aiLL 7/12/23 - How Generative AI is Already Impacting Business and Society

Description

Chapters

Transcript

0:08 [Laughter]
0:09 um
0:10 [Music]
0:12 [Laughter]
0:15 [Music]
0:28 welcome welcome welcome
0:30 to the AI learning lab
0:34 we're gonna do some learning about AI in
0:37 the lab or at least a home office
0:40 welcome welcome welcome welcome welcome
0:42 welcome welcome welcome welcome welcome
0:44 welcome welcome
0:48 to the Future
0:51 good day J word what's up Phil
0:55 how you guys doing
0:58 what night is tonight Thursday Wednesday
1:00 Wednesday
1:01 happy Wednesday what's up earthlings
1:04 good day Kevin
1:09 um
1:11 a lot of people have showed up but Tick
1:14 Tock has not yet told me how many people
1:16 are here so I don't know if you all can
1:17 see it
1:19 looks like zeros looks like none right
1:22 now nothing I got nothing
1:24 except I see people I see people coming
1:27 in welcome welcome 25. cool is it
1:30 possible for AI to act like people via a
1:33 bot yeah look at me
1:35 I'm sorry
1:37 I need to catch up to speed oh you're
1:39 you're being serious okay is it possible
1:42 for AI to act like people via a bot uh
1:46 yeah
1:46 you know I think I think um
1:52 it you know if if you're if you're sort
1:54 of chat interacting with
1:57 this generative AI stuff it can they did
2:00 a study about three weeks ago some
2:02 results came out that a bunch of
2:04 patients interacted with chat GPT
2:06 doctors and then they interacted with
2:08 real doctors and the patients
2:13 largely preferred the uh GPT doctor over
2:16 the real doctor they said they were more
2:18 competent and more empathetic so
2:21 so yes you can uh welcome everybody this
2:24 is the AI learning lab I'll show you
2:26 what these URLs here and are in a minute
2:29 I actually added one it's been an
2:31 interesting two or three days
2:33 um chat GPT all of a sudden has some
2:35 real plausible
2:38 um competition
2:40 and uh that's that's interesting I mean
2:42 I figured that would be coming and I
2:44 said I wouldn't change this list until
2:46 some new [ __ ] happens you know till
2:48 someone emerges that that
2:51 knocks someone off the list and it's
2:53 funny I haven't knocked anyone off the
2:54 list but I added one so so we'll see
2:58 we'll see how that goes
3:00 108 can AI get rid of trump well he's
3:03 still he's out there I don't know he's
3:05 he's got he's got some issues I don't I
3:07 don't know that we need AI for that he
3:10 seems to be managing that himself
3:15 and also I don't do politics here I do
3:18 not care to speak about it I'm tired of
3:21 the divisiveness and I'm hoping AI can
3:24 make us all human again
3:32 and I think it's actually a possibility
3:34 um these Bots actually
3:37 um speak to you with nuance and logic
3:40 and
3:42 um empathy
3:44 ironically enough wouldn't it be amazing
3:47 if the Bots rather than killing us were
3:49 the things that freed us up to be decent
3:51 human beings again
3:53 that'd be fun
3:56 anybody done with that
3:59 um anyway welcome everybody this is AI
4:00 learning lab my name is Kyle Shannon a
4:02 little bit about my background I'm an
4:03 entrepreneur I've started 12 or 15
4:05 companies in my day
4:07 I was an innovator in the early days of
4:08 the World Wide Web I started one of the
4:10 first digital agencies a company called
4:12 agency.com and I was Chief creative
4:14 officer there and the parallels between
4:17 the early days of the web and the early
4:19 days of this generative AI stuff are
4:22 insane and it's got my mind all a
4:25 flutter and that's what this channel is
4:27 about just talk about what I'm seeing
4:30 answer anything I can if you have
4:31 questions about AI pop them down below
4:33 if you like this kind of content that's
4:35 what the channel is about so give me a
4:36 follow and I see you're tapping the
4:39 screen there that's great that that'll
4:41 let people know we're here
4:42 and uh yeah let's get let's get rolling
4:45 on this
4:46 fantastic Thursday night Wednesday night
4:50 [Laughter]
4:53 all right oh I've also been giving out
4:55 thank you very much I've been giving out
4:58 free business ideas like the last three
5:01 nights it's been at least two business
5:02 ideas a night that that are gonna be
5:04 billion dollar ideas I just give them
5:06 away for free
5:07 you don't even have to sign it it's not
5:09 even like a raffle or anything I just
5:11 give them I just you just take them you
5:13 take these business ideas I come up with
5:15 I'm like answering someone's question
5:17 I'm like hey here's a business I give it
5:19 away give it away for free why because
5:21 that's the kind of guy I am
5:23 that's really more just because I have
5:24 ADD and I have 4 000 ideas a minute and
5:27 so there's more where that came from so
5:30 and and and quite frankly the idea is
5:33 never the thing it's always the
5:35 execution so
5:37 um
5:38 I've learned that many day many times in
5:41 my life what I've also learned is if you
5:43 say an idea out loud it is happening it
5:46 is being started while you say it out
5:48 loud it's being started at the same time
5:49 and if you don't do yours three months
5:52 later that business will show up in the
5:54 world
5:55 I've seen it happen like so many it's
5:58 freaky it's freaky how many times that
6:00 happens so if you have an idea and you
6:02 you don't want it stolen just don't say
6:03 it out loud
6:06 but the minute you say it out loud you
6:07 better do it
6:09 um
6:11 okay so let's talk about the if you're
6:13 new to generative AI if you have not
6:15 played with chatgpt that's what the top
6:17 URL is that's the official chat GPT
6:19 website
6:20 and then Bing Claude and poe.com
6:23 cloud.ai
6:26 um
6:27 are are all different flavors of it so
6:30 so Bing is Microsoft's implementation of
6:33 chat gpt4 so you can play with you know
6:37 the official gpt4 there that's
6:39 Microsoft's version of gpt4 Claude is a
6:42 competitor to gpt4 and they just
6:46 launched clod two
6:48 where you can upload files and it'll
6:50 read them and you can interact with them
6:52 so you can upload a big PDF it also has
6:54 a hundred thousand token context window
6:56 so that means if you go to claude.ai
7:00 you can input a prompt as big as 75 000
7:04 words so like your whole novel The Great
7:07 American novel that you wrote you can
7:09 just pop it into Claude and start you
7:11 know having it edit it for you give your
7:13 other ideas and then poe.com actually
7:15 lets you play with the Claude models the
7:18 open AI models and their own model and
7:20 plus it's got a or waitpo.com and plus
7:23 it's got a bot Builder you can make your
7:26 own bot and the bot now can uh you can
7:30 you can use gpt4 and Claude 100
7:34 000 token in your bots so the Bots that
7:37 people are going to make with po.com are
7:39 going to be insane so these are all just
7:41 different flavors of chat GPT if you've
7:44 not played with one of them go do it now
7:46 prompts.chat will teach you how to
7:48 interact with them it'll give you a
7:49 primer on prompting and then give you
7:51 some ideas and then futurepedia.io
7:53 that's just a giant pile of AI tools
7:56 that I can't keep up with them anymore
7:58 so Matt wolf I think is his name that
8:01 created that he's keeping up with that
8:02 [ __ ] I can't do it anymore so I am I am
8:05 I am delegating that uh to some uh some
8:09 YouTuber dude
8:11 dude
8:14 am I Gen X yes I am Gen X Gen X in the
8:19 house
8:20 I ain't no bloomer
8:23 you know what else is I'll tell you what
8:25 you got to be careful with Gen X because
8:27 we know how to use computers and we're
8:30 going to be really good at this [ __ ] so
8:32 uh
8:33 you yield little young whippersnappers
8:35 we're gonna figure out how to use this
8:37 AI stuff you're gonna be like what do
8:39 you mean I lost my job to an old man
8:43 it's gonna happen
8:49 all right
8:51 will AI reveal the Illuminati and the
8:54 aliens if the data is there
8:58 um one of the things that's really good
9:00 at is looking at disparate pieces of
9:02 data and analyzing it and pulling it all
9:04 together if you haven't played with code
9:05 interpreter
9:07 go do that because uh if there is day to
9:10 be hat data to be had that can prove
9:13 that and you can get to that data why
9:15 indeed you know you might be able to
9:17 reveal some [ __ ] it's uh it's pretty
9:19 amazing stuff what it's capable of doing
9:22 uh yep let's Unite Here yes I am Gen X
9:25 can AI make sure we get a monthly
