AI Learning Lab

aiLL 5/7/23 - I Turn 58 Years Old and Chat About The Wild World of AI

9cQh_55WPAc
Video2023-08-062:51:2615 views

Description

It's my birthday! 🎉 In this May 7, 2023 live session, I discuss how generative AI like ChatGPT is transforming industries from advertising to education. Learn my predictions for Apple, opportunities in business, and risks we must mitigate. It's a wild time in AI: ChatGPT hit 100M users in 6 weeks, new generative techniques are emerging rapidly, big tech races to release products amid rising concerns, and open source threatens dominance. Dive in and get curious! Chapters: 0:00​ Intro - The Hair, The Face, The Birthday! 4:38​ How AI Will Change The Advertising Industry 10:51​ My Background & How I Got Into AI 19:11​ AI Assistants Doing Everything For You 34:12​ How To Teach AI Ethically In Schools 48:33​ Why I'm Now A Microsoft Fanboy 1:15:46​ Fun With Avatars and Virtual Agents 1:29:22​ A Vision For AI As A Global Public Resource #AI #ML #GenerativeAI #ChatGPT #Birthday

Chapters

Transcript

0:09 please
0:19 hey hey it's Kyle Shannon with the AI
0:22 learning lab
0:23 there is something special going on with
0:25 the hair today I look like I've been
0:28 hanging out with a Tesla coil
0:31 so if you have joined my live welcome uh
0:35 I kind of like chatting about AI stuff
0:38 uh I can't seem to keep away from it
0:40 it's Sunday night
0:42 what is it nine o'clock here in Denver
0:44 and uh if you have any questions about
0:46 AI jump in ask any questions I'll do my
0:49 best to answer them if I can't answer
0:51 them I'll either make stuff up like chat
0:53 GPT or I'll tell you I don't know
0:57 and uh I'm happy to show things I'm
1:00 happy to talk about things
1:03 um it was a big week let me pop up some
1:07 everyday AI this is my newsletter
1:12 and what were some of the big stories uh
1:15 open ai's code interpreter I think was
1:17 the biggest one that's why we put it
1:19 first
1:21 um bing.com is now open to everyone you
1:25 don't need to be on the wait list
1:26 anymore so if you want to use gpt4 and
1:29 if you want gpt4 that's connected to the
1:31 internet go to bing.com download the
1:33 edge browser and uh yeah you're you're
1:36 Off to the Races
1:40 um
1:40 there's been a tremendous amount of
1:44 progress in the open source Community
1:48 um really remarkable stuff there's a new
1:50 open source model that can do 65 000
1:52 tokens
1:54 so what that means
1:56 to human beings is you can paste up to
1:59 48 000 words into a prompt
2:03 um so so one one of the uh one of the
2:06 guys put in as the prompt the entire
2:10 text of The Great Gatsby and then it
2:12 then then just said write an epilogue to
2:14 The Great Gatsby and it did
2:17 um so yeah so if you've got questions
2:19 about AI
2:21 um put them in the chat uh looks like
2:22 we've got a bunch of people here you're
2:24 probably not going to hang around I'm
2:25 just an old guy sitting around talking
2:26 about AI crap on a Sunday night but if
2:28 you have questions about AI let me know
2:31 um so big big moves for the open AI
2:34 Community there was also if you didn't
2:35 see it there was a leak from a Google
2:37 employee there was an internal Google
2:39 memo uh that basically talked about
2:43 um we have no moat I think I have a link
2:45 to it here
2:48 um hang on let me see if I can get the
2:49 focus to go better yeah there we go
2:52 um Google has no mo basically what
2:54 they're saying is that the
2:57 um
2:59 the open source Community is doing for a
3:02 hundred dollars what it took Google 10
3:05 10 million dollars to do and what it
3:07 took open AI five million dollars to do
3:10 um they're not the open source models
3:12 aren't quite there but they're basically
3:14 taking really small large language
3:16 models and fine-tuning them in such a
3:18 way that they're getting essentially
3:19 similar results as Google's barred which
3:23 is arguably as good as chat GPT
3:26 um but uh but yeah so
3:29 um so that's another thing going on
3:31 that's just wild and what else do we
3:35 have here there's cool video search tool
3:38 open AI uh Sam Altman said this week
3:41 that he thinks they need to raise they
3:44 being open AI need to raise 100 billion
3:48 dollars to get to AGI which sounds
3:52 absolutely insane but they've already
3:53 raised 10 billion dollars from Microsoft
3:56 so they just need another 90. so if you
3:59 have a spare 90 billion dollars give Sam
4:01 Altman a call sounds like he could use
4:03 it
4:04 um if I want to use chatgpt4 without
4:07 being forced to use Bing are there
4:10 Alternatives I'm pretty sure Microsoft
4:12 is not forcing anyone to use Bing
4:16 um you can use the edge browser which
4:18 has a little
4:20 um
4:21 being chat button and it gives you a
4:24 sidebar that that's cheap GPT for but it
4:26 sounds like you think that Microsoft is
4:29 the evil empire which I do too but quite
4:31 frankly they've been acting more like a
4:33 startup in the past six months than uh
4:36 than anyone else
4:39 um so
4:41 um you know I feel like they're
4:43 innovating on a level uh beyond what
4:46 Apple's doing like apple hasn't
4:47 innovated since Steve Job died Steve
4:49 Jobs died in my opinion
4:52 um but yes if you want to use
4:55 um
4:56 gpt4 without being forced to use Bing
4:59 there's a couple of uh places you can go
5:02 to poe.com poe.com I think they limited
5:06 if you're not doing their paid thing
5:08 they limit it to maybe 10 searches a day
5:13 um or you know 10 10 queries a day
5:16 um if you pay open AI 20 bucks a month
5:20 for
5:22 GPT plus
5:24 then you get access to to gpt4 and I
5:28 think there's probably some others but
5:29 those are the ones that I know about
5:31 so um feel free to
5:35 um
5:36 yeah to go like you know pay 20 bucks a
5:38 month and just use it on on open AI site
5:42 that's what I do like I don't use Bing
5:43 not because I think they're the evil
5:46 empire I'm actually kind of impressed
5:47 with them lately I I I'm really enjoying
5:49 the Glee that their CEO is taking in
5:51 [ __ ] with Google it's really
5:55 hilarious
5:56 um it's free with pin yeah exactly so if
5:58 you want it for free go use Bing
6:01 um do you think AI can streamline pen
6:03 testing and when will it replace
6:05 traditional pen testers in the future
6:08 um yeah I think I think especially when
6:11 you when you start to look at things
6:13 like Auto GPT and baby AGI
6:16 um sorry I can't get my my little
6:20 my chair is knocking into my little
6:22 tripod this is my man it's hard live
6:25 casting is hard people
6:28 um what's the question uh oh pen testing
6:32 um
6:32 yeah I think when you look at things
6:34 like Auto GPT that that
6:38 um are using large language models to
6:40 prompt itself and take actions that's
6:42 kind of the perfect thing to do pen
6:44 testing right you know do pen testing
6:46 you know on this particular
6:49 um server and and just have at it
6:52 um I have a funny story about pen
6:54 testing so we were
6:56 we were doing a security audit I think
6:58 it was with Pfizer it's one of the big
6:59 Pharma companies I forget who it was but
7:02 um
7:03 we had this pen tester guy and he asked
7:06 for
7:07 credentials for our server
7:11 and and he was in India so the time was
7:14 was all messed up and and so this was
7:16 like it just kept happening in 24-hour
7:17 chunks where we would give him the
7:20 password and then 24 hours later we
7:23 would discover that he couldn't get in
7:25 that the password didn't work and this
7:27 happened like three times in a row
7:29 and then
7:30 we're finally like well let's just get
7:32 on the phone with this guy like what the
7:33 hell is going on here and what happened
7:36 was he he would
7:39 um
7:40 he would log into the system and then he
7:43 would run a script that
7:46 um automatically changed the password or
7:49 something like that and then the
7:51 password wouldn't work anymore
7:54 so so we told him
7:57 um you know don't run that script before
7:59 you log in so you know you can't make
8:04 this stuff up so the guy that was
8:05 running the pen testing was incompetent
8:07 this was where a major Pharma company
8:09 and we had to figure out why he couldn't
8:11 log into his own
8:12 to into our system it was absolutely
8:14 ridiculous so yeah uh from my experience
8:18 I think AI is going to do better at pen
8:20 testing I think it'll be very good for
8:21 it actually and if if you're if you're
8:23 in that world I mean I would say dive in
8:26 start looking at Auto GPT and those all
8:29 of those autonomous agents there's
8:31 probably a dozen of them out there now
8:33 um
8:34 all right so so what I do here is I just
8:36 sort of scroll through the
8:39 um
8:39 the comments kind of from the top down
8:41 so if you're adding comments at the
8:44 bottom I won't be able to see them for a
8:45 while so and if if you think I'm not
8:47 answering your question it's just really
8:49 because I haven't quite scrolled down to
8:50 it yet
8:52 all right this is uh the yeah I learned
8:55 you know
8:57 this is the AI learning lab I'm not
8:59 drunk I promise
9:01 um I'm Kyle Shannon if you want to
9:03 subscribe to my lives you can I don't
9:06 know why you would but Tick Tock told me
9:08 I can turn that feature on and I did so
9:10 now you can subscribe and you get
9:12 special little banners you can do and I
9:15 don't know what else you can do you know
9:17 I'm not an influencer
9:20 especially with that hair
9:22 um but best resource to learn about more
9:25 about AI um
9:27 well find people that you trust on on
9:30 YouTube and Tick Tock actually Tick Tock
9:32 seems to be kicking the [ __ ] out of uh
9:35 out of LinkedIn in terms of business
9:37 content for me lately like I'm I'm
9:39 learning more from for me it's kind of
9:42 tick tock Twitter like following people
9:44 I trust on Twitter and YouTube or the
9:47 big three
9:47 um where I learn about this stuff
9:49 anyone who's teaching a course in AI or
9:52 they're like I'm an expert in Ai and
9:54 I'll teach you everything you need to
9:55 know run run
9:57 like you're on fire from them don't
10:01 nobody's an expert right now this stuff
10:03 is changing so fast it's crazy
10:05 um I do uh this newsletter everyday AI
10:08 with
10:10 um Greg mushin who's an xxpedia product
10:13 guy so we put out a weekly email that
10:17 that you know sort of covers all the big
10:18 stuff the small stuff you know
10:20 interesting tweets eye candy things like
10:22 that
10:24 um so so that's another good way to do
10:25 it there's other good
10:27 um there's other good newsletters out
10:29 there um Rachel Brown's a good tick
10:31 talker who's who's really really tight
10:34 with keeping you up to speed on stuff
10:36 she's really good so I would encourage
10:38 you to follow her you can follow me
10:40 um I try to I try to keep everyone up to
10:43 up to date but this stuff's moving fast
10:45 and don't feel like you're behind
10:46 because it's this is all very very new
10:50 um I use Linux so Edge is not an option
10:53 yeah so um if you were the one that was
10:55 asking about being forced to use
10:56 Microsoft you can go to bing.com but I
10:58 would say just if you don't want to
11:00 support Microsoft then just go to open
11:02 AI
11:03 I use both Googles and gpt4 yep I do too
11:07 oh I don't know I don't sorry I don't
11:09 use Google's I don't use bard
11:11 um I haven't found it to be all that
11:13 useful
11:14 um and I don't I don't want to use lots
11:16 and lots of tools but it but it doesn't
11:18 matter if it works for you that's great
11:20 um
11:21 okay
11:23 um I'm 73 what do you think is going to
11:25 happen to society in the next 25 years
11:27 without AI hell I don't know what's
11:29 going to happen in the next year
11:32 um well it's it's funny I so I was on a
11:34 panel
11:35 um a few weeks ago
11:38 and on the panel I said you know when I
11:40 experience what
11:43 I recognize that the World Wide Web is
11:45 going to be a really big deal in 1994
11:48 like I discovered this thing I put an
11:50 online magazine up and within three
11:52 three weeks of publishing it there was a
11:55 full page article written about this
11:59 magazine I put up
12:01 um in Paris
12:02 um so so I kind of experienced
12:06 um
12:08 how big a deal the internet was very
12:10 early I had a big Epiphany about it and
12:12 I think that AI this is like the next
12:14 big thing this is I feel like this is
12:16 the second time in my life I've gotten
12:18 to sit at the precipice of one of these
12:19 things and I feel really lucky about
12:20 that
12:21 the difference between then and now is
12:24 with with the web I could kind of see
12:27 okay
12:28 um geography collapses and time
12:31 collapses right you've got instant
12:33 access to everything and it doesn't
12:34 matter where you are in the world right
12:36 and from that you could kind of see what
12:38 the future looked like you could kind of
12:40 look 10 years down the road and there
12:41 was a really clear picture of it I don't
12:44 I don't have that with AI because the
12:47 what these tools do is so profound and
12:50 it's going to so up and
12:53 everything from knowledge work to
12:55 creative work to
12:58 um the complexity of work to the
13:00 complexity of software all of this stuff
13:02 is going to be upended and and I can't
13:04 really look at it so I said this on a
13:07 on a panel I was on and a friend of mine
13:09 my nose is running like crazy
13:11 um
13:13 I am not doing coke if you think I'm
13:15 doing coke it's funny every time I see
13:17 someone scratch their nose I'm like yeah
13:18 cocaine
13:20 ah okay
13:23 um
13:24 so I said I said I couldn't see what the
13:26 future looked like with the say I stuff
13:27 and a buddy of mine reached out a guy I
13:29 really respect and he said he's been
13:30 thinking about a lot
13:32 and he's got kids that are like 10 and
13:34 12 or something like that
13:37 he said I've been trying to think about
13:38 what I have them
13:39 study in school and he said if I look
13:42 five years out I can't see it but but I
13:44 went 30 years out and he said here's
13:46 what I think it's going to be here's
13:47 what I think AI is going to do
13:50 he said
13:54 the entire world right now is run on
13:57 incredibly complex systems
14:00 that some of them are complex by Design
14:02 like health insurance certainly in
14:04 America is complex by Design right the
14:07 more complex they make it the less you
14:09 can actually get them to pay you because
14:11 they've make it there's so many you know
14:13 Hoops to jump through paying your taxes
14:16 is incredibly collect you know complex
14:19 buying a house buying a car refinancing
14:22 your house
14:24 um everything is is incredibly
14:26 complicated and there's whole Industries
14:27 built on
14:31 them dealing with the complexity he said
14:34 imagine a world 25 30 50 years out where
14:38 all the complexity is gone
14:40 where everyone has access to the best
14:44 accountant in the world like right now
14:45 all the tax loopholes just go to the
14:47 rich because the best accountants in the
14:50 world read all of the really complicated
14:52 tax codes and they figure out all of the
14:54 ways that you can move money around and
14:56 put things in the Cayman Islands and all
14:58 that [ __ ] they do well imagine when
15:00 you could just have an autonomous agent
15:02 an AI agent do that for you and I was
15:05 like holy [ __ ] I said that changes
15:07 everything like that would change the
15:08 tax code because all of a sudden their
15:11 revenue would plummet because everyone's
15:14 using the tax loopholes right and so so
15:17 they they would obviously fix some of
15:19 those tax loopholes
15:21 um but I think there's something about
15:22 that I think there's something about AI
15:25 is going to help us remove the
15:27 complexity of the systems that the
15:30 entire world's been built on
15:32 um and and it's
15:34 what I'm really excited about with with
15:36 that vision of the future is that it
15:39 allows us as humans to kind of return to
15:41 Being Human not being cogs in a machine
15:44 but you know just being human and
15:47 reconnecting and being self-expressed
15:49 and all that sort of stuff so that's a
15:52 possibility you know or it kills us all
15:54 [Laughter]
15:58 ah it's never a dull moment here at the
16:01 AI learning lab
16:02 all right opening igpt4 is also
16:05 available for developers by request I
16:08 have requested developer access to gpt4
16:10 for months and I don't have it so if you
16:13 know a secret let me know
16:15 um I also don't have access to the
16:18 browser plugin for chatgpt I also don't
16:20 have access to code interpreter so if
16:23 anyone is on here from openai give me
16:26 access damn it I deserve it it's called
16:27 the AI learning lab I talk about open AI
16:30 all the time like um it's like like it's
16:32 like I'm an apple Fanboy but for open AI
16:34 you should give me access to this stuff
16:36 right
16:38 tap that screen if you think I should
16:40 have access to all the goodies because
16:42 I'll tell you about them
16:44 all right no one's tapping the screen
16:45 fine don't tap the screen
16:49 um
16:51 thank you very much being able to
16:54 practically use this [ __ ]
16:57 is the hard part being able to
16:59 practically use this [ __ ] is the hard
17:00 part well
17:02 yes and no okay so here's been my
17:05 experience with it
17:07 if I just sit down at my computer with
17:09 Chachi PT in front of me or even even
17:11 some of these image models
17:14 I don't know what to do it's like I'm
17:16 staring at the blank screen right it's
17:18 like because it can do so much
17:21 I tend to do like a little parlor trick
17:23 like I'll have it write a poem or I'll
17:25 have it explain quantum physics like I'm
17:27 a five-year-old
17:29 so here's the trick that I found and I I
17:31 think this can help you is rather than
17:33 think about what can I do with chatgpt
17:37 um pick a project and and you can make
17:40 up a project pick a project either for
17:42 yourself or for someone else ideally for
17:45 someone else
17:46 so maybe it's a community group
17:48 maybe it's your work
17:50 um maybe at work people have been
17:52 bitching about how
17:53 the marketing isn't working anymore
17:56 you could say you know what
17:58 can I try and experiment can I create a
18:01 marketing plan for you I'm learning
18:03 about all these AI tools and I've heard
18:06 that it can write marketing plans like
18:07 do something for someone else a specific
18:09 project and if you use chat EBT for a
18:13 project like that it it totally
18:15 transforms
18:18 how quickly you learn what's possible
18:21 um because because you kind of stay
18:22 focused on solving the problem and the
18:24 other thing that shifted it for me is
18:26 this don't think of chat EBT like a
18:29 Google search engine where you just type
18:31 in a thing and then you get back a list
18:33 of responses and you're kind of done
18:34 think of it more like you've got a
18:37 creative collaborator
18:39 um like like a partner that you can
18:41 solve a problem with and go back and
18:43 forth with it so tell it you want a
18:45 marketing plan and then it'll write you
18:47 a shitty marketing plan and then you're
18:49 like oh no it needs to be way more
18:50 specific than this you know here's four
18:53 bullet points about my business and then
18:55 it'll give you a slightly better
18:56 marketing plan you're like well but we
18:58 don't have budget for paid media so I
19:00 need a marketing plan you know so so
19:01 just go back and forth with it do
19:03 something specific pick a pick a thing
19:05 it could be
19:07 um do meal planning for your family for
19:09 a week it could be
19:11 um I don't know organize an event for a
19:15 community center
19:16 um pick an event do it that'll that'll
19:18 help
19:19 um with with the practicality of what
19:21 these tools can do and then once you
19:23 sort of crack the nut of oh if it can do
19:25 that for that maybe you could do this
19:28 for this
19:29 besides having strict rubrics like
19:31 college kids have I don't know what the
19:33 beginning of that was oh
19:37 part of the one I don't quite know what
19:39 to do with just pick a project and solve
19:40 it yeah yeah trying to invent or or
19:44 research all the best prompts and like
19:47 have the perfect prompt it it's I don't
19:50 it's a mess I can't I can't work like
19:51 that either it just it just feels like
19:53 gratuitous use of Technology
19:56 um so I'm totally with you I'm trying to
19:57 figure out the practicality of it
20:00 okay sorry I'm scrolling back to God
20:03 damn Tick Tock in there
20:06 this is all my lives are is me bitching
20:09 about The Tick Tock interface because I
20:11 because I'm old I'm old and tired and I
20:14 don't like these newfangled
20:15 multi-layered interfaces
20:18 get off my lawn you kids get off my lawn
20:25 I'm getting Hearts I assume that means
20:27 you're you're liking me losing my
20:29 [ __ ] mind
20:32 oh my God
20:35 I don't know if I've told you this I'm
20:37 nuts
20:40 um okay so I got back to this uh best AI
20:43 way to revamp a resume for a different
20:46 industry this is
20:48 so cool so
20:51 so copy your resume
20:54 also copy your LinkedIn profile
20:58 um
20:59 then copy a job posting for for whatever
21:03 industry you want to go in
21:06 and so here I'll show you
21:10 all right so I'll go 3.5 so because even
21:14 if you don't have access to gpt4 so let
21:16 me go to my LinkedIn profile
21:18 and
21:22 hold on hold please and so I'm just
21:26 going to copy just a big pile of crap
21:29 from my
21:31 LinkedIn profile
21:34 and then I'm going to go back to chat
21:36 EBT I'm going to say okay
21:38 um based on
21:42 based on this
21:45 um Linked In
21:47 profile
21:49 and then I'm hitting shift return twice
21:51 and I'm pasting in just whatever I
21:53 copied out of LinkedIn
21:56 and let me see what else can I put in
21:58 there
22:00 um do I have a bio somewhere let me see
22:02 if I can find a bio
22:04 hold please
22:08 where is my Gmail
22:13 come on
22:16 uh-oh I think Microsoft Edge
22:20 its memory has gone gone away all right
22:24 um let's go
22:26 um
22:27 tile bio
22:29 see if I can find one here
22:34 did I give a bio here
22:37 yeah all right this is a crappy one but
22:39 whatever okay so here's a bio
22:42 copy so go back to chat DVD so I I said
22:45 based on this LinkedIn profile and then
22:48 I pasted it now I'm going to say and
22:50 this
22:52 um bio about AI
22:54 and I'm going to paste that
22:57 um
22:58 write a resume
23:02 for Kyle
23:05 um
23:06 that
23:08 positions his experiences
23:12 for a job in the
23:16 um
23:17 let's see gas and oil industry which I I
23:23 would never get hired for the gas and
23:26 oil industry all right so so it read all
23:28 that crap
23:30 and it is now I think writing something
23:36 so it says resume for Kyle Shannon
23:40 um
23:41 my objective is to apply my diverse
23:43 skills and experiences in digital
23:45 marketing storytelling Innovation and
23:47 Leadership to contrib can contribute to
23:50 the success of a gas and oil company
23:54 um
23:55 so it grabbed my storyvine experience
23:57 myagency.com experience so those are the
23:59 two you know big kind of core you know
24:02 sort of successful businesses that I've
