbt blah
2:30:01 blah
2:30:02 accelerating research cause and effect
2:30:05 llms can synthesize vast amounts of
2:30:08 information
2:30:10 and provide insights
2:30:12 or summaries to researchers affect the
2:30:15 capability could speed up research yeah
2:30:16 that seems pretty straightforward data
2:30:19 analysis I think the data analysis thing
2:30:22 is huge like the the ability to to
2:30:25 instantly analyze massive amounts of
2:30:27 data
2:30:29 in fields like particle physics or
2:30:31 massive amounts of data are generated
2:30:33 llms might play a role in quickly
2:30:34 identifying patterns or anomalies
2:30:36 collaboration tools
2:30:38 by serving as intermediaries that
2:30:41 translate domain-specific jargon
2:30:44 they could be used okay so this is all
2:30:47 oh wait this is all generic speculative
2:30:49 theories and simulations
2:30:52 regarding undisclosed physics
2:30:55 while open research and peer review are
2:30:57 the Cornerstone
2:30:59 it's conceivable that certain
2:31:01 advancements especially those with
2:31:03 potential defense applications might
2:31:06 remain classified or undisclosed
2:31:09 uh let's say
2:31:11 um
2:31:14 [Music]
2:31:17 what about identifying
2:31:20 and developing technologies that may
2:31:23 threaten the current power structure
2:31:26 period
2:31:27 how might AI
2:31:34 change the status quo
2:31:38 now we're getting into some conspiracy
2:31:40 theory crap oh yeah
2:31:45 the interplay between technological
2:31:47 advancements and the status quo
2:31:48 especially regarding power structures
2:31:50 has been a recurring theme throughout
2:31:52 history hasn't it though
2:31:54 from the printing press in the rentals
2:31:57 science which democratized information
2:31:58 to the rise of the internet in the
2:32:00 modern area technological shifts blah
2:32:02 blah blah
2:32:03 democratizing information
2:32:07 the democratization could challenge
2:32:09 Gatekeepers of information allowing a
2:32:12 more informed citizenry
2:32:14 well they're they're sort of handling
2:32:16 that by getting us all to fight each
2:32:17 other so so that they seem to be
2:32:20 handling right now
2:32:22 Grassroots movements could Empower
2:32:24 smaller groups to have a louder voice
2:32:26 that that I think is possible it's
2:32:28 actually fascinating look at what's
2:32:29 happening with the LK 99 stuff right now
2:32:31 where smaller groups are trying to
2:32:33 figure that [ __ ] out too
2:32:36 changing economic models bias and
2:32:38 control enhanced surveillance Ai and
2:32:41 governance
2:32:43 decentralized AI development
2:32:46 I don't know
2:32:48 some shit's gonna go down I'll tell you
2:32:50 that
2:32:51 some [ __ ] down it gonna happen it gonna
2:32:55 happen it won't happen
2:32:57 they're going to turn their robots loose
2:32:59 and gonna happen gonna happen it's gonna
2:33:02 happen
2:33:03 I don't know if you know that but it
2:33:04 gonna happen it's gonna happen
2:33:08 all right
2:33:09 retina skin to join Iris yes it might be
2:33:12 I think it's Iris uh for the world coin
2:33:15 World coin issued a privacy statement
2:33:18 they're using oh wait wait that was old
2:33:22 what
2:33:25 why
2:33:28 tick tock
2:33:30 I don't understand what's going on
2:33:32 Silver Fox is out of here by silver fox
2:33:36 Emilio's wife showed up hey Emilio's
2:33:39 what's happening what's shaking
2:33:46 Kyle if I call money just know it is me
2:33:49 that's funny that's a good one all right
2:33:52 I'm gonna do a few more questions and
2:33:54 then I gotta go what time is it oh it's
2:33:55 late it's late I I thought I was gonna
2:33:58 stay till 10 it's like 10 30.