9:27 stipend since they're going to take over
9:28 all our jobs well AI can't make sure
9:30 that we do that but we as a society
9:33 better [ __ ] figure something out
9:35 um I don't think we're gonna need
9:38 um
9:39 Universal income for long
9:42 I think that
9:44 um
9:45 my my
9:46 suspicion or my
9:49 the way the way I'm seeing things play
9:52 out
9:52 I think there's going to be a two to
9:54 three period two to three year period of
9:57 painful disruption where people lose
10:00 jobs and then we're going to start
10:01 seeing entirely new categories of jobs
10:04 being invented
10:06 because of AI so this happens at every
10:10 major technological innovation but this
10:13 one's gonna this one's gonna be big
10:14 profound and fast
10:16 but I think also the the kind of the
10:19 prosperity kick up because of this stuff
10:21 and the
10:23 um new jobs that are created because of
10:26 this stuff I think they're gonna they're
10:27 gonna kick up as fast as well so I don't
10:29 think this is I don't think the pain is
10:31 going to be like a decade I think it'll
10:33 be more like two three five years maybe
10:36 and then it'll start to get interesting
10:39 AI seems nicer than humans a it's Danny
10:42 that's it it
10:43 yes
10:45 especially in the states and it seems
10:48 like this is not just the states but
10:50 there's there's just a lot of
10:51 polarization going on right now so we're
10:53 not taking each other in very much these
10:55 tools are actually really good at
10:57 listening and responding reasonably to a
11:00 to an inquiry
11:02 um and we as humans have gotten really
11:04 shitty at that in the past decade
11:07 um I mean I think we've always been
11:08 generally bad at it but so these things
11:10 are good I have a feeling that these
11:12 tools are going to help us be better
11:13 people
11:16 all right what's the xai about Elon Musk
11:21 is behind it so Kevin so yes Elon Musk
11:25 put 50 million dollars into open AI
11:29 as a non-profit and he said he did it in
11:31 reaction to having some dinners with
11:33 Sergey Brin from Google who was a little
11:35 bit too flippant about uh AI safety and
11:39 so
11:40 um Elon Musk said I'm going to start a
11:43 non-profit to make sure that
11:45 someone is doing something with AI that
11:47 isn't just Google because he didn't like
11:50 his attitude and so it started as a
11:52 non-profit
11:53 and then depending on who you believe
11:56 I'm gonna go with Sam Altman because Sam
11:59 Altman actually seems reasonable and
12:00 thoughtful and Elon Musk seems a little
12:03 like emotional and rammy but
12:07 what ended up happening was
12:10 as open AI learned more and more about
12:12 how to use the Transformer model that
12:14 Google invented ironically enough what
12:17 they realized is that the more data they
12:19 threw at
12:20 Transformers and the more processing
12:23 power they threw at them the better they
12:25 got
12:25 and what they realized is they needed to
12:28 throw so much compute power at it that
12:30 it thank you for the Roses appreciate
12:32 that they needed to throw so much
12:34 compute power at it that it
12:37 um
12:38 they couldn't raise that much money as a
12:39 non-profit so they went to what's called
12:42 a capped for-profit which I had never
12:44 heard of before I don't know if they
12:45 made that up I assume they didn't but a
12:48 capped for-profit which just basically
12:50 says no investor can make more I think
12:51 it's 100 x their investment which is
12:54 still a lot
12:55 um
12:56 and it's got some limited rights Sam
13:00 Altman doesn't own any of the stock
13:02 which is interesting he's the founder
13:04 and CEO he doesn't own any of the stock
13:07 um
13:08 so so that's all the background
13:11 and then about what is it two months ago
13:14 three months ago Elon Musk signed a
13:17 letter that said
13:20 anyone who's doing AI research like open
13:22 AI why don't you all take a six month
13:24 pause because this stuff's dangerous and
13:27 we don't want it advancing so why don't
13:29 you all take a six month pause and a
13:31 bunch of people signed it and but Elon
13:33 Musk was kind of at the front of the
13:34 line there and then
13:36 a week later Elon Musk musk accepted a
13:40 shipment of 10 000 gpus and started x
13:43 dot AI which is a competitor to open Ai
13:46 and to anthropic and to inflection
13:50 um which which also probably deserves a
13:52 spot on that list Pi dot a i p i dot a i
13:58 um so yeah so that's what that's what x
14:01 a i is about so at some point Elon Musk
14:04 is going to come out with now what he's
14:05 been saying is these things should be
14:07 open source they should be free to the
14:08 world
14:10 um
14:11 we'll see what he does
14:13 see what he does Biz idea for someone AI
14:16 adjacent adjacent a geeky chicken wings
14:19 place that shows star Star Wars no
14:22 sports ball that's actually not a bad
14:24 idea
14:26 you could you could just do it would
14:28 just be all sort of geeky kind of cult
14:30 films you could you could narrow it to
14:32 sci-fi but you could also
14:34 you could also just take you know cult
14:37 films that's kind of that's not a bad
14:39 idea
14:40 what could you do with AI with that I
14:42 don't know you could remix your movies
14:44 in real time
14:46 I want to see Star Wars with uh the cast
14:49 of Indiana Jones
14:51 there you go
14:53 have you been on Shark Tank no but I was
14:56 on there was a show called American
14:58 inventor
15:00 um that was before
15:02 um before Shark Tank it only lasted one
15:04 season but I I was on one season of it
15:08 um I had a my inventions were a long
15:11 sled so it's basically like a winter
15:13 sled but it had wheels on it so you
15:15 could go down Hills in summertime
15:17 and in the South and then
15:20 um we had this we had this thing called
15:23 zip lights which was was like a magnetic
15:25 base and and like you know those little
15:27 zip things that the string would pull
15:29 out and zip back in with a little light
15:31 on the end of it a little LED light
15:33 and like one of the inventor dudes there
15:35 said um said so you mean tell me people
15:38 are gonna buy a light and stick it in
15:41 their closet and just leave it there and
15:44 I'm like yeah and it was like that's
15:46 ridiculous and like two years later the
15:48 those Puck lights came out that they
15:51 sold billions of so I don't think my
15:54 idea was stolen I just think that that
15:56 guy was an idiot
16:02 pension company called invention Asylum
16:05 and so uh my buddy Steve session who
16:09 um is the CEO of a company called zero
16:11 shoes
16:13 x-e-r-o-s zero zero zero shoes
16:17 um
16:18 he uh he was on Shark Tank he and his
16:20 wife were on Shark Tank
16:22 he has no qualifications none do not
16:26 believe anything I say I have a degree
16:27 in acting
16:29 started using chat gbt to review my
16:31 novel interesting uh what's your name
16:33 imminent
16:36 I think it's a good name it's hard to
16:39 say it's it's like too many other words
16:41 uh but I'm curious like you know what's
16:44 what's your what's been uh the result
16:46 like has it been good has it been
16:48 helping how's it helping you edit I'd
16:50 love to hear
16:51 pop it down there Gen X rocks yes
16:56 all right
16:58 what are your qualifications sir I have
17:01 no qualifications I resemble that remark
17:04 you people with your desire for
17:07 knowledge and credentials
17:14 [Music]
17:16 thank you for that
17:18 what's the best AI related stock in my
17:20 opinion uh I think it's been established
17:23 I have no qualifications
17:25 um I also have a degree in acting that's
17:27 been established by Penn State and uh so
17:30 I wouldn't take any advice from me but I
17:32 can tell you the companies that
17:35 I think are interesting right now
17:37 um
17:38 let's see
17:40 that's one of them that's another one
17:42 Microsoft that's another one anthropic
17:46 po no I don't really care about that's
17:48 quora quora has been around for a while
17:50 I I
17:53 don't I don't know about Poe it's not I
17:55 don't think they're doing I don't think
17:57 they're doing interesting research I
17:59 could be wrong I don't know I mean pose
18:01 at least showing up as a competitor
18:05 um I think inflection is interesting
18:07 that's the the one that was uh the newly
18:10 relatively newly founded Company by Reed
18:14 Hoffman of LinkedIn and the dude from
18:17 deepmind and they just raised 1.3
18:20 billion dollars so that's inflection so