24:04 had
24:05 then it grabs some of my AI stuff
24:08 um grabs my education grabs my skills so
24:11 so yeah so you can just use it to
24:14 um just give it a bunch of raw materials
24:16 and then tell it to rewrite it and if
24:17 you don't like what it wrote for you
24:19 just tell it it sucked and give you
24:21 other stuff
24:24 it's it's it really is quite remarkable
24:26 and you know using these tools I've said
24:28 this before on this channel that um
24:32 I think it was Arthur C Clarke said any
24:35 sufficiently advanced technology is
24:37 indistinguishable from Magic this shit's
24:40 Magic
24:45 all right
24:47 I love hearing you talk about this cool
24:49 I'm glad this this helps and makes sense
24:52 tree grown or negro I have no idea what
24:56 you're talking about
24:57 um when do I think Google and Microsoft
24:59 will release their Suite of AI
25:01 integrated apps I think Microsoft will
25:03 do it before Google will I think Google
25:05 is scrambling to catch up I I think they
25:08 have screwed themselves they invented
25:10 this [ __ ] they invented the core
25:12 technology that sits beneath all these
25:15 um
25:16 all the toys we're playing with today
25:18 um and then they kind of kept it behind
25:20 the firewall because I think they knew
25:22 that it would threaten their business
25:23 model so they are playing catch-up and
25:26 and the CEO of Microsoft is taking
25:29 particular Glee in [ __ ] with Google
25:33 um so I think as soon as they have
25:34 something that's solid they're gonna
25:36 they're gonna release it so I would
25:37 think that Microsoft is coming sooner
25:39 than Google but they're both coming
25:42 um
25:44 I don't know my guess would be
25:47 probably within three months but maybe
25:49 by the end of the year I don't know I
25:51 don't I mean you know the Office 365
25:54 Suite is an incredibly complicated
25:56 series of applications that have lots
25:59 and lots of features so it's not just
26:01 like sort of slapping on
26:03 a you know a GPT window on top of
26:06 Microsoft Word right like some of the
26:08 demos they showed were like you know
26:11 take this earning report that is in
26:16 um Microsoft Word and write a
26:18 presentation about that and so they're
26:21 having to integrate it across the suite
26:22 and they're probably incorporating it
26:24 with Outlook emails and they're probably
26:28 incorporating it
26:30 um with uh
26:32 uh with
26:34 um you know their CRM Dynamics 365. so
26:37 so it's probably pretty complicated but
26:39 what I've seen from how they integrated
26:41 gpt4 into Bing and Edge is very well
26:45 thought out even the early stuff was
26:47 quite well thought out so it's coming
26:49 soon I'm actually quite excited about it
26:51 is it a good time to jump into ml what
26:53 type of work should one get into in ml
26:57 yes I think it's not only a good time to
27:00 get into ml I I think it's I think it's
27:04 quite important like why I do this
27:06 channel is I think it's imperative that
27:09 anyone who's doing any kind of work with
27:12 computers in any way any kind of
27:14 knowledge work any kind of creative work
27:16 um
27:17 the amount of disruption that's coming
27:19 thank you for that the amount of
27:20 disruption that's coming is massive and
27:23 I think that the people that learn it
27:26 now are going to have big opportunity
27:28 and those that don't are kind of going
27:30 to get blindsided
27:33 um what type of work should you get into
27:35 I would say what are you good at
27:37 um I think the skills moving forward are
27:39 going to be things like those that are
27:41 good at curating Big Ideas strategy
27:43 thinking horizontally like t-shaped
27:46 skills someone said to me yesterday
27:48 which I forgot about that term but so
27:51 the the the sort of vertical part of a T
27:53 is like a vertical skill like
27:54 copywriting or illustration or
27:58 um uh programming engineering
28:00 the horizontal thing is if you if you
28:04 accept that these tools are going to
28:06 make it very very easy to do the
28:07 executional part of all the vertical
28:10 things then what we need is people that
28:12 can you know operate across the
28:15 horizontal part of the T so think
28:18 avoid being a virtuoso that's only good
28:21 at one thing and think more like being a
28:23 conductor of the symphony
28:25 so learn as much as you can take what
28:28 skills you're good at if you're really
28:29 good at project management figure out
28:31 how to do Kick-Ass project management
28:33 with ML if you're really good at
28:35 creative
28:36 figure out all the tools that you know
28:38 amplify your creativity so I would say
28:40 take your core
28:42 the core thing of what you enjoy and
28:44 what you're good at you know hopefully
28:45 those things are decently aligned and
28:48 then have
28:49 that's where I would focus uh on on the
28:52 ml stuff I am an influencer yeah I got
28:54 to get a pink stripe in my hair to truly
28:57 be an influencer or twerk I will not
28:59 twerk you're welcome
29:01 uh uh uh uh
29:04 I'm an expert on AI ask me anything I
29:07 assume that's a troll for me saying
29:10 I shouldn't be answering questions on AI
29:13 I'm not an expert I'm just learning and
29:15 sharing what I learn along the way so
29:17 feel free to troll I don't give a [ __ ]
29:21 um I have a question can government stop
29:22 it from being super cool
29:24 government can stop everything from
29:26 being super cool government doesn't
29:28 understand how Wi-Fi [ __ ] works so I
29:31 I'm a little terrified about government
29:33 getting their hands on AI
29:34 here's the thing
29:36 I think the thing that that would have
29:38 kept all this stuff from being cool is
29:41 if it had gone the way it was going
29:42 which was Google invented it kept it
29:45 behind the firewall Meadow was keeping
29:48 it behind the firewall all these
29:49 companies were kind of keeping this AI
29:51 stuff behind the firewall it was sort of
29:52 living in the AI and and machine
29:56 learning is for the data scientists and
29:59 for big corporations
30:01 and then open AI said you know we're
30:04 going to put this out there and
30:05 mid-journey started doing this with
30:07 images and stable diffusion started open
30:09 sourcing their images and then Chachi PT
30:12 caught fire November 30th 2022 is the
30:15 date that was launched I think that will
30:18 be the day that goes down in infamy for
30:20 AI because within six weeks 100 million
30:23 people were using chat GPT I don't even
30:25 know what the numbers are now but I
30:26 would assume it's two or three hundred
30:28 million it's insane
30:30 um
30:31 so
30:33 because of that because of the success
30:36 of chat GPT
30:38 and the fact that Microsoft you know got
30:40 this sort of you know glow of of being
30:44 ahead of the curve with this chat gbt
30:46 stuff
30:47 it forced Google Out of the Shadows to
30:50 say well we've got AI too right then it
30:53 forced meta Out of the Shadows Apple's
30:55 coming next
30:57 um and then these tools like metas
31:00 uh large language models leaked right
31:03 and then the open source came out
31:04 community pounced on those and now
31:07 there's a number of Open Source large
31:09 language models stability just released
31:11 one
31:12 so now that this stuff is proliferating
31:16 and the entire development community and
31:18 the entire investment Community everyone
31:20 is putting focus on this
31:22 um I don't think any single entity is
31:26 going to be able to Reign it in anymore
31:27 I think they could have
31:29 um
31:30 six months ago but I but tattoo PT
31:32 changed everything in my opinion
31:38 how are you associated in AI good
31:41 question I have a degree in acting eh
31:45 um
31:46 I just so so I
31:48 ah
31:50 I love like I love figuring out whatever
31:53 is the new thing
31:54 um
31:55 and always have and I did it
31:59 um in 1994 when I when I kind of
32:02 stumbled upon this thing called the
32:03 world wide web and I started one of the
32:05 first online magazines and one of the
32:07 first digital agencies company called
32:09 agency.com and built a lot of the first
32:12 websites for the Fortune 500 like we
32:14 built British Airways website we built
32:16 their ticketing system so it was wild
32:19 times so
32:21 um
32:22 I thought that would be the last time in
32:24 my life that I would see anything that
32:26 was a technological Revolution that was
32:29 going to change the game you could argue
32:31 maybe that the iPhone was one of those
32:32 but
32:33 I kind of see the iPhone as just kind of
32:35 an extension of you know it's computers
32:38 just got smaller and you're still kind
32:40 of using the internet on phones right
32:43 but it wasn't it wasn't necessarily A as
32:45 big a shift as as as
32:47 the popular internet was the world wide
32:50 web
32:51 um and then and then I stumbled upon AI
32:53 so I I spent about six months just
32:57 playing with images I was using a an nft
32:59 project called pixelmind
33:02 um to make stable diffusion images and
33:04 mint those into nfts and then I decided
33:07 to geek out a little bit
33:09 and install stable diffusion and that's
33:11 the the pit where I just made thousands
33:14 and thousands of of images
33:16 um you know I uploaded you know images
33:18 of me into dream booth and then made you
33:22 know all these kind of crazy images
33:24 of myself and I put together an art
33:27 project for that and then kind of in the
33:29 middle of doing that art project
33:32 um chat EBT came out and I had an
33:34 epiphany I had I I uh my Epiphany with
33:38 chat GPT was
33:40 um
33:42 I was I was just watching TV with my
33:45 wife one night it's like three weeks
33:46 after it came out and I thought oh I saw
33:48 Tick Tock videos that you can do coding
33:50 on it'll do coding
33:53 and so I said well I've heard Python's a
33:55 cool language so I said hey chat EBT
33:57 write me a python code or python script
34:00 to take an input or just do something
34:02 like that
34:03 and it did and then I realized I don't
34:05 know how to run python I I'm into
34:08 programming in decades and so I said hey
34:11 chat EBT I don't know how to run python
34:13 it said it said oh if you're on a Mac
34:15 open your terminal and type in this
34:16 command line and I'm like I don't want
34:18 to do command line is there a website I
34:20 can just you know paste some code and
34:22 have it run and said sure here's the
34:23 three websites and so I picked one
34:25 and it said paste the code in the left
34:27 hand window hit the Run button so I did
34:30 that and it gave me an error and so I
34:31 copied and pasted the error and I put it
34:34 in and it told me oh you don't have this
34:36 Library installed and it told me how to
34:37 install the library and then I got the
34:39 thing running within 90 minutes I had
34:41 written
34:43 a fully functional python application
34:45 that asks someone about a topic and
34:48 asked a few questions about it and then
34:50 wrote five pieces of social media
34:51 content using the openai API
34:54 within 90 minutes never in a line of
34:56 python in my life still haven't written
34:58 a line of python in my life and I saw I
34:59 had this Epiphany and I just had this
35:01 other moment of
35:04 holy [ __ ] the world just changed and no
35:06 one knows it and
35:09 um I started the AI salon so I have a
35:13 bi-weekly Salon where we meet we have a
35:15 Discord server I started this Tick Tock
35:17 Channel I started the everyday AI
35:20 newsletter with my friend Greg mushin
35:23 um I just went
35:25 all in I went you know full ADHD hyper
35:29 Focus
35:30 just [ __ ] dove in it for
35:33 well I haven't stopped right like I
35:35 started kind of second week of December
35:38 or something like that and I just
35:39 haven't stopped and I'm unsufferable
35:41 right now it's all I talk about people
35:42 can't talk to me
35:44 um
35:45 and um yeah that's how I got into it but
35:48 but I'm you know I'm I'm like rolling AI
35:50 into my business I've completely
35:52 rethought what my business product is my
35:55 company's called storyvine it's a
35:56 automated video storytelling platform
35:59 I've got a whole new vision for that
36:00 it's really exciting so
36:02 that's how I'm Associated I just get
36:05 curious about [ __ ] and I think this is
36:08 this is the next big thing I I think AI
36:11 is significantly more profound than the
36:14 World Wide Web was I think it's
36:16 happening way faster well and here's a
36:19 point
36:19 um the World Wide Web took eight years
36:21 to get from zero to 100 million users
36:24 Chachi PT took six weeks so this is
36:27 coming fast and the sophistication and
36:30 the power of these tools is like nothing
36:32 we've ever seen this is Industrial
36:35 Revolution level printing press level
36:39 um change and and it's coming fast
36:43 um possible to change the camera angle
36:45 so we don't have to look up your nose I
36:47 could I'd have to put my computer higher
36:49 but yeah you get to see my neck and my
36:51 boogers tough [ __ ] sorry
36:54 I'll I'll improve my when I when I film
36:58 my static videos I film those at like a
37:00 45 degree angle up to the ceiling so you
37:02 don't have to see up my nose but I'll
37:04 I'll fix it at some point but sorry I'm
37:08 just gonna have to deal with my my uh
37:11 bad production values AI is a very Star
37:14 Trek Vibe about it yeah I like I I just
37:17 I've said I said this last night I feel
37:18 like we're moving into this post sci-fi
37:20 world where almost everything in sci-fi
37:23 is now starting to happen like time
37:25 travel is the only one that's not quite
37:27 there but all the other shit's happening
37:29 it's like Jesus
37:31 can technical intelligence control
37:33 blockchain
37:35 well
37:37 yeah I mean
37:41 you'll be able to create contracts and
37:44 things like that and with things like
37:45 Auto GPT you'll be able to
37:48 um
37:48 you know execute things on the
37:50 blockchain but I I think that blockchain
37:52 is actually going to become very very
37:54 important for
37:56 um because of AI I think AI is going to
37:58 drive web3 adoption because
38:01 we're quickly moving I don't know if
38:03 you've seen the there's a there's a a
38:05 2024 political ad about Biden and Kamala
38:09 Harris that's fully AI generated and and
38:11 it's there's nothing special about the
38:14 the the ad particularly it's just kind
38:17 of a run-of-the-mill attack ad
38:19 um but it's all generated with with this
38:22 AI stuff but but we're coming into a
38:24 world where
38:25 um
38:26 the ease with which we're going to be
38:28 able to generate content and
38:30 misinformation I think is going to lead
38:32 to a
38:34 a backlash to to be able to verify the
38:37 authenticity of something so if I can
38:40 say yeah this definitely came from CNN
38:42 this piece hasn't been altered in any
38:44 way
38:45 blockchain's a perfect way to verify
38:47 that stuff so I think blockchain and
38:49 nfts and all that technology
38:51 um is is gonna quickly become a tool
38:55 that that creates some level of
38:57 authenticity
38:58 um that I think we're going to need and
39:00 I think I think it's going to get really
39:02 ugly for for I think this election cycle
39:04 is just going to be a flipping disaster
39:06 so we'll see
39:08 um so I think it could actually drive
39:10 the adoption of blockchain also the
39:12 other thing AI does for for blockchain
39:14 and nfts is
39:17 um
39:18 the World building stuff so creating all
39:20 the metaverse stuff right now you have
39:22 to hire like you know rooms and rooms of
39:25 people to to do all the 3D modeling and
39:27 you have to pick a look and feel and you
39:29 kind of have to lock that [ __ ] down
39:30 we're quickly moving into into a world
39:33 where you just ask for a world and it
39:35 will be generated on your behalf and
39:37 then you enter it right so so I think I
39:39 think all sorts of cool things are gonna
39:41 happen between web3 and AI is an AI just
39:45 a database it only knows what it is told
39:50 well
39:52 yeah yeah I mean technically it's a
39:55 database it's a
39:56 it's a it's a database of embeddings
39:58 it's a vector database so but but it's a
40:02 vector database that's kind of been
40:03 trained on all the words right like at
40:06 least all the words of the public
40:07 internet and as these things get more
40:09 and more sophisticated it's gonna just
40:11 kind of be all the words so I kind of
40:13 think of it like this I think of AI more
40:16 like AI is a reflection of humanity
40:19 um where it kind of learns all of the
40:21 words and their semantic relationships
40:23 and then when you start interacting with
40:25 it it's pulling those semantic
40:27 relationships not unlike how a brain
40:30 does like
40:31 a lot of these AI tools you know are
40:33 based on neural networks which are based
40:36 on biology they're based on how our
40:37 brains work so what it's actually doing
40:39 if you look at how these things actually
40:42 work it's predicting the next word that
40:44 is the most likely response to the big
40:46 prompt that you gave it and it's doing
40:48 that over and over and over again and I
40:50 someone told me yesterday that it's
40:51 doing it on phrases now
40:54 but it's not cop Y and pasting crap out
40:56 of the database it's because it knows
40:59 the semantic meaning of all these words
41:00 it's actually composing original content
41:03 based on what you ask it
41:06 um so so it's it's not just a parrot
41:08 that's that's I think that's one of the
41:09 misconceptions so AI will destroy the
41:11 world I think that's it's easy to go
41:13 there that's what Hollywood has taught
41:15 us surveillance will be out of control I
41:17 do agree with that surveillance will be
41:20 out of control and I don't think it'll
41:21 just be big brother I think it'll be all
41:23 of us by the way on the Privacy thing
41:25 with AI we already gave up our privacy
41:27 20 years ago we gave it to Google and
41:30 and Facebook you know that old adage If
41:33 the product is free you're the product
41:35 do I think AI will launch nuclear
41:37 weapons I hope to God they have
41:41 controls in place that that computers
41:44 can't just launch that [ __ ] I'm pretty
41:46 sure they do I'm concerned about
41:48 surveillance close your window
41:51 um
41:53 I just applied for a job at open AI cool
41:55 I gave a speech
41:57 to uh creative mornings two weeks ago
41:59 and this woman came up at the end she
42:01 goes uh I train models at open AI I was
42:04 like ah it's the coolest thing
42:05 it's like it's the hip company to work
42:08 for
42:09 um I've I applied for code interpreter
42:11 multiple times too I know give me access
42:13 to code interpreter I'm an influencer
42:16 man and I want to talk about you too
42:19 what are your thoughts about rewind AI I
42:22 don't know it let's go look at it
42:25 rewind I assume it's a video thing wind
42:28 dot AI is that it
42:33 [Music]
42:34 what if you had Perfect Memory
42:38 about anything oh yeah so so I think
42:41 this kind of stuff's cool so so what
42:43 this is is
42:45 um I assume you kind of aim it at your
42:46 social media or you aim it at your email
42:48 or something like that so so what this
42:51 thing's doing is it's taking all of your
42:53 personal data it's throwing it in an
42:55 embedding database in into a vector
42:57 database after embedding it and then
43:00 it's basically making it like chat GPT
43:02 searchable so you can you can
43:03 essentially interact with your your your
43:07 um digital footprint just like you
43:09 interact with chat GPT I think it's I
43:11 think these kind of things are brilliant
43:13 um apple has a big their developer
43:15 conference is in June and they've got
43:17 big announcements coming there I think
43:19 they're AR VR goggles are coming out
43:22 I have a sneaking suspicion and I've
43:24 heard Robert scoble said a little things
43:27 about it that they're going to be doing
43:29 something with AI I think Apple could be
43:32 really good at agents
43:34 um so so AI that take actions on your
43:37 behalf and things that exploit your data
43:41 because their data privacy stuff is
43:43 really good
43:44 um I think Apple could be really good at
43:46 this stuff so I haven't played with
43:47 rewind but thanks thanks for popping
43:49 that up I'll I'll go play with that
43:50 because
43:51 um this idea of being able to mine our
43:55 own data whether it's business data or
43:57 personal data is is going to be like a
44:00 flip and super power right imagine being
44:02 able to say I know I I
44:05 I know I said in email at some point in
44:08 the last decade something about X Y or Z
44:11 and you know how you spend like four
44:13 hours trying to find that and the next
44:15 two weeks trying to obsess about where
44:16 did I say it who did I say it to did I
44:18 put it in a word doc where is it imagine
44:21 just being able to go did I ever say
44:23 this boom yes here it is like that kind
44:25 of stuff's coming so I'm pretty excited
44:27 about this any alternatives to code
44:29 interpreter while we wait yeah you could
44:31 play with Lang chain
44:33 um it's not exactly the same thing but
44:35 Lang chain allows you to kind of point
44:38 um these tools at you know say a
44:41 directory of your data a directory of
44:43 you know spreadsheets or other kinds of
44:45 data and then you embed them and you put
44:48 them in a vector database like pine cone
44:50 and then you can search on them so I
44:52 think
44:53 um
44:54 um code interpreters just a fancier
44:56 version of that kind of in the form of a
44:58 plug-in but Lang chain is probably the
45:00 tool that you could use in the meantime
45:03 and there's there's there's decent
45:05 videos on it it's still you have to be a
45:07 little geeky to be able to do it but
45:09 have cheat chat GPT teach you how to do
45:11 it
45:12 how long have I been in the AI business
45:14 I'm not am I in the AI business yeah
45:17 sort of I guess I mean I'm not really
45:19 making money on AI but um
45:22 I've been
45:24 curious about AI for about a year I've
45:27 been obsessed with generative AI for
45:30 about six months and
45:32 um I've been insufferable
45:36 talking like this about AI for about
45:39 four months so four or five months
45:41 pretty new and here's an important thing
45:43 I I get asked a couple of times on these
45:46 lives like am I in in machine learning
45:48 or you know am I somehow involved in the
45:50 model generation side of this business
45:52 I'm not so so this to me is where
45:54 there's another parallel to the early
45:56 days of the web where the internet had
45:59 been around for decades and and the
46:01 people that created the internet and
46:02 most of the users of the internet early
46:04 on were scientists and researchers and
46:06 things like that the World Wide Web
46:08 comes along and it it was a tool that
46:11 gave access to creatives and business
46:14 people and and just basically anyone
46:16 access to this really powerful thing
46:18 called the internet
46:20 um chat gbt in these and these image
46:22 generation models and now all these
46:24 other tools are are the same thing
46:26 so I would say that I represent
46:29 um sort of a new class of AI
46:33 um you know super user which is not