2:34:02 um Troy Decker is not ready for the iris
2:34:05 scan ah Troy come on where's your sense
2:34:08 of adventure
2:34:09 they're using your iOS scan for AI and
2:34:12 use my DNA for Human Genome Project 2.0
2:34:14 yeah exactly
2:34:19 Facebook already have all the data like
2:34:22 does it really [ __ ] matter if they
2:34:23 have my Iris like maybe at some point in
2:34:27 the future where you know in order to
2:34:29 get out of the the work camps you gotta
2:34:34 have a
2:34:35 an unscanned iris or some [ __ ] like that
2:34:38 to let the hackers get break you out I
2:34:41 don't know
2:34:43 we're not there yet
2:34:46 all right
2:34:50 [Music]
2:34:53 I got mine in San Francisco last week
2:34:55 painless have you checked out deep dream
2:34:58 gent generator
2:35:00 for image manipulation no what's deep
2:35:02 dream generator is it a dream thing I do
2:35:06 know some people that are using
2:35:08 mid-journey to visualize dreams that
2:35:11 they've had they basically write their
2:35:13 dreams out when they wake up and then
2:35:15 put them in mid-journey
2:35:17 which I think that's a fascinating use
2:35:19 case
2:35:24 what's it called
2:35:26 deep dream generator deep
2:35:29 dream
2:35:31 generator
2:35:58 I guess I've been here before
2:36:00 generate
2:36:04 uh this looks like I think this is just
2:36:07 a web front end sitting on top of stable
2:36:09 diffusion
2:36:14 on Ancestry since the beginning 850
2:36:17 ancestors till 1400s in Germany oh
2:36:20 that's pretty cool Matthias
2:36:22 that's that's pretty cool 953 Colorado
2:36:24 time I'm missing class but I'll watch it
2:36:27 on the tubes later yeah you can I can
2:36:29 see the catfishing with AI generated
2:36:31 photos too yep exactly
2:36:35 um online dating with a AI you are very
2:36:38 funny thank you
2:36:39 [Laughter]
2:36:41 all right what AI platform can I use to
2:36:44 crawl my website for SEO errors
2:36:47 um any of any of the ones on screen
2:36:48 right now
2:36:51 um you can just copy your code paste it
2:36:52 in there and if it's too big use Claude
2:36:57 um but you could also you know what you
2:36:59 could also do is you could use web pilot
2:37:01 the web pilot plugin
2:37:03 and just point it to specific web pages
2:37:05 and tell it to do SEO analysis for you
2:37:15 [Music]
2:37:17 if only have to build me a boyfriend
2:37:19 like build a bear
2:37:23 have a family safe word I know safe for
2:37:25 safe words have escaped the dungeon and
2:37:28 now they now they're good for families
2:37:29 too
2:37:31 uh smart idea to have a code word code
2:37:34 word super smart yep I think that's
2:37:36 going to be necessary
2:37:39 uh what do you think about the Senate
2:37:41 talking to the guy that was connected to
2:37:43 the government and Aliens I don't know I
2:37:46 mean
2:37:47 I thought the fact that they they held
2:37:49 the the hearing was interesting but
2:37:52 neither of the guys that were talking
2:37:54 had any [ __ ] evidence it was all
2:37:57 hearsay I know someone that did
2:37:59 something well that's what they've like
2:38:02 we've heard that for [ __ ] decades
2:38:04 like it was it was a more impressive
2:38:06 venue than what we've heard before but
2:38:08 it was the same [ __ ]
2:38:11 um I lost the keys to my car your
2:38:14 connection is spotting I know no I can't
2:38:16 hear you internet is lagging bad I know
2:38:20 all right listen I love VW buses oh I'm
2:38:24 I'm very far behind I don't even know
2:38:25 where I am in the comments I'll look at
2:38:27 some bottom ones here deep dream
2:38:29 generator
2:38:33 um
2:38:34 oh I have a dream journal I need to see
2:38:37 what it does are they using GPT or an
2:38:40 image AI so Rebel
2:38:42 um if you've got a dream journal
2:38:44 go to you could do it in bing just go to
2:38:46 Bing and say make an image of and then
2:38:50 paste your dream in there or if you've
2:38:52 got a mid-journey account do it in
2:38:54 mid-journey because mid journey is just
2:38:56 way better from an image generation
2:38:59 point of view I would just say go start
2:39:01 popping your dreams in mid journey and
2:39:03 see what comes out I bet I bet it'll be
2:39:05 insanely powerful all right listen I
2:39:08 gotta go it's late here I got work
2:39:10 tomorrow you got work tomorrow you
2:39:13 shouldn't be watching me
2:39:14 but I really do appreciate y'all
2:39:21 wait I just I just uh swiped my my app
2:39:25 away
2:39:27 um peace out everybody I really do
2:39:29 appreciate you hanging out with me it's
2:39:30 been fun have a good one peace