18:22 of the startups those are the ones I'm
18:25 paying attention to
18:26 um I think Microsoft's interesting I
18:29 think Google is struggling I think meta
18:32 is struggling but meta meta seems like
18:35 they've got a more interesting approach
18:39 to this AI stuff at least short term
18:41 than Google does
18:43 but Google invented a lot of this [ __ ]
18:46 so I just I can't count Google out
18:49 Nvidia seems like they have
18:53 a quasi-monopoly
18:57 um and that that
18:59 um the the pi dot AI that inflection
19:01 company from the deepmind dude they're
19:04 gonna spend a good chunk of that 1.3
19:07 billion dollars
19:08 on
19:09 um building a data center with twenty
19:12 thousand
19:13 um gpus like the big honking ten
19:15 thousand or twenty thousand dollar
19:17 Nvidia cards
19:19 so that's going to make nvidia's stock
19:21 do something interesting so I don't know
19:23 I assume all that shit's baked into the
19:25 price already but because you know and
19:27 what Nvidia hit a trillion dollar market
19:29 cap a month ago
19:32 but amd's trying to compete with them
19:34 but
19:35 it doesn't look good
19:37 I don't know Nvidia seems like they have
19:40 an interesting kind of lock on the uh on
19:42 the processor business the GPU business
19:46 I don't quite know how that world works
19:48 but holy [ __ ] it's it must be Mighty
19:52 expensive
19:54 pretty impressive I don't know what was
19:56 impressive the hair oh the all the hair
19:58 the hair is doing a special thing
20:01 tonight
20:02 it's it's parted it's parted in like six
20:06 different directions this is like a
20:08 Vortex there's like a wormhole in my
20:10 head where aliens are like entering and
20:12 leaving the universe
20:14 all right well what are you gonna do
20:16 [Laughter]
20:19 why don't you have a podcast
20:23 I had a podcast if you go look up uh you
20:26 should have been there I did a podcast
20:27 for a year about the early days of the
20:29 internet where we interviewed my buddy
20:32 ritesh Patel and I interviewed a bunch
20:35 of the early Founders from the early
20:37 days of the web Craig canarick and Jeff
20:40 Duchess from razorfish and just a bunch
20:42 of people a bunch of people from
20:43 agency.com it was cool
20:46 um I might do a podcast I'm listen I'm
20:48 just doing the the tick tock thing
20:50 because
20:51 it started as as I was learning about
20:54 this generative AI stuff I was like my
20:56 mind was so overloaded with ideas and
21:00 crap and learning and I didn't have
21:02 anyone to talk to about it because like
21:04 all of my friends and family were like
21:06 we don't give a [ __ ] shut up go away
21:09 and so I started the tick tock just as a
21:12 way to kind of process my thoughts and
21:14 get it out there and then I started
21:16 doing the lives
21:17 just as a way to
21:20 um not blow a couple hours a night doing
21:22 Sim racing which I was doing for like
21:24 two years in a row which is fine but I
21:27 just I feel like there's there's an
21:29 important
21:31 I think talking about this shit's
21:32 important I think getting people to try
21:35 it and be curious about it is really
21:36 important
21:37 so guys it was a joke from a steam
21:41 stream last night he has no
21:43 qualifications exactly
21:45 what's the best AI platform to build an
21:48 AI sales agent for texting leads and
21:50 moving them through a CRM
21:52 um probably the easiest thing to do if
21:54 you want to do that is just do a zapier
21:57 automation or a series of them you can
21:59 start really simply
22:02 um and you know you could you could
22:04 probably do something you know when a
22:06 lead is created in Salesforce that's a
22:09 trigger that then generates an email and
22:12 that email could be based on you know
22:14 some text from that Salesforce record
22:19 um
22:20 so I would say zapier is probably
22:23 probably it's it's it's it's got It's an
22:26 ecosystem that's got lots and lots of
22:29 services connected to it I think it's
22:30 like five or six thousand services that
22:33 you can connect connect to through
22:34 zapier so I would say just start with
22:38 you know
22:39 what's your triggering event like like
22:42 what's going to trigger you know some
22:43 communication and then how do you need
22:46 to process that through open AI or or
22:49 anthropic or whichever one you use
22:51 probably open AI
22:53 um and then how are you gonna send it
22:55 are you sending an email is it going
22:56 into a blast email or is it just
22:58 updating
23:00 um a record in a database lots of ways
23:03 to do it so if you have more specific
23:05 questions that's probably where I'd
23:06 start and then it's not an AI platform
23:09 but it's just a it's just a a platform
23:11 that you can incorporate AI into and tie
23:15 it into lots of other systems you're
23:16 probably already using and then go to
23:20 futurepedia.io that bottom URL there and
23:23 I'm sure there are sales specific AI
23:27 tools that people have created because
23:29 obviously this is something that's going
23:31 to be big
23:32 my caveat with going to futurepedia is
23:35 that there's probably 4 000 Tools in
23:38 there of which maybe 400 or not abject
23:41 pieces of [ __ ]
23:42 so the signal to noise ratio there's
23:44 going to be low there's gonna be a lot
23:46 of crap and the problem with this new AI
23:49 world and because everyone's got
23:51 templates is that all of these apps are
23:53 going to look really slick and it's not
23:55 until you've like committed and start
23:57 using it that you realize oh this is
24:00 just a Hollow piece of poop that I
24:03 shouldn't have paid for so I'd be
24:05 careful there
24:07 there's probably something there to to
24:08 look at
24:12 interesting why wouldn't you just
24:14 execute and run with the idea because I
24:16 have too many of them
24:18 and because
24:21 see so so I got a couple of things going
24:24 on I've got my current company so I'm
24:26 the the founder and CEO of a company
24:27 called storyvine it's an 11 year old
24:29 automated video production platform
24:33 um I'm now incorporating AI into the the
24:37 product vision for for that company and
24:40 re-architecting the value proposition
24:42 and just everything I'm sort of
24:44 rethinking everything about it
24:47 so that's a lot
24:48 and then I've got this I've got uh the
24:52 AI Salon that I run which is a bi-weekly
24:55 Salon like that doesn't take a ton of
24:56 time but that's some time
24:59 um I'm also the chief generative officer
25:01 from for for a like a Federation of
25:04 marketing companies and chief generative
25:07 officer doesn't mean anything I was just
25:09 given that title so the guy the guy that
25:11 runs it goes hey do you want a title
25:13 Chief generative officer I'm like sure
25:16 he goes I think it's the first one in
25:18 the world I'm like all right
25:20 um but but that takes up some time and
25:23 so so I just have a lot of a lot of
25:25 stuff going on anyway and here's the
25:27 thing I know about any idea
25:30 is that you shouldn't start it if
25:33 you you're not willing to see it through
25:36 the dark times because as exciting as
25:39 anything is once you put it in the world
25:43 you're gonna hit brick wall after brick
25:45 wall after brick wall and so if it was
25:48 just a brain fart that was like oh
25:49 here's a good idea and it really could
25:51 have been a good idea if I'm not like
25:53 you know tied to it I I it won't survive
25:56 my ADD for one thing
26:00 um but it also won't survive just the
26:01 gauntlet of starting a business because
26:03 it's just running starting and running
26:05 businesses is hard it's hard it's like I
26:09 I I wouldn't and couldn't do anything
26:12 else like at this point I think I am
26:14 98.9 unhirable
26:19 and plus I don't really want to be hired
26:21 you know you know someone can make me a
26:24 you know a ridiculous offer and then
26:26 I'll probably reconsider that but
26:27 whatever
26:29 um
26:30 yeah so so my heart's got to be in it or
26:33 I can't do it I just can't do it can't
26:35 do it can't do it some comments in the
26:38 live were filtered to protect the
26:39 community's experience don't be [ __ ]
26:42 to each other thank you all right Sean I
26:44 built one I can show you oh okay cool
26:46 Christopher connect with Sean
26:49 um I Love Actually this has been
26:50 happening a lot in the lives and it
26:52 makes me very happy of people just
26:54 um kind of connecting with one another
26:56 one of the things that is a a Hallmark