46:35 um I'm powerful on the model generation
46:38 and training side I'm more powerful on
46:40 now that these models have been
46:42 generated and trained and we have access
46:44 to them how can we use those in our
46:46 business in our lives things like that
46:48 so so it's more about the application of
46:50 these tools rather than the creation of
46:51 these tools I think that's a massive
46:53 shift we're we're right in the middle of
46:55 which AI platform should I use if you're
46:58 just getting started chat GPT is the big
47:00 one and I think it's also worth
47:02 um generating images just see experience
47:05 what it's like to write words of like a
47:08 penguin in a pickup truck driving over a
47:11 cliff
47:12 and watch this thing generate that for
47:14 you that experience is like magic right
47:18 so just start with chat GPT start with
47:20 image generation and then the other
47:22 thing is this don't just sit in front of
47:24 them wondering what to do pick a project
47:26 and and do that project using chat like
47:29 learn chat GPT by solving a problem for
47:32 someone else it'll totally transform how
47:34 you think about it
47:36 what are your thoughts on Bard my
47:37 thoughts on Bard are that it's not as
47:39 good as Chet GPT
47:41 um
47:42 I heard it's getting better at coding I
47:44 think
47:45 um Google is a company full of Engineers
47:49 they're not as good at the human empathy
47:51 stuff so I think the chat EBT has got a
47:54 real leg up on that I think that's
47:56 something Google's going to struggle
47:57 with but I also think that Google's got
47:59 you know a gazillion dollars and a
48:02 gazillion smart people and they can
48:03 figure this [ __ ] out so I think they're
48:05 going to get better really quick but
48:08 Google just there was a a memo that
48:11 leaked
48:12 um two days ago from a Google employee
48:14 that that you know they they feel that
48:16 these open source large language models
48:19 are getting good enough fast enough that
48:22 they don't think that they Google or
48:24 open AI or anyone can really maintain
48:27 um sort of technical dominance on the
48:29 large language model side so I think
48:31 large language models are just going to
48:33 become ubiquitous like the internet
48:35 they're going to be everywhere and
48:36 they're going to all be good
48:38 um so I think Bard's behind the eight
48:39 ball and you can use it I just haven't
48:41 used it because chechi BT is so good all
48:44 right I just lost my place again dang it
48:46 dang it tick tock
48:48 somebody should know someone
48:52 I don't know what happens it's like when
48:53 I start scrolling it like updates the
48:55 last thing and it jumps me all the way
48:57 to the bottom and then I gotta scroll
48:58 back up
49:00 if anyone knows how to use this app let
49:02 me know
49:04 like the the guy that doesn't want to
49:05 look up on my nostrils
49:09 [Music]
49:12 what app am I using to show my screen
49:14 it's just Tick Tock there's a when you
49:15 go go live there's a dual I forgot it's
49:18 called like dual camera
49:19 so
49:21 that's just my back camera and then I
49:24 which chat gbt platform are you using
49:26 just
49:28 chat.openai.com don't you don't need to
49:30 use anything else other than that but I
49:32 will give you this
49:36 um
49:39 trying to get it to focus hold please
49:41 God damn it
49:43 this is like the watch Kyle tried to get
49:46 his technology work Channel
49:49 these four things are are what I would
49:51 use so so one of the things that just
49:55 burns my engines
49:58 I couldn't think of the phrase so that's
50:01 wrong one of the things that pisses me
50:03 off is um
50:05 there's so many apps out there that are
50:08 trying to exploit people searching for
50:10 chat GPT so it's not an app it's not an
50:12 app on the Apple Store it's not an app
50:15 on Google Play
50:18 chat GPT exists at
50:21 chat.openai.com so just point your
50:23 browser whether you're on a phone or a
50:24 web browser on your on your desktop
50:29 um chat.openai.com is chatgpt if you
50:32 want to use chat ept4 you can pay them
50:34 20 bucks a month
50:36 um you can also use the openai
50:38 playground that's platform.openai.com
50:40 playground that's sort of the more geeky
50:43 version that's the engine that sits
50:45 under the hood of chat EPT
50:47 um but Bing this week went live so
50:50 there's no more wait list for Bing so if
50:52 you go to bing.com you're actually using
50:55 gpt4 that's connected to the internet
50:57 it's a little bit neutered because they
51:00 got in trouble with the New York Times
51:01 so it's getting better and better so
51:03 it's not quite as good as the open AI
51:05 chat gpt4 but it's free and you can also
51:09 download the edge browser and that's got
51:11 it in there and then finally poe.com
51:13 poe.com they do have an app as well Poe
51:16 and Poe on the two app stores
51:19 um what's cool about Poe is you can use
51:22 chat gbt 3.5 chat gbt4 you can also use
51:27 anthropics Claude and the new Claude
51:29 plus which I think is anthropic's kind
51:31 of gpt4 kind of competitor so po.com is
51:36 really cool you can also make your own
51:37 botswithpo.com so hope that helps
51:41 eh
51:49 sorry I'm trying to find my way again
51:52 just purchased an auto renewal
51:54 subscription Auto PT seems to create
51:58 endless tasks for a simple goal what am
52:00 I doing wrong that's good what's a good
52:02 use case Okay so if you don't know what
52:05 Auto GPT is it's it's part of a thing
52:07 that are called autonomous agents and
52:10 they are what they sound like
52:12 they are tools that just
52:15 Auto
52:17 prompt themselves and then use tools
52:19 like searching Google and
52:21 um using large language models to prompt
52:24 themselves here's the thing about Auto
52:26 GPT it is brand new it's maybe six weeks
52:30 old I think it's five and a half weeks
52:32 old maybe six weeks old it is
52:35 um really janky
52:37 janky is a technical term for it sucks
52:40 it's shitty software
52:43 um
52:43 it will very often get in these
52:45 recursive Loops where it just is keeps
52:47 running and running and it can't do
52:49 anything it's over complicated so so
52:52 um I would say give it really simple
52:54 tasks but but I don't know I I mean
52:57 I think I think the interesting thing
53:00 to do with auto GPT right now is kind of
53:03 play with it enough to understand how it
53:05 works and see if you can get it to do
53:06 things in any kind of predictable way
53:09 um but don't take it too seriously
53:11 unless you're a real developer and you
53:13 really want to develop something around
53:14 the tool or do a spec you know a
53:17 specific version of it then you can play
53:19 with it but if you're just kind of
53:20 [ __ ] around with it I'd say just keep
53:22 playing with it if you want to play with
53:24 auto GPT and you're not a developer you
53:28 can go to I think it's agent GPT agent
53:34 gpt.com probably not
53:37 nope God damn it hang on agent GPT
53:43 gotta goes God damn it agent
53:47 agent
53:49 GPT
53:52 uh okay it's
53:54 agent GPT got reworked without the e so
53:59 r e w o r k d dot a I so agent GPT so so
54:06 what you do with these agents is you
54:07 give them a goal like here let me do one
54:09 now I'll do
54:11 um Story vines my company storyline
54:13 agent we'll call this thing we'll see if
54:16 it works it's kind of cool when you
54:18 watch watch them do what they do the
54:19 goal is going to be find the
54:23 marketing taglines of
54:27 the top 10 competitors to storyvine
54:36 so I got this to work once and it was
54:38 pretty slick and then I got it to not
54:40 work like 10 times
54:42 so what happens is so so with these Auto
54:44 agents you just give them a goal like go
54:47 find all this stuff
54:49 and then you say deploy agent oh I've
54:53 got to sign in
54:54 hold please
54:59 I'd have to retype this crap God damn it
55:02 see it's early software I told you
55:04 that's all crappy right now
55:06 Story vine agent
55:10 or agnet that's fine
55:12 um
55:13 please
55:15 find the top
55:18 find the marketing taglines for the
55:23 top 10 compete it tours to storyvine
55:33 with a typo and then I hit deploy agent
55:36 and so now it says thinking
55:40 and it says added task search for the
55:42 top 10 competitors of storybine in the
55:45 market added tasks and align the mark
55:47 analyze the marketing taglines
55:51 from storyline's competitors then it
55:53 says completing search for the top one
55:54 so it says storyvine
55:59 um
56:00 some of the top competitors include
56:02 include vidyard yes wistia bright Cove
56:05 brightcove's not a competitor
56:08 um panopto Vimeo kaltura Sprout video
56:11 hippo video bomb bomb and dub so about
56:14 half of those are competitors half of
56:16 those I would consider you know um
56:18 complementary platforms
56:22 um and then it says uh completed analyze
56:25 the marketing taglines of each
56:27 competitor understood my task is go find
56:29 them so it's giving itself tasks let's
56:32 see has it found anything then it gave
56:35 itself more tasks
56:38 um my task so it's still trying to find
56:40 tag lines now it's giving itself more
56:42 tasks and it says we're sorry because
56:45 this is a demo we cannot have our agents
56:47 running for too long so this one
56:48 basically failed
56:49 um but that's how they work and and
56:51 they're they're pretty slick
56:53 I mean they're pretty slick conceptually
56:54 right it's it's a very cool idea
56:57 um because they're so new and because
56:59 it's actually expensive like uh some
57:01 people have gotten in a decent amount of
57:03 you know Financial
57:04 um trouble because they sort of turn
57:06 these things on and they just go run a
57:08 bunch of really expensive queries to all
57:10 these apis and they run their bills up
57:12 really quick because this thing's just
57:14 executing in the background
57:16 um so you can get in trouble with them
57:17 but I would say just go play with them
57:18 unless you're a developer and in that
57:21 case you know go to some of the GitHub
57:23 repositories and make them suck less
57:25 would be nice
57:30 let's get another picture here
57:32 oh here I got a mob version of me that's
57:35 a good one
57:37 um
57:41 [Music]
57:43 can
57:44 GPT can also write resume for you and
57:47 edit it with all your info yeah it's
57:49 pretty amazing it's pretty amazing
57:56 really good at is if you give it give it
57:59 examples
58:00 so if you go find you know there's
58:03 there's all these um
58:05 courses on how to write a resume right
58:08 and and and you know some of them say
58:10 here's the perfect you know structure of
58:12 a resume
58:13 you can copy and paste a resume that you
58:16 like the structure of and say based on
58:19 this resume
58:21 you know rewrite this resume and then
58:23 paste your resume in to match the
58:25 structure of the one above and it'll
58:27 reformat it like the the one that you
58:29 liked and fill in all the gaps and
58:32 pretty remarkable
58:36 someone got a number one gift or badge
58:38 good for you you can use my platform to
58:41 get your own badges thank you for
58:44 sending me love I appreciate it
58:46 um hello please what app are you using
58:48 to show your screen there you said that
58:51 um it's just the tick tock dual dual
58:53 camera app when you go live how would
58:55 you go about
58:57 a business plan
58:59 um I did that I actually wrote a
59:00 proposal today for for a big client to
59:03 do so I'm doing I'm starting to do a lot
59:06 more workshops on AI for for companies
59:09 um so if you want to do that let me know
59:11 um and so I took some ideas that we had
59:14 for the workshop I took some results of
59:17 a survey that we got I took some timing
59:19 that we talked about
59:21 and then I took basic sort of pricing
59:23 ideas and I literally just pasted a
59:26 bunch of those into chat gbt and then I
59:28 said write me a proposal with a lovely
59:30 intro
59:32 um for this thing and it knocked it out
59:33 and kind of within five iterations I had
59:36 a proposal that was it was as good as
59:38 anything I would have written within
59:39 about five iterations um so I do the
59:41 same thing with a business plan
59:43 um what you can do with a business plan
59:46 depending on what you're trying to do
59:47 with the business if it's a small
59:48 business that's one thing if you're
59:50 trying to get Venture Capital that's
59:51 another thing if you're trying to
59:54 um you know modify an existing business
59:56 that's another thing but I would I would
59:59 have GPT act as a venture capitalist or
1:00:02 is a business coach or a business
1:00:04 consultant and if there's a particular
1:00:07 kind of approach to business that you
1:00:09 like like tell it what kind of business
1:00:11 you're in and tell it that you want a
1:00:13 business plan that takes into account
1:00:15 whatever the realities are I'm not very
1:00:18 well funded or I am very well funded or
1:00:21 um you know I've been in business for
1:00:23 five years and I want to completely you
1:00:25 know make a pivot and so I need to write
1:00:26 a new business plan based on that so
1:00:29 give it lots of context it's one of the
1:00:31 things that it responds well to is
1:00:34 complicated prompts it it does really
1:00:36 well with complicated prompts so if you
1:00:38 kind of talk to it like you know if you
1:00:42 sat down with a business consultant the
1:00:44 first thing you would do is you'd spend
1:00:45 the first hour kind of telling it you're
1:00:47 telling that person your story right
1:00:50 think of it the same way kind of tell it
1:00:52 your story tell it what you want tell it
1:00:54 what its role is so your role is to be
1:00:56 this business consultant and my role is
1:00:58 I'm running this business and want to do
1:01:00 a business plan and it will knock it out
1:01:02 for you
1:01:03 pretty cool pretty cool stuff
1:01:05 what's the best chat GPT app just chat
1:01:08 jpt no app
1:01:10 you can usepo.com because you can use
1:01:12 different models there
1:01:14 but no app
1:01:16 um you can usepo.com because you can use
1:01:18 different models there
1:01:20 but I wouldn't pay for anything unless
1:01:22 you're going to pay for it just go to
1:01:23 openai dot or chat.openai.com that's the
1:01:27 best chat GPT app period hands down for
1:01:30 now
1:01:31 it may change when it does I'll tell you
1:01:33 pay attention to build conference
1:01:36 they will talk a lot about AI co-pilots
1:01:39 for Dynamics 365. interesting yeah I
1:01:42 think I mean Microsoft going all in on
1:01:46 incorporating
1:01:47 these these large language models
1:01:49 throughout their their portfolio from
1:01:53 from GitHub to Office 365 to Dynamics to
1:01:57 Azure is I think it's brilliant and and
1:02:01 they're doing it they're doing it well
1:02:03 can the AIS talk to each other they'll
1:02:05 be able to and those Auto GPT tools are
1:02:09 are connecting multiple tools together
1:02:11 and they are talking to each other so
1:02:13 that's what those things are doing what
1:02:15 is ML MLS stands for machine learning
1:02:17 which is basically
1:02:19 we're calling all this stuff AI but it's
1:02:21 really ml it's really machine learning
1:02:24 and um it's it's just basically
1:02:27 um the act of of getting these machines
1:02:30 to learn like humans do
1:02:33 um it has been the goal
1:02:36 um ai's easier people know sort of what
1:02:39 that means they've at least heard it in
1:02:41 a movie
1:02:43 um
1:02:44 new here what's ml oh machine learning
1:02:46 oh I've got three people that just ask
1:02:48 that including unicorn farts that's a
1:02:51 fun name
1:02:53 um wait why are there four people
1:02:54 talking about what's ml I guess there
1:02:56 was I either I mentioned it or or uh
1:02:59 someone asked about it and I didn't
1:03:00 explain it machine learning so sorry
1:03:02 about that if I if I I try not to use
1:03:05 acronyms
1:03:07 um what if they gave all the NPCs in GTA
1:03:10 their own AI personality so I assume you
1:03:13 know this but if you don't go look for
1:03:15 this paper
1:03:18 um go look for
1:03:21 what is it generative agents is that
1:03:23 what it's called gen
1:03:26 or God damn it erotive agents
1:03:34 uh
1:03:35 see what it's called
1:03:39 hold please
1:03:44 NPC agents
1:03:47 llm
1:03:53 okay here it is okay
1:03:56 so what's the paper called yeah
1:03:58 generative agents okay so this is an RS
1:04:01 Technica article
1:04:03 but
1:04:04 so so that's what this paper is about so
1:04:07 so what this paper is about so you have
1:04:09 this kind of sim sim life kind of game
1:04:12 here and it's got 25 NPCs but these NPCs
1:04:18 I'll use chat GPT to talk to one another
1:04:21 and they have memory and and they have
1:04:23 they're like a signed jobs so like the
1:04:25 artist wakes up and makes themselves
1:04:27 breakfast and then they go do art and
1:04:29 the farmer wakes up and they make
1:04:31 themselves breakfast and then they go
1:04:32 farm and they all kind of walk around
1:04:34 this town and when they encounter one
1:04:36 another they actually have a
1:04:38 conversation and it's not pre-programmed
1:04:40 like all NPCs are it's a generative
1:04:42 conversation where they're using chat
1:04:45 GPT to talk to one another and they
1:04:48 remember [ __ ] they like walk around
1:04:50 acting like humans so one of the
1:04:52 examples they they talk about in this
1:04:54 paper is
1:04:56 um
1:04:56 um one of the agents finds out it's the
1:04:59 other one's birthday and plans a
1:05:01 birthday party for it and like tells
1:05:03 tells the other agents they're having a
1:05:05 birthday party so they start planning
1:05:07 this birthday party so they remember
1:05:08 from day to day what they did in the
1:05:11 previous days so
1:05:12 um this paper's out there and and I
1:05:15 think I I think this idea of of these
1:05:17 agents that kind of have these lives of
1:05:19 their own is really really interesting
1:05:21 and and again this is something I think
1:05:24 Apple's going to be really good at I can
1:05:26 imagine that we will have
1:05:29 um different kinds of agents that that
1:05:32 that represent us right so I might have
1:05:34 one that is like my entertainment agent
1:05:37 and I want them to go out and just find
1:05:39 good music for me based on the music
1:05:41 that I like and what I listen to and
1:05:43 just if it finds a cool you know local
1:05:46 concert that's free or fits within my
1:05:48 budget it just buys me the tickets and
1:05:50 it checks my calendar and then I might
1:05:52 have a legal agent that goes off and
1:05:54 does legal stuff for me and I might have
1:05:55 a dating agent that goes off and does
1:05:58 all that kind of stuff I'm not I'm
1:05:59 married so you know
1:06:01 um so so
1:06:03 I this idea of these agents kind of
1:06:06 taking on these autonomous lives of
1:06:07 their own I think it's going to
1:06:09 transform the game industry but I think
1:06:11 Way Beyond
1:06:12 um just gaming application for these
1:06:14 things
1:06:15 how do I think AI will affect sales
1:06:17 professionals I think every single
1:06:20 profession is going to be upended and
1:06:23 transformed I think sales is a huge one
1:06:25 I so imagine so right now
1:06:29 um using 11 Labs you can synthesize your
1:06:32 voice and it's it it's pretty close to
1:06:34 what you sound like and probably within
1:06:36 a year it will be indistinguishable from
1:06:38 what you sound like
1:06:39 also within a year the speed with which
1:06:42 that agent that uses your voice will be
1:06:46 able to have a real-time conversation
1:06:48 where there's not a noticeable lag
1:06:51 that'll all probably be gone so I think
1:06:53 outbound sales is going to go fully to
1:06:57 fully to automated AI
1:07:00 synthesized Ai and it could sound like
1:07:02 you so all of your outbound sales calls
1:07:04 could you know could go like that I
1:07:07 think how you research
1:07:09 um companies how you personalize emails
1:07:12 I'm already starting to get way more
1:07:14 personalized emails than I've ever
1:07:15 gotten before
1:07:16 um because of AI I think that's going to
1:07:18 get Way Insane I think everything about
1:07:21 it so if you're in sales I would I would
1:07:23 dig deep into
1:07:26 just pop open chat gbt and have it start
1:07:29 doing research for you have it start
1:07:30 writing letters for you
1:07:33 um
1:07:33 learn about the voice synthesis stuff
1:07:36 um just start learning about it you
1:07:37 don't need to do anything right now but
1:07:39 as you learn about it you're going to
1:07:40 realize oh [ __ ]
1:07:43 um if you use notion to take notes it's
1:07:46 already in there if you do video calls
1:07:49 you can take those transcriptions and
1:07:51 automatically have those things
1:07:52 summarized and action items pulled out
1:07:54 of them
1:07:55 using these autonomous agents you're
1:07:58 going to be able to have it
1:07:59 automatically enter stuff into
1:08:00 Salesforce imagine assuming you use
1:08:04 Salesforce if you could just say
1:08:07 add this person into Salesforce
1:08:11 um and and take notes from that last
1:08:12 call and put all the notes in there and
1:08:15 and it will just do it it will navigate
1:08:17 Salesforce for you
1:08:19 put that thing in perfectly I think it's
1:08:21 going to affect sales dramatically I
1:08:23 think it's going to affect almost every
1:08:25 job dramatically
1:08:28 um
1:08:28 [Laughter]
1:08:31 look up three times that we're in the
1:08:33 Bravo hackathon
1:08:36 um I'll go to Bing for that are you
1:08:38 talking to me you talking to me
1:08:41 um let me see Bing
1:08:44 dot com ask me anything
1:08:47 um what were
1:08:50 three teams in the
1:08:54 Bravo
1:08:55 hackathon I don't know what that is
1:08:59 oh that was Department of Air Force wait
1:09:01 let me go to chat
1:09:07 I found information on the Bravo
1:09:08 hackathon series of Air Force and space
1:09:11 force the hackathon allows teams I
1:09:14 couldn't find any information on the
1:09:15 three teams that participated however I
1:09:17 found
1:09:18 that the Department of Air Force will
1:09:20 host a Bravo hackathon to test whether
1:09:22 permissive permissive development
1:09:24 modeling using classified data is more
1:09:26 cost effective than traditional means
1:09:29 and prototyping weapon capabilities with
1:09:31 with a citation
1:09:34 um
1:09:36 the goal of the Bravo hackathon is to be
1:09:39 is sixty percent of participants be from
1:09:42 the government all right that's cool uh
1:09:44 it didn't give it the answer but I know
1:09:46 more about that than I did before
1:09:48 do you think the military used AI before
1:09:51 everyone well listen we've all been
1:09:53 using AI for like the past decade series
1:09:56 Ai and Alexis Ai and your grammar check
1:09:59 and your spell check and your music
1:10:01 recommendations and Netflix all like AI
1:10:04 it's
1:10:05 so so AI isn't new what's new is real
1:10:09 human beings like you and me having
1:10:12 access to these tools you know not being
1:10:14 data scientists or these big massive
1:10:16 corporations
1:10:17 um that's what's new so yes the the
1:10:19 military's been using AI for a while we
1:10:21 just did in in everyday AI
1:10:24 we just did
1:10:27 I think it was issue 15 yeah or 16.