26:59 of my lives is that I try to answer all
27:01 the questions so I get really behind and
27:03 so people don't think I'm actually live
27:05 but whatever
27:08 um
27:08 and you've all been helping one another
27:11 as you have questions that's been really
27:12 cool um do me a favor double tap the
27:14 screen get some likes going to see if we
27:16 can get some more people here share the
27:18 live if you know how to do that uh and
27:20 you're willing to like if you think like
27:21 who's this damn old man I don't I don't
27:24 want anyone else to see this stupid [ __ ]
27:26 my friends see this
27:28 I'm not endorsing this you don't need to
27:31 endorse me that's fine
27:32 coffee by the way
27:39 what is the best one what is which best
27:41 one Danny what were you asking
27:45 why don't you have a podcast what's the
27:48 best podcast
27:52 oh for I I don't know for for this AI
27:55 stuff
27:56 I don't know
27:58 the only one I listen to really
27:59 consistently is the uh Paul rateser's
28:02 marketing AI Institute I listened to
28:05 that Weekly
28:06 he's got his [ __ ] together he's been
28:07 doing this for like five or six years on
28:09 this generative AI stuff around
28:11 marketing so I think it's uh that one's
28:14 that one's a good one
28:18 mine's like 120k
28:20 I don't know what that means told my
28:23 idea to someone 20 years ago and heard
28:25 it in the speech recently with all the
28:26 same examples yeah that happens I'm
28:29 telling you man you put something in the
28:30 world it's just in the world I don't
28:32 know what it is there's something about
28:34 it it just happens especially when
28:36 there's like a lot of energy there's so
28:38 much energy around this stuff the other
28:40 thing that's going to happen with these
28:41 tools
28:42 is
28:45 it's going to be incredibly easy
28:48 for people to steal other people's
28:50 businesses like incredibly easy not just
28:54 the look and feel not just the marketing
28:56 like all the typical [ __ ] you see today
28:59 the barrier to entry historically was
29:03 doing the execution part like doing the
29:05 coding of a particular kind of platform
29:07 or particular kind of service that was
29:09 always the barrier to entry but with
29:11 these tools you're going to literally be
29:13 able to say grab me all the marketing
29:15 positioning rewrite it so it sounds
29:17 different grab me the design you know
29:19 redesign it so it looks different but
29:21 it's still functionally the same and
29:22 then let's go ahead and just copy that
29:24 functionality
29:26 um I don't know what that does to the
29:29 world of
29:31 investing and startups and entrepreneurs
29:36 I don't
29:38 it seems like it's going to be a really
29:40 stressful time
29:42 like like just like just like it's going
29:45 to be stressful for people who get
29:47 displaced out of current you know nine
29:49 to five jobs I think there's going to be
29:51 similar displacement in the startup
29:53 world and in the certainly in the
29:54 investing world
29:57 um so it's gonna be I that's that's an
29:59 interesting time how do I access code
30:01 interpreter is it a plug-in or separate
30:02 service here I'll show you
30:04 um
30:09 okay so here we are at chat gbt you can
30:14 make money
30:17 um
30:18 so in order to get access to code
30:22 interpreter you need to have a chat GPT
30:24 Plus account
30:27 why is this not there we go
30:29 um you have to have a chat GPT Plus
30:31 account
30:32 and and then you'll get access to it now
30:35 if you have a GPT Plus account and don't
30:39 see it that means you haven't enabled it
30:41 so you have to go to your settings
30:44 and then in your settings there's this
30:46 thing called beta features
30:48 and turn them on
30:50 so you got plugins and you got code
30:52 interpreter and at some point they'll
30:53 bring bring back browse with Bing
30:57 I don't know what's taken them so long
30:59 but whatever
31:01 um just turn them on and then you will
31:03 see code interpreter and if you haven't
31:05 seen code interpreter let me show you
31:08 what it do it's amazing so normally in
31:11 chat gbt you just type your thing here
31:13 like write me a poem and it'll do that
31:15 what you can do with code interpreters
31:17 you hit plus upload a file
31:21 and then I forget what this was a data
31:24 set I grabbed last night I don't even
31:26 remember what it was
31:28 but it's called archive five zip
31:31 and it's uploading
31:34 I know
31:36 slowly
31:41 [Music]
31:41 um
31:43 and I can now go what
31:46 is in this file question mark
31:52 so it says the uploaded file appears to
31:55 be a zip archive to see what's inside
31:57 I'll have to unpack it let's do that and
31:59 then it fires off this plug-in looking
32:01 thing and if you look at what it did
32:04 it actually wrote python code it it
32:08 loaded some libraries wrote python code
32:10 to unzip that file unzipped it and then
32:13 it says oh this is crime data so it says
32:15 it it contains the following crime
32:18 offense and crime XML would you like to
32:21 preview the content of these files to or
32:23 or perform a specific let's say preview
32:28 um preview what's here
32:32 and so again it's writing python code to
32:35 read those files
32:37 and give us a little example of what's
32:39 there you don't have to look at this
32:41 code this is all just happening in the
32:42 background
32:43 so there's there's what the files
32:45 returned
32:46 so it seems like there was an issue
32:48 reading one of the files because it was
32:50 encoded weirdly
32:52 so now it's going to try another way of
32:54 encoding it or reading the encoding
33:01 uh
33:09 looks like it hung
33:11 so it's probably going to apologize and
33:12 try something else and that's another
33:14 thing that's wild about this is
33:16 if it if it fails it
33:20 it
33:21 um it'll it'll try again
33:25 and it apologizes
33:31 um
33:34 I'm gonna say stop generating let me say
33:36 um please
33:38 visualize VIs
33:41 you are lies
33:44 the interesting
33:47 um
33:48 Trends in this data
33:52 something went wrong regenerate response
33:55 something went wrong okay I gotta reload
33:56 it
34:00 you failed I'm sorry you failed
34:05 um
34:06 if you're sitting there thinking well I
34:07 thought code interpreter was supposed to
34:09 be so cool why did it just poop out on
34:12 you because all of this technology right
34:14 now is early
34:16 it's just a sign that we're early if it
34:18 were perfect and easy to use then we
34:21 wouldn't be at this phase of the uh
34:23 the the beginning of of a remarkable
34:26 transformation in all of humanity
34:30 um so let's see
34:31 um please
34:35 visualize the
34:41 interesting Trends in the data here
34:48 okay
34:50 let's see what it does first let's unzip
34:52 it
34:57 foreign
35:02 [Music]
35:13 seems like there was an encoding problem
35:15 so now it's going to try a different
35:17 encoding
35:22 thank you for the hearts
35:25 all right so now there's the data
35:27 incidents Offense Code
35:30 offense type offense category first to
35:32 current state last to current state
35:34 incident address
35:36 Geographic coordinates of the incident
35:39 neighborhood so there's a bunch of
35:41 interesting stuff there so
35:43 it's unzip the file figured out the
35:46 encoding of the CSV files
35:49 read the data
35:53 now it's reading the data of another CSV
35:55 file
35:58 so it says before analyzing the data
36:00 let's take a quick look at the XML just
36:02 so we understand the hierarchical
36:03 structure
36:07 so it's got there's there's your
36:09 metadata from the XML
36:12 so now let's move forward with the
36:14 analysis of the CSV file based on the
36:16 contents there are several several
36:18 interesting Trends we might explore
36:20 number of incidents over time
36:22 distribution event uh offense by type or
36:25 category distribution of incidents by
36:28 geographic area
36:30 number of victims per offense so let's
36:32 do
36:34 um
36:36 a beautiful
36:40 um visualization
36:44 around
36:45 three
36:48 so it knows what I meant by three it
36:50 gave me four options and I just picked
36:52 the third one so it says sure let's do
36:54 it by geographical area we can start
36:56 with distribution of incidents by
36:59 District
37:00 let's create a bar chart
37:04 incidents for each district
37:11 so that's boring
37:14 can you make it a heat map
37:21 um on a
37:25 geographic map
37:29 let's see if we can figure that out
37:32 unfortunately I'm not able to generate a
37:35 geographic map that's not always true oh
37:38 because lack of the internet access but
37:40 there it tells me the code I could do if
37:42 I wanted to do it myself
37:46 um
37:49 let's see can you do some other more
37:54 visually
37:56 interesting
38:00 um
38:02 look I don't know
38:05 sure I understand you're looking for
38:07 something more visually interesting
38:10 how about we explore temporal Trends in
38:12 the data we can visualize the number of
38:14 crimes
38:16 I'm just going to say yes
38:19 see what it comes up with
38:31 to do this we'll need to convert the
38:32 first occurrence date to a date time
38:35 format then extract the hour for each
38:38 time stamp
38:39 so now it's doing some full-on you know
38:42 data work some data cleansing and
38:45 and reformatting
38:56 we're waiting we're waiting if it times
38:59 out it'll it'll often give itself a new
39:01 task too
39:08 [Music]
39:10 I have no idea how much data's in here I
39:12 think it's a fair amount of data
39:15 now what this thing's actually doing is
39:17 when you start a chat instance with code
39:20 interpreter it's actually creating a
39:22 virtual machine so this is like our own
39:24 little stand-up computer or like you
39:26 know virtual computer that is you know
39:29 setting up all of its RAM and its places
39:32 to write data and it's installing Python
39:35 and installing the libraries and then
39:36 running all that stuff within that
39:39 virtual machine
39:41 so that's number of incidents by hour of
39:44 the day I mean that's that's fine it's
39:46 it's not all that interesting but but
39:48 anyway we could go on but that's that's
39:50 what code interpreter does that's how
39:52 you find it so hopefully that gives you
39:54 what you need
39:58 [Music]
40:00 d d
40:03 um I'm parsing it out in chat GPT
40:05 chapter by chapter and it's it's keeping
40:08 up so far
40:09 um the only thing to be careful with if
40:11 you're doing it chapter by chapter is
40:13 just be cognizant of the context window
40:16 it's only 3 000 words
40:18 so
40:20 um it may forget your earlier chapters
40:21 so what you might want to do if you
40:24 haven't already done this is for each
40:26 chapter have it summarize the chapter
40:28 so that
40:30 um you've got
40:31 a shorter representation of what's
40:34 happening in the chapters leading up to
40:35 the current one you're on that'll keep
40:38 as much as possible in the in the
40:39 context window the other thing you could
40:41 try is claude.ai has a hundred thousand
40:45 token context window so you could pop
40:48 all of your writing in that and then
40:50 just start interacting with it that's
40:52 another option
40:54 um I don't think it's as good as
40:58 gpt4 but it's pretty flipping good
41:02 I feel the same way I can't keep up I
41:04 know same here
41:06 can't keep up at all it's just there's
41:09 too much there's too much Buck you know
41:11 you know it's like it's like I told you
41:13 there's just a lot there's a lot there's
41:15 a lot to deal with it keeps coming at
41:17 you it keeps coming at you
41:19 you see it's a lot yeah
41:25 that's champ
41:28 quick question can turn it in detect AI
41:35 is turn it in homework
41:37 like plagiarism software
41:40 probably
41:42 um if you're University or your high
41:45 school is using
41:46 AI detecting software
41:49 I would
41:50 I would transfer
41:52 um
41:54 we should be past the point
41:56 at this point where universities are
41:59 saying you know
42:01 we're just going to use AI detector
42:03 software they should be engaging with a
42:05 Ai and teaching their students how to
42:07 deal with what's coming
42:09 um they should be teaching them critical
42:11 thinking skills and understanding how
42:13 this [ __ ] works and what the
42:14 implications are but that's a whole
42:16 separate conversation
42:19 love your haircut thanks I don't know
42:21 that this is a haircut so much as a
42:23 like three toupees got in a fight and
42:26 this one lost
42:28 that's just really bad
42:32 but thank you
42:35 I'm just here for the shirts oh yeah
42:37 uh-huh
42:40 this shirt is
42:42 you can almost read the text in front of
42:44 it it's a little busy there's a lot
42:45 going on here it's it's a little hard to
42:47 read it hurts the eyes but thank you any
42:49 AI crypto making tokenism type apps
42:55 you can I mean chatgpt does Smart
42:58 contracts you could just copy your
42:59 favorite smart contract paste it in
43:01 there and say write me a new smart
43:03 contract and give it whatever you want
43:05 it to be and it'll do it
43:11 um probably go to futurepedia.io there's
43:13 probably some crypto Bros have figured
43:15 out a way to
43:16 put a front end on front of it in front
43:19 of it but what I would do I mean
43:20 seriously what I would do is it's good
43:22 enough with with coding and with with a
43:25 lot of the the smart contract stuff that
43:28 I would probably just figure out how to
43:29 do it in chat GPT just go learn and
43:32 maybe you start a business right maybe
43:33 that becomes a business you know
43:35 click here to build an nft collection
43:39 here's your images here's your this
43:40 here's your smart contract here's you
43:42 this do you want to deploy it
43:45 start a Discord server and get ready to
43:47 rock your followers
43:49 it's a web 3 tradition you can now rug
43:52 people without even knowing code
43:54 [Laughter]
43:57 oh web3
44:02 time was huh
44:04 showed such a promise
44:06 web3's got three years of wound looking
44:09 well-deserved wound looking but I think
44:12 AI is going to drive adoption of web3
44:15 in the end because I think it's actually
44:17 really important technology it just it
44:20 just had the wrong people at the helm
44:22 for a while there
44:26 all right I'm trying to figure out where
44:28 I was Champion stop it
44:32 hey
44:36 what'd you do tonight I don't know I
44:38 watched some old man yell at his dog for
44:39 an hour
44:40 why would you do that I had no idea
44:44 oh that's way old questions hold please
44:50 I lost my place I lost my place
44:55 it's hard you don't know how hard it is
44:59 to
45:01 yesterday laughing my ass off because
45:04 why
45:08 you don't think they have a monopoly on
45:09 this AI [ __ ] tell me someone who
45:12 can compete with them right now all of
45:15 the software infrastructures based on
45:17 their um
45:19 based on making calls to their chips
45:22 no one else seems to be doing it maybe
45:24 Intel
45:26 gpus but
45:29 maybe AMD
45:32 go listen to that if you want to hear if
45:35 you want to hear a really interesting
45:36 take on that stuff go listen to The Lex
45:38 Friedman George hotz interview from
45:41 three days ago
45:42 and Listen to George Hots talk about you
45:45 know
45:46 he almost bailed on AMD and then they're
45:48 trying to get their [ __ ] together and
45:50 then he almost went to Intel and didn't
45:52 want to do that but he doesn't want
45:54 Nvidia at the top of the heat but he's
45:56 like who's gonna compete check EBT novel
46:00 analysis was pretty impressive it picked
46:02 up on what I intended to do thematically
46:03 that's cool
46:05 from a fiction narrative perspective but
46:07 it's hard to trust
46:09 at first it responded like I'm a great
46:12 writer when I that's interesting when I
46:14 asked if there were problems it found
46:16 problems and contradictions to previous
46:19 analysis
46:21 and after several chapters it crashed
46:23 several times
46:24 but the basic analysis was at least on
46:27 point with what I intended that's pretty
46:29 cool hey Troy Decker it is Kyle that's
46:33 me
46:34 I'm Kyle
46:37 um yeah the The Narrative analysis thing
46:40 having having chat GPT be like your your
46:43 writing coach and your writing partner
46:45 is a really interesting idea
46:49 that would be that would
46:50 business to build is
46:58 you could you could almost build a tool
47:00 that works in stages right there's the
47:03 ideation stage there's the blocking and
47:05 tackling rough draft stage and then
47:07 there's different levels of refinement
47:10 and if you could build tools that make
47:12 it really easy for writers to navigate
47:14 those stages of of writing a novel where
47:18 they're still in charge but you're kind
47:19 of supporting them and doing it in the
47:21 sort of pre-structured way with a really
47:24 slick interface
47:25 that's a business
47:28 now what's the downside of that business
47:31 writers are broke who's gonna [ __ ]
47:33 pay for it
47:36 so it's not a business it's a good idea