1:10:30 issue 16.
1:10:32 um we talk about this company language
1:10:34 models so like chat gbt they're using
1:10:37 large language models to automatically
1:10:39 come up with Battlefield logistic
1:10:41 recommendations that they then send
1:10:43 those recommendations up to you know
1:10:46 decision making brass who can then take
1:10:48 action on those so
1:10:51 um yeah so so
1:10:53 so yes I think they were using using a
1:10:56 AI long before everyone and they
1:10:59 probably have stuff that is really scary
1:11:02 but you know
1:11:04 whatever
1:11:07 when things get you know sort of
1:11:09 completely out of my ability to control
1:11:10 them I'm just like yeah I don't know I'm
1:11:12 just gonna
1:11:14 whistle whistle past the graveyard
1:11:19 okay let's see other questions I have a
1:11:22 degree in biology how will I impact my
1:11:25 career prospects I think if you learn Ai
1:11:27 and get really good with it
1:11:30 um it could dramatically improve your
1:11:32 prospects if you ignore it it could
1:11:36 decimate your prospects
1:11:39 I mean quite frankly I mean biology the
1:11:42 the the the amount of medical
1:11:44 advancement and Analysis data analysis
1:11:47 that you can do with large language
1:11:49 models I don't know what particular
1:11:52 flavor of biology you're in but boy you
1:11:54 could do some really amazing stuff with
1:11:56 uh with large language models so if
1:11:58 you're not digging in deep to it I would
1:12:00 say digging deep to it and figure out
1:12:02 how you can apply it to your particular
1:12:04 flavor of AI I have a degree in applied
1:12:07 math and built multiple AIS for the
1:12:09 government there you go
1:12:11 um yeah this is my friend Richard Boyd
1:12:14 has been in in Ai and large language
1:12:16 models for a decade or so he sold a
1:12:20 company to Lockheed Market Martin he's
1:12:22 been doing this stuff for a long long
1:12:23 time so they've had they've had it for a
1:12:25 while and and again there's there's
1:12:27 stuff we don't know about that's
1:12:28 probably pretty intense
1:12:30 but these tools that we have access to
1:12:33 are quite powerful like you know Jeffrey
1:12:35 Hinton the guy they're calling the
1:12:36 Godfather of AI
1:12:39 um who just quit Google he quit Google
1:12:41 because he's like we've accomplished in
1:12:44 five years what I thought would take 30
1:12:46 to 50 years
1:12:48 um it's moving very very fast
1:12:50 um nice I'm wanting to learn so bad if
1:12:53 you're wanting to learn so bad just jump
1:12:55 in go to chatgpt chat.openai.com
1:12:59 pick a project for yourself or for your
1:13:02 business or for a community center and
1:13:05 just do a project using chat GPT as your
1:13:07 partner to to execute that thing write a
1:13:10 marketing plan for your business doesn't
1:13:11 matter just go do a project and you will
1:13:14 learn
1:13:15 what doing a project does is it lets you
1:13:18 learn what these things what's what
1:13:20 these things make possible and it kind
1:13:22 of does that and then you can kind of
1:13:25 see other things that's the best way to
1:13:27 learn it
1:13:29 um uh yodkowski says AI will kill us
1:13:32 what do I think and why
1:13:34 um
1:13:35 yeah I mean he's he's a what they call
1:13:38 doomers there's Boomers there's doomers
1:13:39 and then I'm a gen xer [ __ ] all them
1:13:42 Boomers
1:13:43 um
1:13:47 I think it's very easy to go to the dark
1:13:49 place I think it's actually harder and
1:13:51 way more interesting to go to the light
1:13:53 place
1:13:54 um I think that these these tools are
1:13:58 superpowers and they allow us I think
1:14:00 they they have the very real possibility
1:14:03 of elevating us to a higher level
1:14:11 [Music]
1:14:14 I just had something pop up that I don't
1:14:16 quite understand of tick tock when I get
1:14:20 an invite here I think that's someone
1:14:21 asking to come on stage and talk I might
1:14:23 start doing that Lang chain or llama
1:14:26 index
1:14:27 um I don't know I haven't really dug
1:14:30 deeply into Lang chain and I've heard of
1:14:32 llama index but I just all that stuff is
1:14:35 so new right now I mean Lang chain is
1:14:38 starting to be a bit more established so
1:14:41 um
1:14:42 but I don't know I mean I I think
1:14:44 they're both really remarkable
1:14:47 approaches to tools
1:14:50 um I'm just waiting for apple or
1:14:52 Microsoft or someone to make it just
1:14:55 stupid easy to train large language
1:14:57 models on my own data I'm just being
1:15:00 lazy
1:15:03 [Music]
1:15:05 do I think it's too late to jump on the
1:15:06 bandwagon no we are very early we're
1:15:09 very early very very very very very
1:15:10 early
1:15:12 I'm doing I'm putting together a
1:15:14 workshop for a global company right now
1:15:16 and we we sent out a survey to the
1:15:18 people that might attend the workshop
1:15:20 and 55 percent of them had heard of
1:15:23 Chachi PT and 45 percent of them hadn't
1:15:26 there are a lot of people that just
1:15:27 haven't even heard of chat gbt much less
1:15:30 doing anything about it so
1:15:33 um we are very early we're very early
1:15:37 uh uh literally I made a Holodeck bar
1:15:40 full of aliens interacting with one
1:15:42 another that's super cool
1:15:44 um
1:15:45 can AI get so smart that the government
1:15:47 will shut it down we'll shut down the
1:15:49 internet
1:15:50 I would assume
1:15:52 soup a possibility
1:15:55 um
1:16:00 but I it would have things would have to
1:16:03 be pretty
1:16:04 [ __ ] destructive to shut down
1:16:08 essentially the engine of the economy
1:16:12 um
1:16:13 you know that might be if something
1:16:15 happens globally
1:16:17 um that that sort of globally a bunch of
1:16:20 countries decide to shut it down
1:16:21 machine war is worse than nuclear war
1:16:23 yeah I the the the the the idea of
1:16:26 weaponizing robots that have
1:16:29 um autonomous AI in them is is pretty
1:16:31 terrifying
1:16:32 um really AI is for gaming as it has
1:16:35 very limited information I don't know
1:16:38 quite what that means AI talking to AI
1:16:41 on 20 computers he's going to so far
1:16:44 outrun our abilities yeah these we're
1:16:46 we're already pretty close to these
1:16:47 things out running they're at least out
1:16:49 running our capacity for knowledge and
1:16:53 as these autonomous agents get smarter
1:16:56 and the large language models get
1:16:58 smarter yeah
1:17:01 um hey I will destroy the world I think
1:17:03 that's
1:17:04 that it's easy to go there
1:17:08 yeah I think it is easy to go there
1:17:09 because Hollywood's trained us that the
1:17:11 AI the robots are going to kill us
1:17:13 two eyes quantum computers talking to
1:17:15 each other will be able will be unable
1:17:18 for us to control it's going to be
1:17:19 really interesting that the quantum
1:17:20 computer thing is is quite interesting
1:17:23 where the the training of these models
1:17:27 might get might might be happening in
1:17:29 real time and yeah I don't know it's
1:17:32 really it's fascinating stuff
1:17:35 um so it's just regurgitating
1:17:36 information that may not be correct it's
1:17:38 well it's not regurgitating information
1:17:40 it's not copying and pasting chunks of
1:17:42 information it it's kind of like a brain
1:17:45 where it learns lots of words and it
1:17:47 learns their semantic relationships to
1:17:49 one another and then when you ask it a
1:17:51 prompt it it does its best through
1:17:53 probability to answer your question with
1:17:56 with original
1:17:58 composition of words
1:18:01 um but
1:18:02 it does hallucinate that's the term in
1:18:05 in AI meaning it will confidently answer
1:18:08 you even if it's wrong Chachi BT in
1:18:10 particular has been trained to to
1:18:13 give you what you want and so it will
1:18:16 answer you and I I like to call Chachi
1:18:20 PT mansplaining as a service I did not
1:18:22 come up with that phrase but I love it
1:18:25 it will confidently answer you even if
1:18:27 it's wrong so if you're doing anything
1:18:28 that you need real answers for you need
1:18:31 to
1:18:32 um self-verify so just like with nfts
1:18:35 and blockchain do your own research with
1:18:38 chat GPT do your own verification
1:18:41 um
1:18:43 clothes and windows can't help now they
1:18:46 can see through Wells with new newly
1:18:48 developed technology I'm sure they are
1:18:49 yeah yeah
1:18:52 that they're watching we've already
1:18:53 given away all of our privacy anyway we
1:18:55 did that 20 years ago with Facebook and
1:18:57 Google so
1:18:59 um if the product is free you're the
1:19:02 product
1:19:03 then chat GPT will be my personal slave
1:19:05 yes
1:19:06 it's a tool
1:19:08 have it do
1:19:09 the [ __ ] you don't want to do that's one
1:19:11 of the things that I think is this
1:19:13 stuff's going to be brilliant at truth
1:19:14 gbt that'll come
1:19:16 there's companies like anthropic that
1:19:18 are really looking at the the accuracy
1:19:20 part of
1:19:22 um of these large language models
1:19:24 there's going to be a whole industry
1:19:25 around that how can a AI be beneficial
1:19:28 for those of us with ADD I am a
1:19:31 laminated card carrying member of the
1:19:34 ADHD Club
1:19:35 um
1:19:36 this stuff is so wild and so imagination
1:19:39 sparking that it overwhelms my ADD brain
1:19:44 um
1:19:45 I think it does a lot I think I think
1:19:53 one of the one of the super powers of
1:19:56 add is the ability to synthesize lots of
1:19:59 things take lots of inputs and
1:20:00 synthesize them
1:20:02 and you can think fast and you have lots
1:20:04 of ideas where where I have historically
1:20:06 got hung up when I have an idea is I get
1:20:09 hung up on the execution of of taking it
1:20:11 to the next level right so oh I've got
1:20:14 an idea I need a piece of software like
1:20:16 oh I want to make a flow chart oh I
1:20:19 should get a piece of software for
1:20:20 flowchart and then I spend the next 20
1:20:22 minutes trying to find a piece of
1:20:24 software and two of them suck and one of
1:20:25 them makes me pay before I find out it's
1:20:27 the wrong piece of software and 45
1:20:29 minutes later I have no [ __ ] idea
1:20:31 what the idea was right what happens
1:20:33 with AI is that you have an idea and
1:20:37 with chatgpt you can flesh that idea out
1:20:39 within five ten minutes like fully
1:20:42 fleshed out like holy [ __ ] that was a
1:20:45 great idea let me dig deeper on that or
1:20:47 that wasn't all that good an idea but
1:20:49 you know what that gave me another idea
1:20:51 and then you can flush that out and then
1:20:52 you can jump over to Mid journey and
1:20:53 generate an image and so it allows you
1:20:56 to run down your ADHD hyper Focus rabbit
1:20:59 holes in 10 minutes instead of two hours
1:21:02 course and so the the frustration and
1:21:06 and discouragement of when you spent two
1:21:09 hours and you just wasted all your time
1:21:10 trying to find the right piece of
1:21:12 software and you lost the idea that
1:21:14 that's gone for me so I'm finding myself
1:21:17 living at a much more focused much more
1:21:20 high level
1:21:22 um
1:21:22 creative plane than I've ever done
1:21:25 before so
1:21:27 I think it's super powers for everyone
1:21:28 but I I to agree to a great degree I
1:21:31 think AI right now kind of requires a
1:21:35 little manic a you know 80d brain to to
1:21:38 work well
1:21:41 I just lost all my comments again hold
1:21:43 please while I scroll backwards
1:21:47 oh God damn this happened
1:21:53 okay so what you should do now is smash
1:21:55 that screen so I get lots of heart
1:21:57 buttons that someone told me that
1:21:59 yesterday
1:22:02 while I try to find all your comments
1:22:04 smash that screen show me some love
1:22:06 people
1:22:08 all right thank you let me see all the
1:22:11 love coming okay I found the add
1:22:13 comments I hope that answered okay how
1:22:16 much cash have you got in your wallet
1:22:17 none
1:22:20 um in fact when oh I was at a yard sale
1:22:23 yesterday and they didn't take venmo it
1:22:24 was I wanted a two dollar thing I didn't
1:22:26 have two dollars in my pocket I'm like
1:22:28 how do you not take venma like how do
1:22:30 you not
1:22:31 I don't I just don't have cash anymore
1:22:33 cash is weird
1:22:36 how do you create images with chat gbt
1:22:38 well well you don't you don't with chat
1:22:40 EBT although it'll get there
1:22:43 you create images with things like
1:22:46 stable diffusion and mid-journey and
1:22:48 Dolly if you want to just play
1:22:50 um go to bing.com create and that's the
1:22:54 dolly that's that sort of Microsoft's
1:22:56 Bing but they've got the dolly image
1:22:58 generation engine so bing.com create and
1:23:01 then you just type in something and
1:23:02 it'll make you images
1:23:04 and you can experience that I think it's
1:23:06 actually important for you to for people
1:23:07 to do that if they haven't done it just
1:23:09 to experience kind of the magic of you
1:23:11 put in words and then it's visualized
1:23:13 for you
1:23:15 how do you access chat gpt4 for free you
1:23:17 go to bing.com so bing.com
1:23:20 the chat option is gpt4 connected to the
1:23:24 internet that's free you can go to
1:23:26 po.com you can use gpt4 I think for like
1:23:30 10 queries a day for free and then they
1:23:33 charge you if you want more than that
1:23:36 um
1:23:36 you can pay 20 bucks a month which I
1:23:38 know is not free
1:23:40 it's 20 bucks a month but you can you
1:23:42 can go to uh chat.openai.com pay them 20
1:23:46 bucks a month and they give you access
1:23:47 to gpt4 I'm sure there's others
1:23:50 um I don't understand why Google can't
1:23:52 understand the power of AI
1:23:54 the power is AI powered data and they
1:23:57 are the king
1:23:58 well not only does Google understand
1:24:00 they flip and invented the Transformer
1:24:03 which is the tech that sits underneath
1:24:05 all of these image generation tools all
1:24:08 the video generation tools all the audio
1:24:09 generation tools chat EBT that ship was
1:24:13 invented by Google
1:24:15 they purposefully
1:24:21 didn't use it didn't release it because
1:24:25 it threatened their business model their
1:24:27 business model is based on you clicking
1:24:30 20 times to find something
1:24:32 so if you just give someone the answer
1:24:34 it [ __ ] up their business model and
1:24:36 they had 92 or 94 percent market share
1:24:39 when chat GPT launched
1:24:42 so they didn't have any incentive to to
1:24:45 threaten their own business model but
1:24:47 but when chat EBT got to 100 million
1:24:50 users it forced them out of the out of
1:24:53 the woodwork it forced them to confront
1:24:55 this so so that they're behind because
1:24:57 they they they basically ignored it can
1:25:02 chat gbt drive my car no but Tesla's you
1:25:06 know autopilot is trying to get there
1:25:08 it's taking longer than Elon Musk
1:25:11 thought it would how realistic are
1:25:12 expectations that AI is detectable in
1:25:14 text
1:25:16 well some of it is there's just a
1:25:18 there's a war going on right now whether
1:25:20 or not something has been written by
1:25:22 chat GPT will be irrelevant six months
1:25:26 from now maybe a year from now it's
1:25:28 going to be ubiquitous it's going to be
1:25:29 in everything so who gives a [ __ ] if it
1:25:32 was written by chat TPT you know what
1:25:34 you should give a [ __ ] about is is it
1:25:36 good and in fact Google Google initially
1:25:38 said
1:25:39 it's gonna it's gonna deprioritize
1:25:42 anything that was written by large
1:25:43 language models and then after they
1:25:46 released Bard they said well we're not
1:25:48 going to de-prioritize anything written
1:25:50 by a large language model but we're
1:25:51 going to de-prioritize things that are
1:25:53 low value content so that they're like
1:25:55 you got to do high value content which
1:25:57 basically means you need an editor you
1:25:59 need someone who who actually looks at
1:26:02 the content and goes yeah this is good
1:26:03 valuable content that solves this
1:26:05 particular content goal
1:26:07 um and and if it does that whether or
1:26:09 not it was written by a machine who
1:26:11 gives a [ __ ]
1:26:14 I heard that Bing is a mix of GPT 4 and
1:26:17 3.5 is that true I think it's based on
1:26:19 four from what I understand but but it's
1:26:22 definitely a different
1:26:24 um fine-tuned model and they've got
1:26:26 their own kind of set of rules so when
1:26:29 Bing first came out it was quite quite
1:26:30 powerful
1:26:32 um and and like weirdly more powerful
1:26:34 than GPT 3.5 chat GPT at the time
1:26:38 and then this reporter from The New York
1:26:40 Times kind of pushed it over the edge
1:26:42 and it told him to leave his wife and
1:26:43 that made the cover of the New York
1:26:45 Times and when that happened Microsoft
1:26:47 shut that [ __ ] down so they kind of
1:26:50 neutered Bing they neutered Sydney was
1:26:52 Sydney was the name of the
1:26:55 um what what uh what Bing chat told this
1:26:58 reporter its name was was Sydney so they
1:27:01 uh they lobotomized Sydney
1:27:04 um but but it's getting better again
1:27:05 like they're they're slowly opening up
1:27:08 the capability as as they're getting
1:27:10 safe but they kind of got their asses
1:27:12 handed to him for a week or two there
1:27:13 I'm working with AI to create a meal
1:27:15 plan but it keeps forgetting what I've
1:27:17 asked okay
1:27:19 so yeah this is and this will get better
1:27:23 over time but but the the way the memory
1:27:26 works when you have when you have a chat
1:27:28 conversation with chat EBT
1:27:31 there's a window of 4 000 tokens which
1:27:34 basically equates to three thousand
1:27:36 words so you have about three thousand
1:27:38 words that it can keep in its memory so
1:27:41 you start with just a paragraph and
1:27:43 that's whatever 150 words and then it
1:27:45 gives you some answers and that's a
1:27:46 thousand and then you go back and forth
1:27:48 and so there's kind of this rolling
1:27:50 three thousand word window so what you
1:27:53 may need to do is if if there's
1:27:56 something from the beginning of the
1:27:57 conversation go copy that or have it
1:27:59 summarize that and sort of put that down
1:28:02 lower in the conversation and that'll
1:28:04 help
1:28:05 um there's a a a
1:28:09 an open source model that we covered in
1:28:12 everyday AI
1:28:13 from Mosaic ml this model called MPT
1:28:18 7B and one of their models is called
1:28:22 story writer and story writer has a 65
1:28:25 000 token context window so that's 48
1:28:29 000 words so it's basically an entire
1:28:31 novel is what it can keep in its memory
1:28:34 so you can go use that open source model
1:28:36 story writer from MPT 7B Mosaic Mosaic
1:28:40 ml
1:28:42 um or you can just wait for gpt4
1:28:46 um when they finally fully release it
1:28:48 we'll have 32 000 tokens which is like
1:28:50 25 000 words so these things are going
1:28:52 to get better but yeah it's pain in the
1:28:54 ass right now I thought Bing is running
1:28:56 gpt4 since about a month not officially
1:28:58 but I've heard that often yes gp24
1:29:00 that's confirmed
1:29:03 I build a trading bot with chat GPT and
1:29:05 it's wicked accurate I know the trading
1:29:06 bot stuff's really cool
1:29:10 um um yeah stock trading um heard that
1:29:12 auto GP racks up a fortune doing the
1:29:14 most simple things yeah if if Auto GPT
1:29:17 gets in a recursive Loop and just keeps
1:29:19 looping through [ __ ] and it's charging
1:29:20 you API fees every time uh yes it can
1:29:23 and if if one of the things it's doing
1:29:25 is embedding like using its memory
1:29:28 embedding is quite expensive
1:29:30 um relatively so yeah you can run up big
1:29:32 bills with that I haven't but I've heard