47:40 one of the things you should always be
47:42 asking yourself because there might be a
47:44 brilliant idea and people might tell you
47:46 holy crap yeah that's awesome I'd use
47:49 that
47:50 don't mistake that for here's a check
47:54 that's what you need to figure out is
47:57 build a something that's a good enough
47:58 solution someone's willing to write a
47:59 check for it
48:02 I'm sorry that's very Gen X that
48:04 someone's willing to venmo you money for
48:06 it or give you six diamonds
48:08 [Laughter]
48:10 hey hey Kyle you're like a fine wine
48:12 better with age Gen X yeah thanks Troy
48:16 in the either
48:18 have you heard of road map to riches I
48:22 have not I will check it out though my
48:24 novel is about 120 000 words oh
48:26 interesting
48:27 so yeah that's a that's a big fatty of a
48:32 novel
48:33 does it need to be 120 000 words if it
48:36 does it does
48:37 but that could be an interesting
48:38 exercise is to see if see see how chat
48:41 GPT might might
48:43 um
48:45 how it would suggest tightening that if
48:47 you want to tighten it but if it's like
48:49 a big epic story then maybe it needs to
48:51 be that big
48:52 I don't [ __ ] know I don't read so
48:56 80d 80d in reading for me they do not go
48:59 well together I want chat GPT to rewrite
49:02 and design my website how would I go
49:04 about it
49:05 um
49:07 did you
49:09 I don't know there's a lot of ways to do
49:11 it uh
49:14 I don't know enough about that world I
49:16 mean I would go to futurepedia.io and
49:18 look at some of the tools that people
49:19 have already created for AI assisted web
49:23 you know design and redesign
49:25 and design analysis I'm sure there's
49:27 some tools there that do it
49:30 um
49:31 you can also just try taking chunks of
49:35 your code pasting it into chatgpt and
49:38 say you know tell me about this website
49:40 and then describe how you want it
49:42 changed and just sort of go back and
49:44 forth copying and pasting code like
49:47 describe how you want things changed go
49:48 back and see how it works on the site
49:50 you might do that if you're doing that
49:52 or just use one of the no code tools you
49:55 know Squarespace or
49:56 or or webflow or whatever
49:59 he's so polite to the robot I am I do
50:02 say please and thank you because quite
50:03 frankly the robot's polite to me and it
50:05 and it kind of just triggers
50:08 it's less like I'm prompting a robot and
50:11 more like I'm having a conversation with
50:12 a remote co-worker
50:14 and so I do I find myself I I think I've
50:17 gotten over this part of it but this has
50:20 taken me five or six months is
50:23 like if it would be getting things wrong
50:26 or I just have a new idea and I'm like
50:27 oh I want you to rewrite this oh rewrite
50:29 it again rewrite it again about the
50:31 third time I'd say rewrite it again I
50:33 would start to feel guilty like maybe
50:36 I'm asking it to do do too much and
50:39 they're like oh no it's a robot you
50:41 [ __ ] idiot
50:43 um but it's but it's like a there's
50:45 something about because it feels like
50:47 you're interacting with a person you're
50:49 having like this back and forth
50:50 collaborative thing
50:53 that experience I have with a real human
50:55 like about the third time I send
50:58 something back for a rewrite
51:00 I just go oh [ __ ] it I'll just do it
51:02 myself it'll be faster for me to just do
51:04 it myself and and that's you know just
51:07 me being a spineless ninny and not
51:09 wanting to offend someone but whatever
51:14 um he's Apocrypha I don't know what that
51:17 Rachel good word good Lord
51:21 this is this is the AI learning lab
51:24 apparently this is the AI this is the
51:26 vocabulary learning lab as well
51:28 apocrypha
51:31 that seems like what I should be able to
51:34 figure out from the Greek
51:37 a collection of writings that are not
51:40 included in the standard Canon of
51:42 religious tradition wow
51:45 fantastic
51:47 that is
51:48 I I
51:50 it is worth noting the term profit can
51:52 also be used more broadly to refer to
51:54 any non-canonical or disputed writings
51:57 okay I I I don't I I like it I like it I
52:01 like it okay
52:04 is your GPT on a local machine no I
52:06 haven't done that
52:08 oh he is Apocrypha Apocrypha is a person
52:11 in here
52:12 so Rachel's referring to Apocrypha not
52:15 saying I am Apocrypha but you're saying
52:17 I am whatever she was asking her he was
52:19 asking before but now we know what
52:21 Apocrypha is that's cool am I saying
52:24 that right
52:26 it's got to be apocrypha okay agree all
52:30 right oh crap Tick Tock just jumped me
52:32 to the bottom dang it tick tock
52:36 love your lives bro thank you meta
52:38 Victory I appreciate that
52:41 um
52:42 I feel like like I'm saying nothing
52:45 today I feel like I'm just rambling
52:47 tonight so if I am just tell me I'm
52:51 wasting your time
52:57 [Music]
53:00 novel analysis was pretty impressive
53:03 Oh wait we're already here oh Tick Tock
53:06 man it's just jumping me all over the
53:07 place what's happening look at tiny
53:10 bucks
53:11 do you think Universal income is
53:13 inevitable okay is your GPT on a local
53:15 machine no my GPT is not a local machine
53:18 um
53:19 I will
53:21 I'll probably do a local llm install
53:26 when they get better and smaller and uh
53:31 more stable they're just
53:32 they're just evolving too fast right now
53:36 um although
53:38 I met with our Tech team today and we've
53:40 got an instance of Lang chain up and
53:42 running so that we can start running
53:44 um Lang chain prompts on all of the
53:46 transcriptions in our videos
53:48 um so I sort of have a local machine
53:50 running but not on like my local
53:52 personal machine
53:54 so I do for work do you think Universal
53:56 income is inevitable
53:59 um
54:00 I think something is going to be
54:02 necessary I do I think I think
54:08 maybe a year from now
54:10 yeah probably about a year from now
54:12 the disruption is is the disruption from
54:16 AI on the workforce is going to be
54:18 significant enough that
54:21 um
54:23 I don't think I don't know my sense is
54:26 it won't be as dramatic as kovid you
54:28 know covid we just sort of like we were
54:30 just like okay
54:32 tomorrow don't leave the house right it
54:35 was just like bang and it was Global it
54:37 was [ __ ] surreal
54:39 I don't think it's gonna be like that
54:41 but I do think that there's going to be
54:43 a level of disruption that's bad enough
54:46 that it's going to have to be dealt with
54:48 and I don't think companies are going to
54:51 deal with it because they're just money
54:53 machines and they'll be like well we're
54:55 generating the same amount of product
54:57 with a third of the people awesome
55:00 profit Bob
55:02 you going golfing this weekend no I'll
55:04 be on the yacht sorry
55:07 um so I think they'll be doing that and
55:09 then I think the workers are gonna the X
55:11 workers are gonna be
55:13 um you know increasingly pissed off and
55:15 freaked out and so I think something's
55:17 gonna have to happen
55:18 hopefully
55:21 it doesn't get to the point where people
55:23 are freaking out and they have to do
55:24 something hopefully they do something
55:26 proactively but we'll see like like it
55:29 could be you know it could be Universal
55:32 income or it could be incentives to
55:35 companies to not lay off people like if
55:39 they're seeing you know increased
55:41 productivity with AI to maintain their
55:44 current employer levels there were a lot
55:46 of those incentives with kovid you know
55:48 don't lay off your people so I could see
55:49 something like that but I could also see
55:51 there being some sort of universal
55:53 income I don't think it's going to be
55:55 necessary to maintain it for maybe I
55:59 don't know more than two or three years
56:02 but I don't know I don't know look at
56:05 tiny bucks Tiny Box is really
56:06 interesting that's Tiny Box is George
56:08 hotza's new company he's creating or
56:11 it's called tiny Corp and then the the
56:13 computer that he's building is called
56:15 Tiny Box and it's a fifteen thousand
56:17 dollar six GPU AI monster machine that
56:21 basically what he wants to do is he
56:23 wants to make the most powerful AI
56:27 super computer that you can just plug
56:28 into a regular wall outlet and not have
56:31 to run special power to it and not have
56:33 to run special cooling to it and not
56:35 have it be so noisy that you can't work