1:29:35 that people have run up big bills
1:29:37 just maybe damn it would be great other
1:29:40 than God damn it sorry
1:29:41 I swear I'm not Gary Vee swearing but
1:29:45 when I get excited I tend to Cut Loose
1:29:48 uh
1:29:49 damn it in the presence of the Lord
1:29:53 um I left an agent overnight and it
1:29:55 spent my whole API allowance I know
1:29:58 that's what she could that's that's the
1:30:01 price innovation
1:30:03 uh it's free but neutered paid API for
1:30:06 full yeah
1:30:07 can GPT be used
1:30:11 by a doc yeah so there's a there's a
1:30:14 there's a uh a really cool Tick Tock
1:30:16 video I saw of this Doctor Who told Chet
1:30:20 GPT
1:30:21 about a patient case and it got the
1:30:23 diagnosis right and then gave it a
1:30:25 harder case and it got the diagnosis
1:30:27 right and then it gave it like the
1:30:29 hardest case you know he had ever dealt
1:30:31 with and it got the diagnosis right and
1:30:33 he goes oh my God this is remarkable and
1:30:35 then he said you know give me you know
1:30:36 cite the research
1:30:38 that is the foundation for you know how
1:30:41 you got to that diagnosis and it gave
1:30:43 him these citations and they were all
1:30:45 made up it was always all [ __ ]
1:30:47 um so so but yeah doctors can use it for
1:30:50 in fact there was a there was a a study
1:30:53 that just was published this week where
1:30:57 they had real doctors and Chachi PT
1:31:00 doctors interacting with patients and
1:31:01 the patients by far preferred the chat
1:31:04 gbt doctors and and they were by far
1:31:07 more empathetic than the real doctors
1:31:09 were
1:31:10 um so I think there's lots and lots of
1:31:11 ways doctors are going to be using this
1:31:13 stuff how can I make images in chat EBT
1:31:15 you can't I inadvertently put in the
1:31:18 birthday with a 2095 instead of a 1995
1:31:21 and now it won't let me set up oh that
1:31:24 sucks I know that's that's the problem
1:31:26 with a lot of these new technology
1:31:28 companies is they don't have [ __ ]
1:31:29 tech support because you get it for free
1:31:32 you're the product and they just don't
1:31:35 give a [ __ ] I'm sorry about that it
1:31:37 makes good python code I know that was
1:31:39 my big breakthrough Epiphany about Chad
1:31:42 jpt it wrote me a python application the
1:31:44 problem is that I only it only has data
1:31:46 until 2021. yeah the large language
1:31:48 models they stop at September 2021 but
1:31:50 Bing is connected to the internet so
1:31:53 Bing can combines web searches that are
1:31:56 more recent with the large line huge
1:31:58 model and and they're all going to start
1:31:59 to do that so that limitation won't be
1:32:02 nearly as apparent as things move
1:32:03 forward well AI really replace software
1:32:06 Engineers um some of them some of them
1:32:08 and and
1:32:10 but but I I still think certainly for
1:32:12 the next two years
1:32:14 um you know software Engineers are going
1:32:15 to be super required and if you know
1:32:17 anything about Ai and how to how to
1:32:19 Wrangle these tools there's going to be
1:32:21 a lot of demand for that
1:32:23 um but I I think that that that we are
1:32:25 quickly moving to
1:32:28 um a place where
1:32:31 these tools might dynamically create
1:32:34 software
1:32:35 for for one-time use right like I ask it
1:32:38 to do something and it will say is there
1:32:41 an app that does that no okay let me
1:32:43 just write one it'll write an app it'll
1:32:45 do whatever I ask it to do and then when
1:32:48 I'm done with it it'll just go away so
1:32:50 um so I think the world of software
1:32:52 engineering is going to change
1:32:53 dramatically and if you're if you're not
1:32:56 using copilot on GitHub
1:32:59 um I would start just get get used to
1:33:01 you know sort of think of it like this
1:33:04 your pair programming with a large
1:33:06 language model
1:33:09 um
1:33:11 yeah someone said use Bart if you need
1:33:12 past 2021 seriously
1:33:15 gosh damn it is offensive thanks well
1:33:18 sorry I I would unsubscribe because I do
1:33:20 swear every now and again
1:33:23 um I've decided not to be a software
1:33:25 engineer but because this will be
1:33:27 automated in the future yeah I mean
1:33:30 I you know I still think there's a role
1:33:32 for engineers and programmers I just
1:33:34 think what they do if you compare what
1:33:36 you do now as a software engineer to
1:33:37 what you're going to do it's going to be
1:33:39 very different and I will take over the
1:33:41 90 of the jobs in the next five years
1:33:45 90s probably
1:33:47 hyperbolic but you know I could see 30
1:33:50 to 50 percent
1:33:52 but it will also generate jobs that we
1:33:54 have no clue are coming so that there's
1:33:57 going to be new new categories of work
1:34:00 that are coming
1:34:01 that are probably going to be remarkable
1:34:02 and interesting so so it's not just Doom
1:34:04 and Gloom again I I there's plenty of
1:34:08 ways to look at this and just look at
1:34:09 the the dark side look at the light as
1:34:12 well
1:34:13 I I mean think about this think about
1:34:15 what what these AI tools are is they are
1:34:19 Brilliance amplifiers they take what
1:34:21 you're brilliant at and they amplify it
1:34:24 so if you were
1:34:26 50 percent more brilliant than you are
1:34:28 because you're using these tools what
1:34:31 could you contribute
1:34:32 what could you do at your job what could
1:34:34 you do in your community what could you
1:34:35 do for your family
1:34:37 that's what these things make possible
1:34:38 so don't just go to the dark side
1:34:41 I've been using AI for a while now and
1:34:43 most people have no clue what's coming
1:34:44 that's why I'm doing this channel this
1:34:46 is absolutely about that why are we so
1:34:48 enthralled
1:34:49 with AI when the Terminator movie
1:34:52 predicts it will be our demise
1:34:54 because that's a movie
1:34:58 I don't know
1:35:00 what if
1:35:01 things got dramatically better instead
1:35:04 of dramatically worse I mean
1:35:06 we're screwing up this
1:35:09 globe and Country right now without AI
1:35:14 so what if things got better
1:35:17 why are we selling okay that was that
1:35:19 one in the future only the only jobs
1:35:21 will be personality based I think
1:35:23 authenticity human connection empathy
1:35:27 creative expression
1:35:30 um are all going to be in high demand in
1:35:32 the future I think I think and this is
1:35:34 one of the things I'm excited about if
1:35:36 you if you think about the mundane
1:35:39 low-level repetitive soul-crushing work
1:35:43 in our society a lot of that is going to
1:35:47 be automated out and then what's left I
1:35:50 don't know go start a music career or
1:35:51 not even a music career go teach
1:35:54 yourself songwriting and sing to your
1:35:56 friends
1:35:57 and reconnect with the world the AI rap
1:36:00 song My Gosh the the one with Drake in
1:36:02 the weekend yeah amazing
1:36:05 amazing and then uh Grimes uh elon's ex
1:36:09 uh just said last week
1:36:11 um synthesize my voice all you want just
1:36:13 give me 50 which I think is a brilliant
1:36:15 move
1:36:16 how soon are you thinking I don't know
1:36:17 in response to what three months and you
1:36:19 will
1:36:20 hear more and more Yep Detroit Detroit
1:36:23 become human
1:36:26 is this what our future will be I don't
1:36:28 quite know what that means like will
1:36:30 cars become human will robots become
1:36:32 human
1:36:34 um they're going to be a hell of a lot
1:36:35 more human than they are now so Steve
1:36:37 from the IRS will no longer sound like
1:36:39 he's calling from Calcutta correct I
1:36:41 that's actually one of the things that I
1:36:43 think is really interesting is is um
1:36:46 you know you're gonna you're gonna have
1:36:47 voice synthesis things where you know
1:36:50 Steve whose name is obviously not Steve
1:36:52 will actually sound like Steve
1:36:55 um but
1:36:56 it will also probably not be someone in
1:36:59 Calcutta on the other end of the phone
1:37:00 it'll probably just be purely
1:37:02 synthesized voice and you're having a
1:37:04 conversation with with a chat GPT bot
1:37:06 that's been trained for that particular
1:37:08 business
1:37:10 chat GPT gave me a business plan
1:37:11 congrats that's awesome so if AI can
1:37:14 replace customer service what if the
1:37:16 customer contact also works with AI yeah
1:37:19 well I think that I think we're
1:37:22 absolutely moving to a world where it's
1:37:24 going to be like I'll have my AI reach
1:37:26 out to your AI and they'll work it out
1:37:28 and write whoever's got the best AI sort
1:37:30 of comes out on the top of that
1:37:31 transaction I think this will happen
1:37:33 with things like insurance claims
1:37:36 um legal interactions customer service
1:37:39 interactions all that sort of stuff yeah
1:37:43 if you haven't played with it there's a
1:37:45 there's a tool called
1:37:47 um
1:37:48 do not disturb
1:37:55 let's see do not disturb
1:37:59 where is it is it.com
1:38:02 do not
1:38:04 disturb is it do not disturb is that
1:38:06 right
1:38:08 disturbed.com no oh dang it
1:38:13 I didn't say
1:38:15 the GD you'd be proud of me those of you
1:38:18 that
1:38:19 hate the Lord and damn being in the same
1:38:22 word together
1:38:23 um do not disturb
1:38:26 foreign
1:38:36 no do not what the hell's the name of
1:38:39 that
1:38:42 there's a your words the world's first
1:38:44 virtual lawyer that's maybe how you find
1:38:45 out world's first virtual lawyer
1:38:55 world's first robot lawyer
1:38:58 this is it
1:38:59 sorry about this sometimes the brain
1:39:01 just don't work
1:39:07 [Music]
1:39:14 fight speeding tickets this is it
1:39:17 an exit called
1:39:19 do not pay do not disturb I'm such an
1:39:22 idiot
1:39:23 do not pay so if you this is
1:39:27 um a virtual lawyer that will like fight
1:39:28 your speeding tickets for you and and so
1:39:32 so you kind of give it a speeding ticket
1:39:34 or you give it whatever sort of minor
1:39:36 infraction you have and it will go do
1:39:37 all the legal work it does it for all
1:39:39 the municipalities things like that this
1:39:41 was started by a 19 year old
1:39:44 like in his mom's basement I think it's
1:39:46 five or six years old now
1:39:47 um but like this is going to
1:39:51 um
1:39:52 you know impact
1:39:53 lawyers and so and but this is exactly
1:39:56 what you're asking there do you also
1:39:58 believe it will replace most jobs within
1:40:00 less than a decade I mean replace jobs I
1:40:03 think it will transform most jobs within
1:40:05 a decade
1:40:07 um I think there will be a lot of new
1:40:09 jobs that we can't even imagine what
1:40:11 they are right now that'll probably be
1:40:13 really exciting and really interesting
1:40:16 um I think there will be a significant
1:40:19 chunk of people displaced that you know
1:40:22 don't have the capacity or the desire to
1:40:24 learn this stuff that are going to be in
1:40:26 trouble I didn't understand Andrew yang
1:40:30 going on about
1:40:32 um Universal basic income in the in the
1:40:35 last presidential election I find like I
1:40:37 get it now
1:40:38 um I don't think it's the complete
1:40:40 solution but I I think that there's
1:40:42 going to be enough disruption that we we
1:40:45 probably need some sort of backfill to
1:40:47 not screw up the economy now I'm afraid
1:40:50 to swear because people are offended
1:40:52 what you should be offended at is my
1:40:55 hair
1:40:56 okay
1:40:58 um can you get current information or
1:41:00 statistics using chat EBT no but you can
1:41:02 with Bing go to bing.com it's connected
1:41:05 to the internet and with the chat gbt
1:41:07 browser plugin you'll be able to just
1:41:09 not right now
1:41:10 what is chatgpt and how do you get it
1:41:12 chat EBT is
1:41:14 uh a simple front end on on top of in
1:41:19 front of an incredibly powerful
1:41:22 knowledge engine that sits underneath it
1:41:25 you go to chat.openai.com
1:41:28 and
1:41:29 um
1:41:31 start asking it stuff and it will blow
1:41:33 your mind ask it to explain
1:41:37 quantum mechanics to you in the style of
1:41:40 an Eminem wrap and watch what it does
1:41:45 generative models in general along with
1:41:47 classifier super powerful in BIO cool
1:41:50 got a link I don't know what you're
1:41:52 talking about how to
1:41:53 how do data scientists equal AI
1:41:57 don't know what you mean
1:41:59 using chat GPT is My Gym trainer to get
1:42:02 fit yeah my buddy that writes the uh
1:42:04 everyday AI newsletter he just got he
1:42:06 wrote a tweet that just totally blew up
1:42:09 he's lost I think 45 pounds chat gpts
1:42:13 helped him lose 45 pounds since covid
1:42:15 Gen X all the way Gen X baby
1:42:17 we're bitter we're angry
1:42:21 we're the ignored bitter people we're
1:42:24 cynical but we're fun
1:42:26 35 more production 20 more enjoyment
1:42:29 using Chachi because yeah they just did
1:42:31 a study
1:42:33 they had two groups of people one used
1:42:35 chat EPT one didn't
1:42:38 um they gave them identical tasks the
1:42:39 Chachi PT group was 37 percent more
1:42:42 productive at a 20 higher quality so so
1:42:46 think in a five day work week
1:42:48 two of the days you don't need to work
1:42:50 if you're using chat EBT and you're
1:42:52 outputting 20 higher quality work than
1:42:55 the people not using it
1:42:57 like and and and that's using like you
1:43:00 got to realize that that the tools that
1:43:02 we have access to to right now are like
1:43:04 these janky primitive
1:43:08 quickly tossed together like chat GPT
1:43:10 was written in two weeks I'm sure it's
1:43:12 better than it used to be but it was
1:43:14 written in two weeks it's a very thin
1:43:16 piece of software there's not much to it
1:43:19 um and
1:43:21 and we're we're seeing essentially 40
1:43:26 productivity improvements
1:43:29 like just wait till these things get
1:43:31 good
1:43:34 um
1:43:36 how do you get or use chat GPT
1:43:38 chat.openai.com
1:43:41 in fact I'll I since there's probably
1:43:43 some new people here
1:43:45 I play with if
1:43:50 sorry my app just crashed
1:43:53 um tap the screen if you can still hear
1:43:55 me am I still live
1:43:57 okay
1:44:00 um
1:44:02 yes yes okay good
1:44:04 um
1:44:06 so chat.openai.com sorry about the the
1:44:10 um
1:44:11 the focus thing God damn I hate machines
1:44:14 sometimes oops I just said goddamn oh
1:44:16 well what are you gonna do
1:44:19 um you could not watch
1:44:22 um jeez that's not readable at all is it
1:44:26 so anyway chat.openai.com bing.com or
1:44:30 poe.com poe.com the the focus is just
1:44:33 disaster
1:44:35 um so that's how you can use chat GPT
1:44:39 let me go get another picture
1:44:43 let's see
1:44:52 other signal I like that one
1:44:54 these images are from a project called
1:44:57 Kyle Shannon dreams so these were all
1:44:59 trained up on like 20 pictures I
1:45:01 uploaded of myself to this thing called
1:45:03 dream booth and they were made in stable
1:45:05 diffusion so they're sort of me but not
1:45:08 me
1:45:09 kind of pictures
1:45:11 um
1:45:12 fine-tune your digital twin
1:45:18 any guidance on how to use it for
1:45:20 language learning yeah it knows all the
1:45:22 languages I had I had chat gbt I took a
1:45:25 paragraph I translated it to
1:45:28 Portuguese then Hungarian then Japanese
1:45:32 then German then back to English and I
1:45:34 can still read like the paragraph still
1:45:37 made sense it knows languages in fact I
1:45:40 I was I was on in one of my videos I
1:45:42 said to someone they asked a question
1:45:43 about
1:45:45 um
1:45:47 like does it understand languages and I
1:45:49 said it can probably make up a language
1:45:50 and a guy got back to me like a week
1:45:52 later and he said I had it make it make
1:45:54 up a language for me that was sort of
1:45:56 based on some Star Wars something or
1:45:58 other
1:46:00 um so so
1:46:03 okay so if you want to use it for
1:46:06 language learning here's I'll give you
1:46:08 this is good for everyone go to
1:46:10 prompts.chat
1:46:15 and then
1:46:17 it this is called awesome chat GPT
1:46:19 prompts and so the the first part of
1:46:21 this document explains how to prompt but
1:46:23 then it's it's it's all just a series of
1:46:25 act as a so the second one in this list
1:46:30 is act as as an English translator and
1:46:33 improver so you can tell Chachi PT to
1:46:36 act like a
1:46:38 um language teacher and and you could
1:46:40 say a language teacher that specializes
1:46:42 in you know teaching someone who only
1:46:45 knows one language to learn this
1:46:47 specific language and I want you to
1:46:49 create customized training for me and it
1:46:52 will just do it it'll
1:46:54 it'll become your personalized language
1:46:56 tutor
1:46:59 even AGI at human level is going to be a
1:47:02 thousand times faster yeah I know is
1:47:04 this a real live no this is gpt5 and
1:47:08 they have chosen this face uh as the one
1:47:13 that that you would believe is a real
1:47:15 human being but this is all synthesized
1:47:18 you on Twitch two or something I'm not
1:47:21 on Twitch because I'm not a gamer and I
1:47:24 I could probably do things on Twitch
1:47:25 that are not game related but I just
1:47:27 haven't I just started on Tick Tock
1:47:29 because I did my Kyle Shannon dreams
1:47:32 project I created tick tocks for that
1:47:34 and then at some point I was like oh
1:47:35 screw it let me just
1:47:37 make a tick tock to talk about this AI
1:47:39 stuff because I I really started this
1:47:41 channel in in large part to just be able
1:47:44 to process what I was learning to get it
1:47:46 out of my head and then I realized oh
1:47:49 wait this is actually it's actually
1:47:50 really important to demystify this stuff
1:47:52 that that we're in the middle of a
1:47:55 transformation and it's it's it's
1:47:57 I think it's vitally important that we
1:48:00 get curious about this stuff and don't
1:48:02 demonize it and understand it and start
1:48:06 learning it and so that's what this
1:48:07 channel is all about
1:48:08 if you could only be part of one
1:48:10 ecosystem sagemaker hugging fish or
1:48:13 steamship
1:48:14 Etc what would it be
1:48:17 that's a good question
1:48:19 um I don't think it would be hugging
1:48:21 face or GitHub or things like that I
1:48:24 haven't even heard of the other two
1:48:25 things
1:48:28 um
1:48:31 if I had my brothers the the large
1:48:34 language models would be just ubiquitous
1:48:36 in every ecosystem that I'm in
1:48:39 and
1:48:41 my personal autonomous agents
1:48:44 um could navigate between the systems
1:48:46 for me
1:48:48 um I would probably prioritize
1:48:52 human communities like like communities
1:48:55 of connection rather than communities of
1:48:57 technical
1:49:00 um
1:49:02 homogenization
1:49:05 because I think the AI is going to just
1:49:07 automate all of the translation between
1:49:09 all of the different platforms so I
1:49:10 don't I don't think the platforms
1:49:12 ultimately matter which which again goes
1:49:14 back to the thing that I think is a
1:49:16 really bright silver lining of all this
1:49:18 stuff is there's the chance that we
1:49:20 reconnect as humans that's the thing I
1:49:22 think we're missing and it's the thing I
1:49:23 think we need
1:49:24 and I think it's the reason I think
1:49:26 everyone's so sad and cranky and angry
1:49:28 with each other right now
1:49:31 um so I want to get back to like not
1:49:33 hating each other wouldn't that be nice
1:49:35 um I don't notice any questions on here
1:49:37 he's replying to because I'm basically
1:49:40 starting at the top and then just
1:49:41 working my way down I'm not doing it
1:49:43 from the bottom up so there's probably a
1:49:45 better way to do this but oh by the way
1:49:47 today's my birthday I'm 58.