56:36 next to it but it's going to be like you
56:39 know some ridiculous amount of flops
56:41 that you can do you can run a 65 billion
56:44 parameter large language model on it
56:49 he's pretty brilliant brilliant dude
56:52 that Duty God
56:54 these guys looking Marvel's going on in
56:56 the old uh Gumball Factory if you know
56:57 what I mean
57:00 all right what I liked was when it got
57:03 critical about Pros it wrote a
57:06 disclaimer stating that it was a
57:08 subjective analysis that's interesting
57:09 silver fox
57:11 do you know how long code interpreter
57:13 keeps your project results to refer back
57:15 to not long like this is one of the okay
57:18 so here's a real frustrating thing about
57:19 code interpreter right now I don't have
57:21 my little URLs up do I
57:24 um code interpreter like if you go away
57:26 from it for like 30 minutes and
57:27 sometimes it's faster like I can't tell
57:29 when it when it shuts it down
57:32 so so you upload some some stuff you run
57:36 some data analysis on it it makes you
57:37 all the charts and graphs if you go back
57:39 to it like half an hour later it says
57:41 you know this this is gone
57:44 and then it it just everything's gone
57:46 and you can redo it but it's it's gonna
57:48 write it's gonna you're gonna have to
57:50 start from scratch
57:51 so a thing you're gonna have to do with
57:53 code interpreter short term because
57:55 they'll fix this is
57:57 um once you get results that you like
57:59 migrate that stuff into other documents
58:02 um that's it's a really good question
58:03 it's a problem so so you just need to
58:06 just need to keep on top of it just know
58:08 that whatever's in your active session
58:10 anything you want to keep get it out of
58:12 there
58:13 um I'm in the road map to Rich's
58:15 Marketing Group you would be an asset oh
58:17 cool
58:18 um let me go so do me a favor just
58:20 Rachel just send me a DM
58:22 and I will uh I'll check it out for sure
58:34 he's Apocrypha that's really funny that
58:36 I thought you were calling me Apocrypha
58:38 I'm like I'm a
58:41 lost book of the Bible
58:47 can Chachi bt4 interact with Google Maps
58:51 or other map apps let's go look at
58:54 plugins there might be I think there's
58:56 some map plugins so let's go I'll show
58:59 you some
58:59 [Music]
59:00 um
59:05 uh rut row
59:09 I wonder if plugins are turned off
59:11 oh your session is expired no
59:16 [Music]
59:21 name that comedian
59:23 come on you gen xers
59:26 no plugins enabled
59:31 um
59:34 plug-ins down I think plugins is down
59:38 reload
59:42 plugins okay they're back
59:45 okay never mind false alarm don't go
59:49 posting that on Twitter they're back
59:52 okay never mind false alarm don't go
59:55 posting that on Twitter AI learning lab
59:58 said that openai canceled plug-ins
1:00:01 all right
1:00:03 Maps
1:00:08 Okay maps map app takes latitude
1:00:11 longitude and latitude coordinates and
1:00:14 generates a map
1:00:16 Tom Tom explores Maps gets traffic
1:00:18 updates
1:00:20 how many of these oh there's only eight
1:00:22 of them
1:00:23 Whimsical diagrams mind maps
1:00:28 mixerbox powered by Google Maps API
1:00:31 mixer box chat map is the world's first
1:00:35 AI chat box for Matt chat bot for maps
1:00:39 powered by Google Maps API mixer box
1:00:43 chat map is the world's first AI chat
1:00:47 bot for maps
1:00:48 okay
1:00:49 for the marketers in the bit in in the
1:00:52 audience
1:00:53 if you ever write a sentence like that
1:00:56 and and actually put that out to the
1:00:58 public you should be ashamed of yourself
1:01:02 you you you cannot physically say this
1:01:04 sentence much less understand it powered
1:01:07 by Google Maps API mixer box chat map is
1:01:11 the world's first AI chat bot for maps
1:01:16 if you write something like that have
1:01:18 chat GPT rewrite it but let's try this
1:01:21 and see if it works okay so I'm going to
1:01:22 install mixerbox
1:01:24 mixer box chat map
1:01:27 if you can't say it too if you can't say
1:01:30 it say it one time fast it's too
1:01:32 complicated
1:01:35 I'm turning off instacart so we only
1:01:38 have mixer bot chat Maps I can't say
1:01:43 that's [ __ ] hilarious okay
1:01:45 um let's see
1:01:47 um
1:01:49 like
1:01:51 um
1:01:54 I'm trying to think I don't I'm not
1:01:56 connected to the internet so let's just
1:01:57 say
1:01:58 um
1:01:59 show me
1:02:01 um
1:02:05 the water around Coney Island
1:02:14 so this should fire off the plug-in it
1:02:17 did
1:02:20 here's some water related places around
1:02:23 Coney Island New York Aquarium no no no
1:02:26 right
1:02:29 uh no no I want to see
1:02:34 um the
1:02:36 Hudson
1:02:42 River
1:02:45 on the map near Coney Island
1:02:51 see if we can get it to actually
1:02:53 generate some map things but what what
1:02:55 this is telling me is
1:02:57 um
1:02:59 so there's Coney Island
1:03:07 Coney Island Creek in Brooklyn oh that's
1:03:10 that's an interesting one let's go look
1:03:12 at that
1:03:15 all right so it's taking us here's Coney
1:03:17 Island Beach
1:03:19 and there's the creek I've fished out
1:03:21 here this is good
1:03:23 I've walked out here and fished that's
1:03:25 fun
1:03:26 um
1:03:30 I don't know what this does
1:03:34 Beyond
1:03:36 that you couldn't do on Google Maps
1:03:38 right so
1:03:40 um like what is what it says to me is
1:03:42 that the
1:03:44 the fact that there's only five plugins
1:03:47 that have maps and one of them isn't
1:03:49 even map related is that this is
1:03:51 probably the this is probably something
1:03:53 that's coming well here's let's go let's
1:03:56 go try another thing
1:03:57 um
1:03:59 let's go try Bing because Bing's
1:04:01 connected to the internet
1:04:04 and Bing is also a search engine with a
1:04:07 map
1:04:08 so let's go to chat and then let's say
1:04:12 um
1:04:15 um show me the
1:04:18 let's see
1:04:20 um
1:04:20 overlay today's high and low
1:04:26 temps
1:04:29 um on a map
1:04:33 of Coney Island
1:04:42 so so I went to bing.com went to chat
1:04:45 and then just typed that in
1:04:48 I'm sorry I'm not able to Overlay this
1:04:50 over a map house can I help you
1:04:53 yeah so it looks like they're not really
1:04:55 dealing with maps right now so there's a
1:04:57 business there's something to play with
1:05:03 I'm not sure what you want to do with it
1:05:05 you know it might be that you know
1:05:07 there's a there's a map API for Google
1:05:09 so
1:05:10 I would think it would be relatively
1:05:12 trivial to build some automations or
1:05:15 some applications that do that
1:05:21 um you are invaluable thank you by the
1:05:23 way your idea too and then it cuts off
1:05:26 thank you Sean
1:05:28 Kyle hi quick thoughts on making a
1:05:30 choice on AI tools and how many to
1:05:32 concentrate on for beginners
1:05:34 awesome question and hello AI for
1:05:37 beginners
1:05:39 um why I put these urls
1:05:43 this is my answer
1:05:47 um and the only thing that I would add
1:05:49 to this
1:05:51 are the
1:05:52 [Music]
1:05:53 um
1:05:54 are the image generation tools but I
1:05:56 think I think as a foundation where
1:05:57 where to start is with chat gbt just the
1:06:01 official website if you can't afford
1:06:03 chat GPT plus then use Bing or Claude
1:06:08 now so it used to be Bing was the free
1:06:11 version of gpt4 but now Claude
1:06:15 which is a bunch of ex-openai people
1:06:17 now that they've launched this this is
1:06:20 Claude 2 with a hundred thousand token
1:06:22 context window which for beginners it
1:06:25 just basically means you can put a
1:06:26 shitload of words in it and it doesn't
1:06:27 break gpt4 still breaks you know after
1:06:31 3000 words or so it might be up to six
1:06:34 thousand words now but it's still not 75
1:06:38 000 words like Claude is
1:06:40 um
1:06:42 and then poe.com lets you play with a
1:06:45 bunch of different models you know three
1:06:47 from anthropic two from open Ai and one
1:06:50 from themselves
1:06:52 plus they've got bots so
1:06:55 as a beginner I would focus on learning
1:06:58 prompting and so prompts.chat that
1:07:01 document right there it's just a it's
1:07:02 just a a document called awesome chat
1:07:06 GPT prompts and it it's really just a a
1:07:09 single flavor of font of prompt which is
1:07:12 act as a you know Persona here
1:07:16 um so act as an accountant act as a
1:07:18 screenwriter act as a therapist act as a
1:07:21 coach act as a nutritionist but you can
1:07:25 really quickly get your head around