1:49:49 um I'm a 58 year old dude so I'm I'm I'm
1:49:52 not a tick tocker I won't be twerking
1:49:54 tonight you're welcome
1:49:57 um
1:49:58 so I'm probably doing this wrong is why
1:50:00 it doesn't look like I'm answering any
1:50:02 questions but I am
1:50:04 um you can use good prompts so that
1:50:06 it won't answer when it's not confident
1:50:10 um okay even can give you a confidence
1:50:13 level I don't quite know what that means
1:50:14 you can use good prompts so that it
1:50:16 won't answer when it's not confident oh
1:50:17 yeah you can you can tell it
1:50:21 um to to
1:50:23 um
1:50:24 one of the one of the hacks of chat GPT
1:50:27 is you can have it give you an answer
1:50:28 and then you can tell it reflect on the
1:50:32 answer you just gave me and and tell me
1:50:35 whether it's good or bad and in
1:50:37 reflecting in the answer and analyzing
1:50:39 its own answer it will get better at
1:50:40 answering your your stuff so so yeah
1:50:43 that's a that's a really good point you
1:50:44 can you can even give it a confidence
1:50:46 level so
1:50:47 so when it gives you an answer tell it
1:50:49 to analyze its answer to give you a
1:50:51 confidence level of how accurate it
1:50:53 thinks it is
1:50:55 um it really interesting I don't see
1:50:57 comments he's talking to that's because
1:50:58 I'm gpt5 I'm not real this is all
1:51:01 synthesized isn't AI cool
1:51:08 what other Tech you think has promised
1:51:11 for revolutionizing the future like
1:51:13 crypto or Quantum computing
1:51:17 um there's a lot there's there's a bunch
1:51:21 um
1:51:21 Quantum Computing feels like it's still
1:51:24 fairly far away but when it gets here
1:51:26 it's gonna be
1:51:28 crazy
1:51:30 um I think all the web 3 stuff is has
1:51:33 shot itself in the foot by being a bunch
1:51:35 of
1:51:36 you know money grabbing crypto Bros
1:51:39 um so I think I think that whole
1:51:41 industry has probably three years of
1:51:45 um eating some Humble Pie until they
1:51:48 actually start putting out real
1:51:50 businesses that that are not just you
1:51:52 know money grabbing scams
1:51:55 um and and I was pretty deep in that
1:51:56 world I'm I'm part of the Neo Tokyo
1:51:59 community and I was pretty deep in that
1:52:00 world for a year and it's just I don't
1:52:03 know what was interesting about crypto
1:52:05 and blockchain and all that nfts for me
1:52:08 was the the deeper I dug the less real
1:52:10 those companies felt and
1:52:13 and the text really interesting just the
1:52:15 companies suck and um the deeper I dug
1:52:18 with the AI stuff the more profound it
1:52:20 got and so I think that AI will drive
1:52:23 web3 adoption
1:52:26 um I'm trying to think if I I mean I I
1:52:28 think AI is going to amplify a lot of
1:52:30 things I think some of the things going
1:52:32 on in in the medical community with
1:52:34 crispr and you know the protein folding
1:52:38 the AI that that you know predicted 400
1:52:42 000 or some ridiculous amount of
1:52:44 proteins like in a two-week period that
1:52:46 that would have taken I don't know some
1:52:48 god-awful number of years
1:52:50 um I think that's going to revolutionize
1:52:51 a lot of things
1:52:53 will this put va's out of business I
1:52:57 don't know what va's are like Veterans
1:52:59 Administration
1:53:01 um
1:53:03 if if if that's what you mean va's of
1:53:07 virtual assistants uh yes I think that I
1:53:11 think that a autonomous agent we're not
1:53:13 when autonomous agents get good
1:53:16 um
1:53:18 you they will be way better than any
1:53:21 virtual assistant
1:53:23 way better they'll go book stuff for you
1:53:28 they'll understand your preferences
1:53:29 they'll fight fights for you they'll
1:53:32 arrange play dates for your kids yes yes
1:53:36 I think that I think that autonomous age
1:53:39 and again I
1:53:40 I just have this sneaking suspicion that
1:53:42 Apple's going to be really good at this
1:53:44 I think I think Apple's gonna gonna be
1:53:47 the first one to really crack the nut of
1:53:48 kind of the consumer
1:53:52 you know the consumer version of AI
1:53:56 and I think it'll be an autonomous agent
1:53:57 that that operates on your behalf and
1:53:59 just does great stuff for you
1:54:04 um
1:54:05 uh can the AI hurt people
1:54:07 um yeah it it can I I mean it like in a
1:54:12 lot of ways but it but it's like I mean
1:54:14 so can everything else right so can
1:54:17 you know a hacker using you know fishing
1:54:21 techniques and
1:54:23 um the AI will be able to do that stuff
1:54:25 a lot more efficiently
1:54:28 um I think we're gonna see
1:54:32 and I don't think it's so this is an
1:54:34 interesting thing like a lot of the fear
1:54:36 is like the robots will get us I don't
1:54:38 think it's so much that the AI is going
1:54:40 to hurt people I think what it is is the
1:54:43 scammers are going to use AI to get
1:54:46 really good at scamming and that's going
1:54:48 to hurt a lot of people
1:54:49 so I think people are going to hurt
1:54:51 people using AI
1:54:53 well AI allows allow us to own AI people
1:54:57 and families who have no idea they're
1:55:00 being watched from a screen I don't know
1:55:02 what that means like you're watching me
1:55:04 right now
1:55:05 I don't know
1:55:08 I'd like to learn how to write AI
1:55:11 um have Chachi PT teach you that because
1:55:13 it can
1:55:15 will AI allow us to own AI people and
1:55:17 families who have no idea I got that one
1:55:20 again I don't I don't know what that
1:55:21 question means if you could have chat
1:55:23 apt rephrase that question so I can
1:55:26 understand it I'm just a simple Country
1:55:29 Doctor
1:55:30 um
1:55:31 they even know the name of our first pet
1:55:34 and the first time we kissed yeah we
1:55:36 gave up all of our private information
1:55:37 20 years ago we give it up to Google and
1:55:40 Facebook so don't blame my AI for that
1:55:42 crap we did that we did that
1:55:45 thank you I'm getting happy birthdays
1:55:47 I'm it jumped me to the bottom again
1:55:49 thanks for the happy birthdays
1:55:52 um I'm old I'm an old Feller
1:55:54 [Laughter]
1:55:56 your controlled cursing is admirable
1:55:58 thank you thank you Emily's wife I know
1:56:00 I I uh I slipped there once but I'm you
1:56:03 know I'm trying I just I get wrapped up
1:56:05 in my I get wrapped up in the the
1:56:07 passion the excitement
1:56:09 and uh
1:56:11 you know I'm a fan of of comics that use
1:56:15 bad words so it's just become my default
1:56:20 so what are you gonna do all right uh
1:56:23 I'm not going all the way back up
1:56:24 because I can't find it but I'll just
1:56:26 start somewhere so
1:56:28 um let's see
1:56:31 I'm a mother shows wait I am mother
1:56:34 shows how AI affects us AI
1:56:38 does depopulate people and then
1:56:40 re-populate us via embryos I have no
1:56:43 idea what that means is there a charge
1:56:45 for AI
1:56:46 um
1:56:48 someone's paying for it
1:56:50 the Bing bing.com is free that's by
1:56:53 Microsoft Microsoft is making their
1:56:56 money on cloud services and Office 365
1:56:59 and probably a little bit of advertising
1:57:01 but not much because they're willing to
1:57:03 lose money on this so they can kick
1:57:05 Google's ass
1:57:07 um
1:57:08 chat GPT is free but you get kind of a
1:57:10 slower version than if you pay for it
1:57:14 um
1:57:15 the dolly
1:57:18 d-a-l-l-e image generation is free you
1:57:21 can also use that on bing.com bing.com
1:57:26 um so there's there's some free stuff
1:57:27 out there
1:57:29 um
1:57:32 what if AI doesn't need us anymore in
1:57:35 this case AGI or ASI I don't know I mean
1:57:38 that's sort of the 20 or 30 year
1:57:41 20 or 30 year Vision I can't I can't
1:57:44 look that far down right now I'm just
1:57:45 trying to figure out what's happening in
1:57:46 the next year because I think the
1:57:49 you know when these things get smart
1:57:51 enough that they're they're they're
1:57:53 starting to create their own governments
1:57:55 and their own
1:57:56 you know governance things and they're
1:57:58 starting to manipulate us into believing
1:58:00 different things
1:58:02 um hopefully we'll be smart enough to
1:58:03 have controls in place if if you're
1:58:06 interested in
1:58:07 a possible direction that those controls
1:58:09 might take start reading up on the
1:58:12 company anthropic which is a bunch of EX
1:58:14 open AI people that use this thing
1:58:16 called the Constitutional learning model
1:58:18 where where basically the AI
1:58:20 measures all of its answers against its
1:58:23 own Constitution so it's a set of rules
1:58:25 that that it checks its answers against
1:58:26 so so there's some models out there that
1:58:29 are pretty pretty interesting
1:58:31 um for that
1:58:32 for me right now I think the next two
1:58:35 years are going to be so disruptive that
1:58:38 um I'm just worried about that right now
1:58:40 I'm I'm kind of in this place of start
1:58:42 getting curious about this stuff so
1:58:44 you're surfing the tsunami rather than
1:58:46 thinking you're in a cozy beach cottage
1:58:49 with with the tsunami coming
1:58:53 um I don't think all the AI tools to
1:58:55 develop software are not in are not
1:58:57 intelligent enough go start writing
1:58:59 software with them they're brilliant and
1:59:01 they're and and they are as
1:59:05 bad now as they will ever be these
1:59:07 things are getting fast better
1:59:09 dramatically fast
1:59:11 the code stuff is I don't think we're I
1:59:14 don't think we're gonna need to
1:59:16 code will become quite invisible
1:59:20 um pretty quickly I think
1:59:23 um what's the largest effect you think
1:59:24 chat GPT will have on business well I
1:59:26 think I think
1:59:32 I think the fact that 100 million people
1:59:34 started using chat GPT in six weeks says
1:59:36 to me that it isn't going away it's
1:59:38 forced all of the AI companies out of
1:59:41 the Shadows
1:59:43 it's dramatically Amplified competition
1:59:46 it's dramatically Amplified the amount
1:59:48 of developers improving these tools
1:59:53 so I don't think that there's a single
1:59:55 business that won't be profoundly
1:59:57 affected
1:59:58 in the next 12 months I think it's going
2:00:01 to have dramatic impact on all
2:00:03 businesses and I mean like
2:00:05 like what companies do may change what
2:00:09 their workers do may change
2:00:11 how they bill may change like like just
2:00:14 take
2:00:15 like one of the worlds I know a decent
2:00:16 amount about is the advertising world
2:00:19 and they use a time and materials model
2:00:22 so they charge hours for people to
2:00:25 generate stuff copywriting images ads TV
2:00:29 shows whatever it is
2:00:33 we're moving into a world where the
2:00:36 generation of that kind of content
2:00:37 happens in three seconds
2:00:40 so if you say to a client I'm going to
2:00:42 charge you
2:00:43 four weeks of 10 people
2:00:47 to generate this amount of content that
2:00:49 I as the client know can be generated in
2:00:52 two days at the same level of quality
2:00:54 you could certainly argue it might not
2:00:56 be the same level of quality yeah but
2:00:58 it's it's orders of magnitude different
2:01:01 um so I think the the advertising
2:01:04 holding companies and and you know just
2:01:07 just ad agencies in general are very
2:01:10 quickly going to confront a reality
2:01:12 where clients are no longer willing to
2:01:14 pay time and materials if you've got a
2:01:16 four decade old holding company that its
2:01:20 entire model is based on
2:01:22 hiring as many people as possible and
2:01:25 charging a hundred and twenty dollars an
2:01:26 hour and up for those employees to do
2:01:29 all this work and all of a sudden
2:01:33 the majority of your clients say I'm not
2:01:35 going to pay time and materials anymore
2:01:36 that will affect that will affect that
2:01:39 business in a very dramatic way I think
2:01:41 that's a very real possibility I think
2:01:43 we're going to see that across almost
2:01:45 every industry
2:01:47 the technology is far beyond what people
2:01:49 are talking about I agree
2:01:51 customer service claims do not pay yeah
2:01:53 do not pay.com oh that was it that was
2:01:55 when I was trying to figure it out what
2:01:57 did I say do not disturb someone should
2:01:59 start do not disturb.com right now
2:02:01 because you know you'll make a fortune
2:02:02 on it because I can remember that
2:02:05 um at least that'll shut down the
2:02:06 Calcutta call centers or free up those
2:02:09 guys for some other Mischief yeah
2:02:10 exactly they'll they'll uh get really
2:02:13 good at the scamming stuff
2:02:14 um I don't take much stock
2:02:16 okay uh I don't know what this says
2:02:20 what is the role of an AI expert called
2:02:24 um
2:02:25 uh
2:02:27 liar
2:02:28 there's there should be no AI experts
2:02:30 right now well there can be machine
2:02:33 learning and data scientist AI experts
2:02:36 but with this generative AI stuff no
2:02:37 one's an expert right now everyone's
2:02:39 trying to figure it out what about trade
2:02:41 box and copyright well my business idea
2:02:42 truly be mine
2:02:45 um
2:02:47 trademark and copyright is an incredibly
2:02:49 complicated nuanced
2:02:52 filled with gray area kind of
2:02:55 topic
2:02:56 um that I am in no way
2:02:58 um uh capable of commenting on with with
2:03:00 any real intelligence I think the
2:03:03 copyright and trademark Mark lawyers are
2:03:06 going to make a fortune in the next two
2:03:07 decades I think everyone's gonna Sue
2:03:10 everyone and I think we are moving into
2:03:13 a world where the role of copyright is
2:03:16 increasingly irrelevant
2:03:19 because we're moving so so copyright was
2:03:21 based on the premise that printing a
2:03:25 book was hard and and those laws got
2:03:29 really weird and had to be modified when
2:03:31 we moved into the internet age when
2:03:32 everything went digital and it was
2:03:34 really easy to send anything and like
2:03:36 they kind of had to artificially put on
2:03:38 sort of copyright layers on top of stuff
2:03:41 we're moving into a world where
2:03:46 content is going to be generated in real
2:03:48 time for me personally like within five
2:03:51 years so I I don't know how you protect
2:03:54 anything in a world where you have when
2:03:56 you where you have infinite content
2:03:59 I I don't know I don't know
2:04:04 people are going to fight and and
2:04:05 people's careers are going to be upended
2:04:07 and that part really sucks
2:04:09 but I just don't know how that concept
2:04:13 survives in a world where you have
2:04:16 infinite instant generation of content
2:04:19 there will be a lot of AI related jobs
2:04:21 lots lots that we don't even know drones
2:04:23 payload drones payload operators will
2:04:27 soon be gone yeah I think drones are
2:04:29 going to get smarter and smarter and
2:04:31 smarter how can I convince my workplace
2:04:33 not
2:04:35 not to go into AI because I don't want
2:04:38 to lose my job your workplace is if
2:04:42 they're going into AI it's smart if any
2:04:45 job that is ignoring this right now I
2:04:47 think is it is at risk of going out of
2:04:50 business
2:04:52 so what I would say to you is get
2:04:54 curious about AI figure out how to
2:04:55 amplify what you do with AI
2:04:58 um
2:04:59 because I don't I don't think you can
2:05:01 hide from this I really don't
2:05:04 um Talk how you talk that's why you're
2:05:06 here we like you as your natural self
2:05:08 why thank you Jasmine I appreciate that
2:05:12 um does it cost money what AI chat GPT
2:05:14 doesn't uh is chat gbt cloud-based yes
2:05:17 it is and and by the way chat GPT if you
2:05:20 don't know it they now have data
2:05:21 controls you can turn off chat history
2:05:22 and you can turn off the ability for
2:05:25 them to use your data but you lose your
2:05:26 chat history so if you turn that on make
2:05:28 sure you
2:05:29 um download your chat history first
2:05:33 what way chat GPT can help with weight
2:05:35 loss um tell it to give you tell it to
2:05:37 make you a meal plan
2:05:39 um if you like a particular
2:05:41 um kind of approach like keto or or
2:05:46 um I don't know whatever some some sort
2:05:47 of approach tell us to give you a meal
2:05:49 plan tell it to give you an exercise
2:05:51 plan I obviously haven't done that with
2:05:53 it that was a fat joke
2:05:56 people are worrying about you swearing
2:05:58 or missing
2:05:59 the point of this stream amazing info
2:06:02 being shared thank you
2:06:05 ah it's a party
2:06:08 how do you use Chet GPT online plugin
2:06:14 uh
2:06:16 I don't know what you mean
2:06:18 there are a lot of uh
2:06:22 like Chrome plugins Chrome extensions
2:06:24 that
2:06:26 tie into the GPT API
2:06:29 um again most of what I use is just I go
2:06:31 to chat.openai.com that's the official
2:06:34 chat GPT site I don't think you need to
2:06:36 use other stuff right now
2:06:38 I need to convince my boss not to get
2:06:40 involved with AI please guide me SOS I
2:06:42 disagree I think you need to
2:06:44 find out I would ask your boss why your
2:06:47 boss is so excited about AI oh I'm in
2:06:49 the advertising industry you're you need
2:06:52 to be an AI like here's here's the deal
2:06:55 a three to four person company
2:06:59 three to four people are going to start
2:07:01 an AI Centric agency that are going to
2:07:04 outperform
2:07:06 a 30 or 40 person agency they're going
2:07:08 to outperform them on turnaround time
2:07:10 cost and quality all three of those
2:07:14 and they're going to start small
2:07:16 companies are going to start taking down
2:07:17 big companies one after the other I I
2:07:20 think if you're in the advertising it
2:07:21 looks like you're a producer
2:07:24 um you got to start learning these tools
2:07:25 the advertising industry in particular I
2:07:28 I do a lot of work in the Pharma
2:07:30 business and
2:07:32 um a number of my clients are fed up
2:07:34 with agencies they're fed up with time
2:07:36 and materials they're fed up with
2:07:38 baiting and switching where the senior
2:07:41 people selling the business and then
2:07:43 they switch them out with the junior
2:07:44 people that are incompetent and
2:07:46 unmotivated
2:07:47 and they charge a ridiculous amount of
2:07:50 money for it they take forever to turn
2:07:51 things around and
2:07:53 it's it's it's a model that is fraught
2:07:57 with with frustration and problems and
2:08:00 as clients get more educated on what
2:08:02 these AI tools can do they're going to
2:08:05 look to agencies that can turn things
2:08:07 around on their expectations at a price
2:08:11 you know that is not the current model
2:08:13 so I think your boss is right I hate to
2:08:17 say it your boss is right to uh to be
2:08:20 leaning into the direction if I were you
2:08:22 I would I would
2:08:24 pile up with your boss and find out why
2:08:27 he or she is so excited about this or
2:08:30 feels it's so important and support them
2:08:34 um because otherwise the the whole
2:08:35 agency is going to go down how do you
2:08:37 use chat GPT online plugin I don't know
2:08:39 I asked that before
2:08:42 um Can AI maintain my home lab
2:08:45 and is smart probably not but still
2:08:47 asking I don't know what your home lab
2:08:50 is like if it's
2:08:53 just like beakers and you're pouring
2:08:55 things into beakers probably not but if
2:08:59 it's anything connected to the internet
2:09:01 um I think it could certainly help
2:09:04 I would like that's something that I
2:09:06 would ask chat GPT I would say you're a
2:09:09 lab assistant
2:09:11 describe your lab describe what tools
2:09:13 you use and then say how could you
2:09:17 help me maintain my home home lab and
2:09:20 see what it tells you
2:09:21 have I seen the latest videos of
2:09:23 Instagram where they use mid journey I
2:09:25 have not I'll go look that up that's
2:09:27 interesting can it help with stocks and
2:09:29 shares yeah you can you can have it make
2:09:31 it will program Bots for you you can
2:09:34 even combine Bots you can if you go to
2:09:37 like a stock trading site and go to
2:09:38 their bot section you could take two
2:09:41 different Bots copy the code paste it
2:09:43 into chatgpt and say combine this into a
2:09:45 hybrid bot between these two or explain
2:09:47 how these Bots work explain what's the
2:09:49 theory behind them and then
2:09:51 take the best parts of that and tell it
2:09:53 to write you a new bot and it'll do that
2:09:54 it's it's it's magic did I say
2:09:57 prompts.chat yes prompts.chat takes you
2:10:00 to awesome chat GPT prompts and it's
2:10:04 just this never-ending list of
2:10:07 um act as a motivational speaker as a
2:10:09 debate coach as a novelist as a movie
2:10:11 critic as a philosophy teacher
2:10:14 pretty cool
2:10:15 all right I gotta go to bed soon I got
2:10:18 big meetings tomorrow big meetings
2:10:23 happy birthday thank you for the happy
2:10:26 birthdays appreciate that I'm old I'm
2:10:29 old and tired but my teeth are still
2:10:32 here some of them not all of them
2:10:36 I love this channel keep it up and happy
2:10:38 birthday thank you fellow jet access gen
2:10:40 xers stick together
2:10:43 bitter you betcha
2:10:46 oh no I came for the twerking yeah sorry
2:10:48 about that there will be no twerking
2:10:52 for the benefit of you but also for my
2:10:54 back do you know what like I would just
2:10:56 drop to the ground you know the spinal
2:10:58 injury if I tried twerking
2:11:00 um it's actually more fun to work with
2:11:02 AI instead of using it yeah I think
2:11:05 that's right like think of think of
2:11:07 cheche EBT and these AI tools like a
2:11:09 collaborator it really does change how
2:11:12 you interact with them like I find
2:11:14 myself someone said this last night
2:11:15 they're like I find myself being very
2:11:18 polite to chat GPT like thanking it and
2:11:20 saying please I do the same thing
2:11:23 what do I think of hate uh hinton's
2:11:26 latest interviews and comments so
2:11:27 Jeffrey Hinton The Godfather of AI um
2:11:29 quit Google so that he could focus on uh
2:11:33 being able to freely talk about the
2:11:34 risks of AI
2:11:36 um
2:11:37 I think it's ultimately a good thing I I
2:11:39 mean what what Hinton realized is that
2:11:42 this stuff's moving way faster than he
2:11:44 thought it would you know we've sort of
2:11:46 accomplished in five years what he
2:11:47 thought would take 30 to 50.
2:11:50 um
2:11:52 and and and what he said was if he's
2:11:54 being paid by Google there are things
2:11:57 that he might need to say like like he
2:11:59 explicitly said Google isn't doing
2:12:01 anything bad
2:12:02 but because Google's a business he might
2:12:04 say something about what the industry is
2:12:06 doing that would have a negative impact
2:12:08 on Google so he didn't want to be paid
2:12:09 by Google and be in that position where
2:12:12 he couldn't compromise his workplace
2:12:15 basically
2:12:17 um so I think that's pretty smart and
2:12:19 he's a smart guy and he understands this
2:12:21 stuff and he understands what the risks
2:12:22 are and he can answer questions like how
2:12:24 will this kill us
2:12:26 um which then means he can also answer
2:12:28 the question what could we do
2:12:30 now to have it not kill us right so so
2:12:35 you know I don't think just having
2:12:38 adorable optimism which I've been
2:12:40 accused of that before
2:12:42 um I don't think just having adorable
2:12:44 optimism is the key I think we have to
2:12:45 be realistic and say what are the actual
2:12:47 risks and how do we mitigate those risks
2:12:50 and part of that is government
2:12:53 um
2:12:54 participation but you have to have a
2:12:56 government that's competent our
2:12:57 government doesn't understand how Wi-Fi
2:12:59 works they don't understand how AI works
2:13:01 I mean come on so like I think you need
2:13:04 people like Hinton
2:13:06 um like like he's probably
2:13:08 um
2:13:09 very well positioned to be in some sort
2:13:12 of advisory role or even a cabinet role
2:13:15 within the government where all of a
2:13:16 sudden the government could have a
2:13:18 [ __ ] clue wouldn't that be nice
2:13:20 I'm not holding my breath
2:13:23 but it would be nice
2:13:26 um
2:13:28 answer my work emails yes yeah it's
2:13:30 going to get really good at that I'm
2:13:31 teaching va's how to use AI That's great
2:13:34 um teach AI how to
2:13:36 interact with your va's
2:13:38 Apple savings already made a billion
2:13:40 dollars yeah apple
2:13:44 I have I listen I've been an apple
2:13:46 Fanboy my first Mac was a Macintosh 512
2:13:49 ke look it up it was the second Mac
2:13:53 um I've been an apple Fanboy for decades
2:13:55 I
2:13:58 don't mind Tim Cook but he's a CFO he's
2:14:01 not a visionary
2:14:02 that company they lost Johnny Ives they
2:14:05 lost Steve Jobs they are missing a
2:14:07 Visionary but they're Apple they they at
2:14:09 least have the DNA to be able to do some
2:14:11 remarkable things and they have the
2:14:13 money
2:14:14 um to be able to hire the talent so
2:14:17 I
2:14:18 I keep
2:14:22 um holding out hope that Apple's not
2:14:24 gonna suck with this AI stuff I think
2:14:26 they could be really good with it
2:14:28 ask me in three months
2:14:30 if they don't make an AI announcement in
2:14:33 June I think I'm I think I'm gonna go to
2:14:35 Android no I'm not I'm gonna always be
2:14:37 an Apple guy but
2:14:39 but seriously like I am more like I was
2:14:41 always a Microsoft hater I think
2:14:44 Microsoft is the their last six months
2:14:47 has been brilliant brilliant they're
2:14:50 acting like a startup it was funny in
2:14:52 the in the early web days Microsoft
2:14:54 because they were
2:14:56 their core business was selling software
2:14:59 and shrink wrap boxes in box stores
2:15:02 and so they were very resistant to the
2:15:04 internet because because it was it was a
2:15:06 direct contradiction to how they made
2:15:08 all their money so they were they were
2:15:10 kind of
2:15:11 two three years just not there and but
2:15:15 when they decided to go internet
2:15:18 they did it like that
2:15:20 and I feel like they've done that with
2:15:21 this AI stuff as well and and they're
2:15:23 also the new CEO is taking such Glee in
2:15:27 [ __ ] with Google it's it's really fun
2:15:29 to see so I didn't think ever in my
2:15:32 lifetime I would be a Microsoft fan but
2:15:34 the last six months have kind of been a
2:15:35 Microsoft fan
2:15:37 like I'm even considering like you like
2:15:39 the office suite of products like I've
2:15:41 been using Google office stuff for I
2:15:44 don't know how long seven eight years
2:15:46 ten years
2:15:47 a long time
2:15:49 but but the the work Microsoft's doing
2:15:51 is is really smart
2:15:55 sorry lost my place again going back
2:15:58 going back going back hold hold please
2:16:01 hold please
2:16:03 I got to go to bed have I said that
2:16:05 before
2:16:07 um but I like doing this I like doing
2:16:09 these things I you know hopefully people
2:16:11 are enjoying
2:16:12 um
2:16:14 it's actually okay I'm back where I was
2:16:16 okay
2:16:19 [Music]
2:16:20 um
2:16:22 we'll use AI to catch the scammers
2:16:25 yeah so so just like you've got you've
2:16:28 got
2:16:29 um hackers right now trying to hack and
2:16:32 you know white hat you know security
2:16:34 people try to fight the hackers like the
2:16:36 hackers are gonna get access to these AI
2:16:38 tools but so are the good guys so the
2:16:40 wars will continue they're just gonna
2:16:42 have more powerful tools
2:16:43 so I agree with that
2:16:46 and yeah you'll be able to if you
2:16:48 actually catch a scammer you'll be able
2:16:49 to take them down in in pretty
2:16:50 significant ways with with these tools
2:16:52 how long will it be before the AI doing
2:16:55 virtual assistant work I think that's
2:16:57 coming very very quickly again I'm
2:16:59 hoping Apple does that really well AI
2:17:01 will certainly make smart contracts
2:17:03 easier boosting the use of dows new
2:17:06 business structures yeah I agree I I
2:17:08 think AI is gonna AI is absolutely the
2:17:11 path that is going to drive web3
2:17:13 adoption
2:17:14 yeah dowels are really interesting but
2:17:16 but
2:17:19 having a having AI assist in all the
2:17:23 scenario planning of what could go wrong
2:17:25 with the Dow
2:17:27 um I think I I think we're gonna I think
2:17:29 we're gonna start to see really
2:17:30 sophisticated structures that are driven
2:17:32 by AI that make web 3 safe palatable
2:17:38 intuitive easy transparent all the stuff
2:17:41 it needs to be right now you still need
2:17:42 to be a [ __ ] programmer to be in web3
2:17:45 and and that needs to go away like I
2:17:47 should just be able to say I want to
2:17:49 create an nft of my artwork and I I want
2:17:52 you know people on these networks to be
2:17:54 able to buy my [ __ ] and it can figure
2:17:57 out all of the translations and
2:18:00 chain swapping and whatever the like I
2:18:03 shouldn't need to know any of that [ __ ]
2:18:04 and I shouldn't need to have 400 wallets
2:18:07 to play in web 3 sorry a little bitter
2:18:11 about web3
2:18:13 um advice for how to introduce
2:18:16 to students on proper use to help in
2:18:19 their schoolwork
2:18:20 um
2:18:22 I would just start interacting with them
2:18:26 like using critical thinking and
2:18:28 exploring the tools together like I
2:18:30 don't think you need to instantly become
2:18:33 an expert overnight I I would make it
2:18:35 safe for them to talk about like who
2:18:37 here's use chat GPT make it safe for
2:18:40 them to say oh I use it and make it safe
2:18:42 for them to say I used it for my
2:18:43 homework and then say well how did you
2:18:46 use it for your homework let's do that
2:18:47 for the next assignment let's have
2:18:49 everyone do that for their homework and
2:18:51 then let's bring in our answers for what
2:18:52 chatgpt created and let's look at them
2:18:54 together and we can talk about what was
2:18:56 good and what was bad and right and you
2:18:58 can engage with them
2:19:01 about hey these tools are kind of
2:19:05 magical and you know I was running the
2:19:08 classroom on uh in a certain way and
2:19:11 then these these tools kind of change
2:19:13 that so let's figure out together how
2:19:16 we're going to change that because maybe
2:19:18 we use these tools to do things that are
2:19:20 more creative than we would have done
2:19:21 otherwise I don't know I mean I I think
2:19:23 you just engage right I think you engage
2:19:26 as humans now you're a teacher so you
2:19:28 need to kind of lead the charge there so
2:19:30 you need to have a plan
2:19:31 but the plan could be step one figure
2:19:34 out what your students are already doing
2:19:36 because they're probably doing something
2:19:38 with it
2:19:40 um I enjoy your analysis what's your
2:19:42 background are you trained and working
2:19:45 in it
2:19:47 um my background is I'm trained as a
2:19:50 Storyteller I have a degree in acting I
2:19:52 moved to New York City out of school
2:19:54 um started and ran a theater company I
2:19:56 wrote seven screenplays I wrote a
2:19:58 one-man show
2:19:59 and then in the mid 90s I was doing
2:20:03 desktop publishing as my day job when I
2:20:05 was trying to be an actor and I wasn't
2:20:07 trying to be an actor I was successfully
2:20:08 being an actor I just wasn't getting
2:20:10 paid for it so creatively successful
2:20:13 financially disaster
2:20:16 um then I discovered the world wide web
2:20:18 and I started one of the first online
2:20:20 magazines I started one of the first
2:20:21 digital agencies company called
2:20:23 agency.com
2:20:25 and then uh for the past decade I've
2:20:28 been the co-founder and CEO of a company
2:20:30 called storyvine which is a automated
2:20:33 video storytelling platform
2:20:35 so I I would consider myself more of a
2:20:38 product guy than I'm certainly not an
2:20:39 engineer and I'm certainly not an I.T
2:20:42 but I work with a lot of those people so
2:20:44 I understand the world
2:20:47 um I'm more interested in and like the
2:20:49 only thing I think I have skill for is
2:20:51 kind of the strategy around how you
2:20:54 think about AI how you might deploy it
2:20:56 how you might use it how you might
2:20:58 leverage it
2:21:00 um and I think that's the the the
2:21:02 transition we're in the middle of is
2:21:04 you don't need to be a data scientist
2:21:06 and a machine learning expert you don't
2:21:08 you don't need to create the models
2:21:10 anymore well people that create the
2:21:12 models can go create the models but now
2:21:15 that the models have been presented in
2:21:17 this magical little tool called chat gbt
2:21:20 and and now that that's going to lead to
2:21:22 all these other magical tools that are
2:21:24 really easy for regular human beings to
2:21:26 access now it starts to be about okay
2:21:29 what are the powers that these tools
2:21:32 give us as professionals and how is that
2:21:34 going to shift every industry so that's
2:21:38 that's how I'm coming at this so
2:21:41 um I'm just geeky enough to be able to
2:21:44 have the conversations and understand
2:21:45 what's going on but the minute you put
2:21:47 me in front of code my ADHD takes over
2:21:50 and I want to smash my face into the
2:21:51 screen
2:21:52 so Engineers I respect them a lot
2:21:55 because they can do that and I can't in
2:21:57 fact my my business storyline basically
2:22:00 exists because I don't have the
2:22:01 attention span to do video editing and I
2:22:03 invented a whole company to automate
2:22:05 that [ __ ] so I didn't have to do it
2:22:08 all right well AI allow you to create AI
2:22:11 personalities on the cloud
2:22:14 um who will have no idea that they are
2:22:16 real yes I think within a year it will
2:22:19 be very hard to distinguish
2:22:23 um real from fake avatars
2:22:26 um and and again this is where I think
2:22:28 that
2:22:29 um blockchain technology could be really
2:22:31 interesting where you could use
2:22:33 blockchain and nfts to validate that
2:22:37 someone is real like yeah this is the
2:22:39 real person this is like in fact someone
2:22:41 people are joking because of how poorly
2:22:43 I manage my question answering
2:22:46 um is this a pre-recorded thing it's not
2:22:48 a pre-recorded thing but it it may seem
2:22:51 like it well so imagine if one of the
2:22:54 you know icons on screen was like a
2:22:56 blockchain verified this is live this is
2:22:58 a real person because I think we're
2:23:00 going to need that
2:23:04 how can I tell the written response to
2:23:07 format better on documents like resumes
2:23:09 different colors indents so
2:23:11 um
2:23:12 so you can if if
2:23:15 it depends on the program you're using
2:23:17 but like like
2:23:19 um
2:23:19 I created a program that generated
2:23:21 emails and then it was just doing plain
2:23:23 text emails and I said
2:23:26 um I want this to look better can you
2:23:28 reformat this in HTML to make it look
2:23:31 better and it did and it just wrote the
2:23:33 HTML of that email and I used that in an
2:23:35 email thing so depending on the on where
2:23:38 you're putting the uh the document you
2:23:41 can have it tag different sections and
2:23:44 have it recommend colors all sorts of
2:23:46 things like that
2:23:48 um
2:23:48 and depending on the software
2:23:50 application you're using it might use
2:23:51 some sort of XML that you can import and
2:23:54 if you're going to Port XML then you can
2:23:55 have chat GPT generate XML it's pretty
2:23:57 remarkable
2:23:58 controls are going to be behind a
2:24:00 paywall or security of sorts no
2:24:02 um for some tools but again
2:24:06 there's so much happening in the open
2:24:08 source world world right now and there
2:24:11 are so many developers pouncing on this
2:24:13 that I think there's going to be all
2:24:15 sorts of options out there that we're
2:24:17 going to have that won't be behind
2:24:19 paywalls what I think is going to remain
2:24:22 behind paywalls and where I think
2:24:23 there's a lot of good business is
2:24:26 proprietary data sets so Bloomberg just
2:24:29 launched or just announced Bloomberg GPT
2:24:32 which is a chat apt-like model based on
2:24:35 proprietary Bloomberg data I think those
2:24:37 kind of things are going to remain
2:24:39 behind firewalls and and those are going
2:24:41 to be worth paying for because you're
2:24:43 going to have a high likelihood of trust
2:24:44 that that data is going to be valuable
2:24:47 and useful and things like that
2:24:49 how can AI be my first personal
2:24:51 financial decision maker I'm really
2:24:53 shitty with money so I I should answer
2:24:55 that one too but you can just have it
2:24:57 tell it you want it to be
2:24:58 a personal financial coach explain your
2:25:01 situation and have it give you a 10-week
2:25:04 plan
2:25:06 um
2:25:06 and and and then break each of the steps
2:25:09 of the plan down into a day by day plan
2:25:11 have it give you a day by day weekly
2:25:13 plan for the next 10 weeks and see what
2:25:15 it comes up with
2:25:18 and 10 weeks might not be right it might
2:25:20 be six months but whatever
2:25:23 um and give it your goals I want to
2:25:24 improve my credit rating I want to start
2:25:26 a savings account I want to understand
2:25:28 investing
2:25:30 tell it to teach you all that and have
2:25:32 it create a plan and then just follow
2:25:34 the plan
2:25:35 you you become the assistant let it
2:25:39 become the boss and you just work for it
2:25:41 gpd4 is blowing my mind with code I know
2:25:44 it's it's pretty remarkable and I don't
2:25:46 even code
2:25:48 um using it so so we had we in my app we
2:25:51 had this obscure bug it was an edge case
2:25:53 bug and it was only on certain browsers
2:25:55 but it like we just couldn't find it
2:25:56 couldn't find it and then our developer
2:25:58 finally just grabbed a pile of code
2:26:00 threw it into gpt4 and said can you find
2:26:03 a bug here and he had two characters
2:26:04 transposed and found that
2:26:07 we untransposed them and assumed that
2:26:09 that wouldn't change anything and it
2:26:10 fixed the bug
2:26:12 it's amazing
2:26:14 I'm going to use it to write my report
2:26:15 cards make lessons and assess yeah yeah
2:26:17 teachers should use it too like you know
2:26:19 like the students shouldn't be the only
2:26:22 ones allowed to cheat right
2:26:24 do you think Charles
2:26:27 that's
2:26:28 that's pretty funny the Charles Kamala
2:26:30 joke pretty good
2:26:32 um what human qualities do you think are
2:26:35 Irreplaceable by AI well I think
2:26:38 so I think we're moving from an
2:26:40 information age to a knowledge age where
2:26:42 where
2:26:43 large language models become knowledge
2:26:46 on demand
2:26:48 I think the thing that remains human is
2:26:50 wisdom
2:26:52 how do we use the tools
2:26:54 how do we
2:26:56 determine what we want to do with them
2:26:58 how do we curate the what's generated
2:27:02 how do we refine it
2:27:05 and how do we apply it and how do we
2:27:09 use these tools in a way that benefits
2:27:11 us as a society rather than destroying
2:27:14 it because we're doing a pretty [ __ ]
2:27:16 good job of destroying our society right
2:27:18 now without AI so maybe we could use AI
2:27:21 to unfuck up what we [ __ ] up for the
2:27:23 past 20 years
2:27:26 so [ __ ] bitter
2:27:28 Gen X baby
2:27:32 I'm not bitter enough to get involved
2:27:34 and like run for office but you know I'm
2:27:37 good at shouting from the sidelines
2:27:40 will future robots sing to their
2:27:42 children old MacDonald had a farm
2:27:46 that's pretty good yes yes they will
2:27:51 AI software developer how can I make an
2:27:53 impact on AI
2:27:56 um apply for a job at open AI
2:27:59 um
2:28:00 I mean honestly
2:28:02 go to GitHub I mean go play with auto
2:28:05 GPT go play with Lang chain go
2:28:09 play with uh baby AGI if you don't
2:28:13 follow David Shapiro on YouTube he's got
2:28:15 a whole project where he so David
2:28:17 Shapiro's written a bunch of books on
2:28:20 machine learning and and and AGI and all
2:28:23 this stuff like he's brilliant brilliant
2:28:25 dude and he's got a project where you
2:28:27 can contribute as a developer to his
2:28:29 project building an AI and AGI system
2:28:32 out of multiple systems so I would say
2:28:36 start developing like like teach
2:28:39 yourself all these tools because
2:28:41 you know all the documentations there
2:28:43 and you can use copilot on GitHub and
2:28:47 you know Chachi BT to
2:28:49 accelerate your learning accelerate your
2:28:51 development I would Dive In
2:28:55 dive in deep quick fast now go and make
2:28:59 [ __ ] better make Auto GPT not suck
2:29:02 because mid Journey the best place for
2:29:03 image making yeah it's by by far the
2:29:05 highest quality images in my opinion um
2:29:08 stable diffusion is a close second and
2:29:10 stable diffusion is more
2:29:12 um
2:29:13 flexible
2:29:15 for sure
2:29:17 um but it's more Technical and it
2:29:20 requires a bit more geeking out
2:29:24 stability AI is a really interesting
2:29:25 company though they keep open sourcing
2:29:27 all of their models I think because they
2:29:29 know they're going to get their asses
2:29:30 sued into Oblivion and they might not
2:29:32 exist
2:29:34 you mentioned Apple what do you think
2:29:36 they'll start with what's the opening
2:29:38 move
2:29:49 it's a great question there could be
2:29:53 an opening move could be around
2:29:55 augmented reality so they're launching
2:29:57 their big goggles which have been
2:29:58 anticipated for what 10 years
2:30:01 um I think they're finally going to
2:30:02 launch those goggles
2:30:04 um I don't have high hopes for them but
2:30:06 you know I think they have to launch
2:30:08 them because everyone's been waiting for
2:30:09 them so I think they could I think they
2:30:12 could roll some
2:30:14 um large language model you know kind of
2:30:16 smart chat stuff into the augmented
2:30:18 reality where your augmented reality is
2:30:19 like really really smart and does really
2:30:21 cool things
2:30:23 the other thing that is the thing that I
2:30:26 think might have more impact is
2:30:29 um they could create AI agents that you
2:30:33 Deploy on your behalf so so think you
2:30:35 know the little characters that can talk
2:30:38 like you and there's like the unicorn
2:30:40 and
2:30:41 so imagine if those didn't suck right
2:30:43 and imagine if they weren't just puppets
2:30:47 um
2:30:49 and you know maybe it's the kind of
2:30:51 thing where they make it really easy to
2:30:53 make it kind of look like you like this
2:30:55 you know the Kyle Shannon dream stuff
2:30:57 like let me find a dude that kind of
2:30:58 looks like me
2:31:00 like there's a bit in the rain so
2:31:02 imagine if I took like this stylized
2:31:04 version of myself and I could turn that
2:31:06 into an agent and but then that agent
2:31:08 could go do stuff for me it could go
2:31:09 book movie tickets and it could build me
2:31:12 websites and it could do things like
2:31:13 that so I could see them doing something
2:31:15 around
2:31:18 super easy to use
2:31:20 agents that have deep AI stuff
2:31:24 underneath them but on the surface
2:31:25 they're like these simple tools that
2:31:27 just do cool [ __ ] for you
2:31:29 so maybe something with ar maybe
2:31:31 something with agents
2:31:33 based on my experience since Steve Jobs
2:31:35 died probably nothing
2:31:37 I just I just hope it's not nothing
2:31:41 um my AI from Snapchat is a favorite
2:31:43 with kids because it's Avatar based
2:31:46 what's another I actually don't know I
2:31:48 heard that some of the Snapchat feedback
2:31:50 was that people were hating it the fact
2:31:52 that that
2:31:53 um it's a kids it's a favorite with kids
2:31:55 is interesting because it's Avatar based
2:31:58 yeah I think I think the avatars
2:32:01 um avatars are going to get really
2:32:02 interesting they're going to get really
2:32:04 interesting they're going to get really
2:32:05 good they're going to get really smart
2:32:08 um
2:32:09 and it's interesting like like avatars
2:32:12 are not gonna start in the metaverse
2:32:14 right it's not going to start like
2:32:16 Zuckerberg you know showed in the shitty
2:32:18 videos right we're not going to go into
2:32:21 a world and then represent ourselves as
2:32:23 an avatar and stand around like awkward
2:32:26 junior high students at a at a at the
2:32:29 dance
2:32:30 um which is which is what you know
2:32:32 virtual reality is like now I think we
2:32:34 start out with these avatars as agents
2:32:36 right so they're kind of independent
2:32:38 like little dudes that that can you know
2:32:41 show up in Tick Tock or show up you know
2:32:44 in chats or or things like that and then
2:32:47 eventually we'll migrate into 3D worlds
2:32:50 um so I think I think we'll start using
2:32:53 virtual agents and avatars soon how can
2:32:56 I make drone delivery business using
2:32:58 geospatial and AI
2:33:01 um what's the name of that
2:33:06 that drone business that Mark Roper just
2:33:09 did
2:33:12 a video on hang on a second
2:33:15 I mean I know nothing about that world
2:33:17 but this company is fascinating
2:33:20 Mark Roper
2:33:24 um drone
2:33:28 so this is the video It's called
2:33:31 this drone will change making lunch so I
2:33:34 place it so the company is called
2:33:38 um
2:33:41 hold
2:33:50 why does not have this in his
2:33:58 hang on a second zip line
2:34:01 so there's so so
2:34:03 um go to YouTube search for Mark Roper
2:34:06 r-o-p-e-r or Rover r-o-b-e-r
2:34:10 and then the video is called this drone
2:34:13 will change everything it's a company
2:34:15 called zipline
2:34:16 um really fascinating company really
2:34:19 fascinating Tech they're using AI
2:34:21 they're doing like real-time delivery of
2:34:25 blood to hospitals and and this thing so
2:34:28 rather than the Drone coming down and
2:34:30 bringing the package down and like
2:34:32 landing on your table and having like
2:34:34 the blades cut your face off it stays at
2:34:36 400 feet above its location and then it
2:34:40 uses a zip line it drops like the basket
2:34:42 down but the basket has its own
2:34:44 propellers and it and and it's like
2:34:47 dinner plate sized accuracy so they're
2:34:50 delivering in real time like blood to
2:34:52 hospitals so so a doctor will just say I
2:34:55 need O negative blood in these rural
2:34:57 areas of Africa and these drones just
2:35:00 get shot out of these catapults and so
2:35:02 so this is the video for it um this
2:35:05 drone will change everything so I would
2:35:07 start with this just because I think
2:35:08 it's quite remarkable and then figure it
2:35:10 out from there I don't know enough about
2:35:12 that world to give you any real advice
2:35:15 other than here's a video
2:35:19 um chat GPT tell me I'm not fat I need
2:35:21 that too
2:35:24 um like Google you can use it for Google
2:35:26 Docs and I'm incredibly likable and
2:35:28 informative thanks so much I appreciate
2:35:30 that
2:35:31 um get on the wait list for the plugins
2:35:32 yes get on the waitlist for chat GPT
2:35:34 plugins they're going to be really cool
2:35:36 I've got access to the plugins but I
2:35:37 don't have any access to any of the ones
2:35:39 that I want so
2:35:41 I'm cranky
2:35:42 [Music]
2:35:43 um
2:35:45 we'll we will all have our AI doctor
2:35:48 that knows everything we do all day I
2:35:50 agree with that and that's a good thing
2:35:52 because
2:35:53 it'll they'll be more empathetic than
2:35:56 current doctors and they'll give us time
2:35:58 and they'll give us answers and and when
2:36:01 these things get you know can validate
2:36:03 answers and validate that it's actually
2:36:04 good advice it's going to get really
2:36:06 interesting I think AI is going to
2:36:08 bridge the gap for people with
2:36:09 disabilities totally agree with that
2:36:11 their their work can now be amazing I
2:36:13 totally agree with that like like I just
2:36:16 I mean these the AI tools are amplifiers
2:36:20 right they're like Brilliance amplifiers
2:36:22 and so you can take things that you're
2:36:24 not good at and it can be good at that
2:36:26 for you and all you have to have is the
2:36:28 intention
2:36:29 so yeah I think I think it's going to be
2:36:31 remarkable for for disabled people how
2:36:34 do I feel coders careers will be
2:36:36 impacted dramatically and fast start
2:36:38 learning AI
2:36:42 hello what can AI what AI can I use to
2:36:45 create and blank I don't know oh wait
2:36:48 one day I can I use to create a
2:36:52 oh a baptism invite
2:36:55 um
2:36:56 go to canva canva's got a bunch of AI
2:36:58 tools I don't know that you need AI for
2:37:00 that you can have chat GPT right the
2:37:02 write the copy and then go to canva and
2:37:04 just pick one of their designs
2:37:06 but canva's got a bunch of generative AI
2:37:08 stuff in it now so I'd say just go play
2:37:10 with canva is chat gbt usable in the
2:37:13 restaurant business chati PT is usable
2:37:14 in every business if so give an example
2:37:17 sure
2:37:19 um
2:37:20 how's your Chef have it come up with new
2:37:23 recipe ideas
2:37:24 um come up with a concept like Tuesday
2:37:27 is
2:37:28 um
2:37:32 uh I I don't know a particular thing
2:37:34 like no chicken night so so tell it to
2:37:37 come up with a menu for for the kind of
2:37:39 restaurant you are
2:37:41 um have it help you shop have it help
2:37:44 you
2:37:45 um organize
2:37:47 um
2:37:48 have it analyze your profitability and
2:37:52 suggest ways to be more profitable and
2:37:53 where you can save money there's just an
2:37:56 infinite number of ways that you can use
2:37:58 it I would say start with tell
2:38:02 tell chat GPT to act like an experienced
2:38:05 successful restaurateur
2:38:09 and you wanted to give you advice on how
2:38:12 to improve your restaurant
2:38:13 have it do marketing come up with a
2:38:15 marketing plan for you have it come up
2:38:17 with a word of mouth campaign for you
2:38:19 literally any part of your business just
2:38:22 start interacting with it pick a
2:38:24 specific problem you have in your
2:38:26 business work with chat GPT to solve
2:38:28 that and then from there it'll the world
2:38:31 will open up
2:38:32 can I explain the prompts
2:38:35 um
2:38:36 yeah prompts are there's the whole thing
2:38:40 what I would say is this go to
2:38:41 prompts.chat
2:38:44 um which takes you to this thing called
2:38:46 that will take you here
2:38:48 um this this top part of the document
2:38:52 will teach you how to prompt and then
2:38:55 um the the whole thing is just a long
2:38:58 series of prompts that will
2:39:01 um put chat chat GPT in a specific kind
2:39:04 of mode like a personality that you can
2:39:07 then interact with that's a good way to
2:39:09 start you just kind of need to get
2:39:11 curious about it and start playing
2:39:13 it's going to be interesting prompting
2:39:15 in the next
2:39:17 two years is probably not going to be as
2:39:19 critical as skill because the machines
2:39:21 are going to start prompting themselves
2:39:23 like a lot of what we have to do now is
2:39:26 going to drift into the background and
2:39:27 there's going to be nice interface for
2:39:28 we'll just click on [ __ ] and it'll it'll
2:39:31 just automatically do stuff
2:39:33 um but right now you kind of need to
2:39:35 understand prompting so I'd say just
2:39:36 start start teaching yourself about it
2:39:39 you can also have chat GPT teach you how
2:39:42 to prompt and it will
2:39:44 students with learning disabilities can
2:39:47 now write their ideas and make it
2:39:48 amazing in seconds I know it's amazing
2:39:52 can we rather have ai managers instead
2:39:55 of replacing employees yeah absolutely
2:39:57 right
2:39:58 and quite frankly like like I feel like
2:40:01 anyone in a company who gets curious
2:40:04 about AI can probably move up the ranks
2:40:06 pretty quickly
2:40:08 if they're if they're not a jerk about
2:40:10 it can I make AI modules that can make
2:40:13 predictive maintenance
2:40:16 for companies I don't know what kind of
2:40:19 maintenance
2:40:20 but probably there's a good buddy of
2:40:23 mine that's part of the AI Salon just
2:40:27 created a an app called maintenance
2:40:30 Mentor that's like a series of of
2:40:33 experts that maintenance people can use
2:40:36 to like
2:40:37 troubleshoot you know air conditioning
2:40:40 issues or Plumbing issues or electric
2:40:42 issues so I don't know if that's the
2:40:44 kind of Maintenance you're talking about
2:40:45 but yeah you can you can absolutely
2:40:48 start to do that it it'll require a
2:40:50 little bit of programming I mean you can
2:40:52 use no code tools like zapier and some
2:40:56 of those other tools but
2:40:58 can I make yeah yeah
2:41:00 um
2:41:02 I just woke up at 7 30 a.m in South
2:41:04 Africa good morning happy birthday thank
2:41:06 you Lindo I appreciate it I'm 48 I feel
2:41:09 you I'm 58 come on man it's it's it's
2:41:12 just awful this aging thing don't do it
2:41:14 don't do that's my advice
2:41:17 maybe I'll live long enough to uh
2:41:20 you know AI will figure out the reverse
2:41:24 aging thing and I I actually think
2:41:26 that's a real possibility now I didn't
2:41:28 think that six months ago I think
2:41:30 Educators who say it's plagiarism are
2:41:32 wrong the ideas are the students the
2:41:34 work is improved yeah the the Educators
2:41:37 that are
2:41:38 embracing chat gbt and finding and and
2:41:42 letting it disrupt
2:41:44 how they've done things in the past and
2:41:46 finding new ways to engage with students
2:41:48 those are the ones I'm really excited
2:41:49 about and there's a lot of them out
2:41:51 there it's cool
2:41:53 how do I feel the programming industry
2:41:54 will be impacted profoundly deeply go
2:41:57 use
2:41:57 [Music]
2:41:58 um
2:41:58 GitHub co-pilot if you haven't have it
2:42:01 write you some documentation or
2:42:04 refactor some code for you
2:42:08 um I need to get back to written tests
2:42:11 in school that might be part of it
2:42:13 how can I get hooked up to office email
2:42:16 oh how can I get AI hooked up to office
2:42:19 email just wait for six months
2:42:22 um Microsoft won't release it
2:42:24 I used to hate Apple now we love it I
2:42:27 used to love Apple now I hate it
2:42:29 I just hate that they don't innovate
2:42:31 anymore and their products suck
2:42:33 like their stuff just doesn't work
2:42:36 good
2:42:38 it always used to just work that was
2:42:40 like one of the reasons I loved them is
2:42:42 that it was just always better now it's
2:42:44 not
2:42:46 I'm a Potter I use AI for lots of other
2:42:48 things I think I'll always be able to
2:42:50 make pots though
2:42:52 experiment query with
2:42:55 um
2:42:55 if you haven't played with mid-journey
2:42:57 yet
2:42:59 experiment
2:43:00 with
2:43:01 [Music]
2:43:02 um
2:43:03 creating pottery designs with
2:43:06 mid-journey with generative AI you can
2:43:09 sort of take your vague ideas of what
2:43:11 you want to see visually and the AI can
2:43:14 really really quickly spin up new
2:43:16 Concepts for you some of which may
2:43:18 inspire you to to go in a different
2:43:20 direction and an expanded direction for
2:43:23 your Pottery so I think that could be
2:43:25 really cool still going I know I'm still
2:43:27 going I got to go to bed it's ridiculous
2:43:29 it's midnight what am I doing what am I
2:43:31 doing with my life I have important
2:43:33 calls tomorrow
2:43:34 I lost my place in the thing though
2:43:36 I'm gonna go I promise I am
2:43:39 why are you all still here well it's
2:43:40 down to 61. we're 150 for a while I
2:43:43 think people like he just won't shut up
2:43:52 I'm going to do a few more of these
2:43:54 chechi PT Guru hey you um you do edit
2:43:57 one of my videos today thank you for
2:43:59 that chat you patina AI tools are simply
2:44:01 just amazing there are no words to
2:44:03 describe what's happening I totally
2:44:05 agree
2:44:07 um that was the first video I ever had
2:44:09 to do edit by the way that was pretty
2:44:11 cool
2:44:12 is dueted right one of the two things
2:44:14 are next to each other that's duet right
2:44:18 how do you teach an old guy Tick Tock
2:44:20 you don't get off here you old man
2:44:23 how do you feel the career of coding
2:44:25 will change
2:44:27 um again deeply profoundly
2:44:30 how long can proprietary data sets
2:44:33 remain private before their data is
2:44:35 mined by other AI
2:44:37 it's a good question
2:44:40 I mean
2:44:41 you know there's an argument to be had
2:44:45 that just like internet
2:44:48 should be a utility that is just just
2:44:50 everyone should have access to
2:44:53 there's an argument that if you take the
2:44:56 concept of museums and universities
2:44:59 where they are designed to house the
2:45:03 output of humanity
2:45:06 there's an argument that you could make
2:45:07 to say the ultimate role of these large
2:45:10 language models and the image models
2:45:12 should be to learn
2:45:15 all of humanity all of the output of
2:45:18 humanity
2:45:20 and then we get to play with it
2:45:23 so
2:45:24 I don't think it'll go there but
2:45:27 you know maybe in a 50-year time frame
2:45:30 you you could get to a place where it's
2:45:32 where it's like you know these machines
2:45:33 are just you have access to everything
2:45:35 that ever was and ever will be and
2:45:38 is ever happening right now
2:45:40 I don't think it's going to happen
2:45:42 anytime soon but that's my that's my
2:45:44 adorable optimism
2:45:46 version of the uh the AI of the future
2:45:48 how do we use it to benefit us as a
2:45:51 species
2:45:52 um we don't hide from it
2:45:54 and jump in now and start using it to do
2:45:57 good [ __ ] and and like this doesn't have
2:46:00 to be on a big scale I mean the the
2:46:02 impact of you taking the time to teach
2:46:06 someone in your community about some
2:46:09 cool thing you discovered that empowers
2:46:10 them to do something else like the
2:46:12 butterfly effect of
2:46:15 embracing this and evangelizing this
2:46:18 with other human beings will be the
2:46:21 thing that will ultimately benefit us as
2:46:24 a species because
2:46:27 good things done now are going to have a
2:46:29 disproportionately
2:46:31 Amplified effect
2:46:34 um on the future of AI
2:46:36 and and similar with bad things done now
2:46:39 because we're so early
2:46:42 all right I'm getting tired I'm gonna
2:46:44 burn
2:46:48 some comments in this live were filtered
2:46:50 to protect the community's experience
2:46:52 hey those of you saying nasty things
2:46:54 don't do that tick tocks tick tock's
2:46:57 watching you and it just told me it's
2:46:59 watching you thoughts on AI crypto coins
2:47:01 I've heard agix Singularity net is good
2:47:04 I don't know I stopped paying attention
2:47:06 to web3 stuff
2:47:08 a year ago when I when I got into AI I I
2:47:11 kind of let that stuff drift to the
2:47:14 background and I'm really happy I did
2:47:18 um is there a place for AI startup ideas
2:47:20 like a Kickstarter for AI yeah like
2:47:23 yes let's go start a business have I AI
2:47:28 help you start that business if you're
2:47:30 starting a business check out GPT boss
2:47:33 McKenzie Bose's startup where you can
2:47:36 hire virtual employees to help you start
2:47:37 your business
2:47:41 um
2:47:43 how do I feel the career of coding will
2:47:45 change is this live or pre-recorded
2:47:48 why why would I record this this is not
2:47:51 something that should ever be recorded
2:47:52 canva is not the greatest AI in AI yet
2:47:55 yeah I know but it's fast
2:47:58 is there a free version of mid-journey
2:48:00 you can do a handful of images before
2:48:02 they start charging you but like the the
2:48:04 the one I pay for is the 10 10 Buck a
2:48:07 month one and I use it a fair amount and
2:48:10 I don't hit my my top if you're going to
2:48:12 do it a lot you might have to do the
2:48:14 more expensive one but 10 bucks isn't
2:48:15 bad for mid-journey because it's
2:48:17 remarkable
2:48:19 our fears of AI reasonable I think fears
2:48:22 of AI are healthy
2:48:25 um I think unreasonable fear of AI is
2:48:28 unreasonable
2:48:30 um the robots are going to kill us
2:48:32 narrative has got to go
2:48:36 you know we need to get involved so that
2:48:38 the robots don't kill us
2:48:40 what is web 3 I never figured it out so
2:48:42 web3 is is what they're talking about
2:48:45 with blockchain like blockchain enabled
2:48:47 stuff and
2:48:49 distributed finance and
2:48:53 um
2:48:53 trustless systems like nfts where the
2:48:56 you can prove that the nft is yours
2:48:58 because it's you know written to the
2:49:01 blockchain and things like that so so
2:49:03 it's basically a um a trackable
2:49:09 um and and autonomous web where you can
2:49:13 do you can create things like Dows which
2:49:14 stand for
2:49:16 um decentralized autonomous
2:49:18 organizations where you can basically
2:49:19 write the rules of a company
2:49:22 and then people can join that company
2:49:24 and they can put money into that company
2:49:26 and the rules of how that company runs
2:49:28 and how people get paid and how people
2:49:31 pay into it is all automated
2:49:33 so you don't have human beings
2:49:36 arbitrarily changing the rules when you
2:49:38 create a Dao for example
2:49:40 the rules are written into stone they're
2:49:43 written into the blockchain it's
2:49:44 verifiable by other people so they're
2:49:47 called trustless systems which always
2:49:49 seems counterintuitive to me because you
2:49:51 can trust them more because you're not
2:49:53 actually trusting a human being so so
2:49:56 web3 is just basically you know nfts
2:49:59 blockchain and and all the like all the
2:50:01 things that come with those Technologies
2:50:03 they're really interesting Technologies
2:50:05 unfortunately sort of the crypto bro
2:50:07 crowd
2:50:09 created a bunch of companies that
2:50:12 were cash grabs and we're kind of
2:50:15 [ __ ] and scammy and then some of the
2:50:17 big exchanges like FTX
2:50:21 did shitty scammy stuff and so
2:50:25 um
2:50:26 the web 3 world needs to go sit in the
2:50:29 corner for three years
2:50:30 um while they get spanked and then and
2:50:34 then you know I think I think the
2:50:36 technology is actually really important
2:50:37 it's going to be valuable for uh for AI
2:50:41 and I think AI is going to drive it so
2:50:43 anyway
2:50:45 where did I get that Avatar I made these
2:50:47 avatars in uh stable diffusion using
2:50:50 dream Booth where I can upload
2:50:53 pictures of myself and it kind of learns
2:50:55 what I look like and I can make like
2:50:56 little dudes little versions it's kind
2:50:58 of cute
2:50:59 it doesn't look dissimilar to how I
2:51:02 looked when I was a kid
2:51:04 that's that's see if I did twerking on
2:51:08 here that's the outfit that I would wear
2:51:10 I think maybe that's I've been doing
2:51:12 this for 14 hours so
2:51:15 um thank you all for your time hope this
2:51:17 was enjoyable
2:51:18 I didn't cost too much which I
2:51:20 apparently did but what are you gonna do
2:51:22 all right
2:51:23 peace