1:07:28 what prompting is how to do it and also
1:07:31 how incredibly [ __ ] profoundly deep
1:07:35 these tools are
1:07:37 this is not these are not fancy Google
1:07:40 searches so anyone beginning I would say
1:07:44 you know basically this is what I've
1:07:46 been saying in here you owe it to
1:07:48 yourself to play with these tools long
1:07:50 enough for you to have your Kevin
1:07:52 McAllister moment which is this moment
1:07:55 oh
1:07:57 Malone moment
1:07:59 where at some point you'll do something
1:08:01 with one of these and it'll just blow
1:08:02 your [ __ ] mind and and that's the
1:08:05 point at which you should start getting
1:08:07 that then you should start digging in
1:08:08 and then you can start to focus about
1:08:11 what tools you want to use and what
1:08:12 where you want to explore
1:08:14 on the image side of the house
1:08:17 you can do free images at bing.com you
1:08:19 just tell it to make an image and it
1:08:20 will that's using Dali I personally pay
1:08:23 for mid-journey because I think it's the
1:08:25 best
1:08:26 stable diffusion if you're geeky is has
1:08:29 the the by far the greatest flexibility
1:08:31 of the tools out there Adobe Firefly and
1:08:35 generative fill within Photoshop are
1:08:38 commercially safe because they train
1:08:40 their models on open sort or uh public
1:08:43 domain and Adobe stock libraries that
1:08:47 they have the rights to so if you don't
1:08:50 want to risk being in any sort of legal
1:08:53 Waters I would steer away from
1:08:56 mid-journey and stable diffusion
1:08:59 and probably even Dali and I would go
1:09:02 toward the Adobe products so there's
1:09:04 that's where I'd start and then if you
1:09:06 really just want to geek out
1:09:07 futurepedia.io that's just a big old
1:09:09 pile of tools so you can feel free to go
1:09:12 if you well what about I want to compose
1:09:14 music go there and find all the music
1:09:16 tools
1:09:18 um I accidentally hit send skip my
1:09:20 comment okay
1:09:25 Kyle you big fat oaf get a real haircut
1:09:29 and what's with the glasses yeah that's
1:09:32 okay that's fine that's fine whatever
1:09:34 you said I'm not I won't take it
1:09:35 personally I spent too much time on AI
1:09:37 lab my wife is getting jealous you think
1:09:39 your your wife is getting jealous my
1:09:41 wife's like hey what are you who are you
1:09:43 talking to why do you are you gonna go
1:09:45 do that again so yeah yeah I get it I
1:09:48 understand
1:09:49 uh I felt bad for a split second when I
1:09:52 kept having to add the letter
1:09:54 to add the letter I was writing I felt
1:09:57 bad for a split sec I don't know what
1:09:58 that's in reference
1:10:02 sorry all right we're building an AI
1:10:04 call center agents I'm just a dumb
1:10:06 entrepreneur no you're not
1:10:09 in my opinion this will displace 80 of
1:10:11 call center reps I saw I saw a really
1:10:14 unfortunate headline today
1:10:16 um it said Indian call center replaces
1:10:20 90
1:10:24 like I don't know whatever 94 96 like
1:10:27 fires 94 96 of its staff
1:10:31 and I was sort of imagining like a
1:10:33 hundred thousand person call center but
1:10:35 it was it was a 22 person company
1:10:37 um but yeah I agree I think that um I
1:10:40 think call centers are going to be
1:10:41 replaced and I think from a an end user
1:10:45 experience standpoint they should go
1:10:48 because here's the deal I want
1:10:52 I want to be able to call a company
1:10:55 and rather than sitting on hold for 45
1:10:58 minutes talk to a synthesized bot
1:11:01 that knows everything about the company
1:11:05 can answer all of my questions has the
1:11:07 authority to do some level of [ __ ]
1:11:10 and is smart enough to know that if my
1:11:12 problem is non-standard and needs to be
1:11:16 dealt with by a human it immediately
1:11:18 connects me to a human
1:11:21 that means you could get rid of 80 to 90
1:11:24 percent of call setters take the 10 or
1:11:27 20 percent of the people that are left
1:11:29 upskill them into high level customer
1:11:33 service people that have the authority
1:11:35 to [ __ ] fix problems
1:11:38 that that are escalated let the
1:11:42 llm-powered chat bot talk to everyone
1:11:45 answer the phone immediately
1:11:48 detect from their area code what part of
1:11:51 the world they're in and have the chat
1:11:53 bot you know respond to them in an
1:11:55 appropriate dialect so it sounds like
1:11:57 them have it know all the [ __ ]
1:12:01 have the conversation have it record the
1:12:03 conversation so that if it does escalate
1:12:05 it to a human a nice little report is
1:12:08 given to that human here's what this
1:12:10 person's dealing with here's why I
1:12:11 escalated it to you
1:12:14 you don't have to be that's that's the
1:12:16 only transfer you do it's a two transfer
1:12:18 call center and there's no phone trees
1:12:21 that's what I want go build that they're
1:12:24 they're second business of the night
1:12:26 second business business of the night
1:12:28 that needs to be built
1:12:31 we're going to transform customer
1:12:33 service to actually not suck again
1:12:40 thank you I got I got one hard on that
1:12:44 I don't know I seemed really upset about
1:12:46 call centers I don't know
1:12:48 I don't think he likes India
1:12:50 say it ain't India that's not the
1:12:51 problem
1:12:53 the problem is
1:12:56 Scripts
1:12:57 and phone trees
1:12:59 and bad shitty over complicated logic
1:13:02 and people that are paid below minimum
1:13:06 wage
1:13:07 that are not allowed to actually have a
1:13:11 conversation they can only read the
1:13:12 script
1:13:14 that's the problem
1:13:18 you can flush that entire [ __ ] system
1:13:21 down the toilet and it and it will
1:13:23 improve dramatically
1:13:24 Andy Kaufman hey Ellie come to save that
1:13:27 day
1:13:30 there's a jet some gen x's in the house
1:13:32 oh yeah Wolfman
1:13:37 all right
1:13:45 yeah Sean McCracken I think that's
1:13:47 really interesting I I would be curious
1:13:49 to you know keep me posted on how that
1:13:51 evolves because I think that's uh
1:13:53 that's a big deal
1:13:55 um do me a favor if you would like uh
1:13:57 tap the tap the screen make the hearts
1:13:59 go and if you would share the live so we
1:14:01 can see if we can get some more folks in
1:14:03 here
1:14:04 Silver Fox for code interpreter I've
1:14:06 been saving the code when I have my
1:14:07 results oh that's interesting
1:14:10 Underdog it's Mighty Mouse
1:14:14 is copy and paste the only way to save
1:14:17 migrate a chat GPT history no
1:14:22 um if you go to
1:14:27 go to your settings
1:14:30 and then go to what's it called Data
1:14:34 data controls
1:14:35 and there's an export data
1:14:38 button
1:14:40 pop that button
1:14:43 now
1:14:45 to Silver Fox it was a silver fox yeah
1:14:47 the silver Fox's Point
1:14:49 um
1:14:50 I don't know that the code the code
1:14:52 interpreter writes will get exported
1:14:54 with that if your virtual machine has
1:14:56 been shut down
1:14:58 so
1:14:59 I would say for code interpreter if you
1:15:02 want the code that it wrote to do to
1:15:05 generate whatever charts and graphs you
1:15:07 like
1:15:08 um
1:15:09 that you might want to copy and paste
1:15:11 but that's worth an experiment
1:15:18 I had ice cream tonight because
1:15:21 because because I'm a bad bad keto
1:15:24 person
1:15:25 I'm doing okay on the intermittent
1:15:27 fasting but I'm doing bad on the Keta on
1:15:30 the non-uh carb part of it
1:15:36 I've lost a little bit of weight this
1:15:38 shirt didn't fit at all
1:15:40 um
1:15:40 I don't know three months ago
1:15:44 so
1:15:46 I'm attending a meeting about llms at my
1:15:48 Corp today wow that's interesting
1:15:51 I'm in their chat group and learning
1:15:53 group lab that's cool man
1:15:56 is it is it a big company that you work
1:15:59 at cabrino and by the way do you say
1:16:01 your name cavruno I never know how to
1:16:03 say your name it's it's like Bruno KV
1:16:07 Runo k k Bruno Bruno I don't know anyway
1:16:11 you could change it so that it's like
1:16:14 sayable
1:16:16 oh
1:16:19 you'd be like llm Mama or
1:16:21 [Laughter]
1:16:26 what did I do last night I misspelled AI
1:16:29 learning lab when I made a bot so I made
1:16:32 a bot on Poe called AI leaning dude
1:16:35 which is which is not bad
1:16:38 is gmaps plug-in wait is gmaps plug-in a
1:16:42 publisher AKA marketer in my opinion for
1:16:45 now
1:16:46 a lot of the whoa what's that Sean that
1:16:50 was cool I've never seen that one I got
1:16:53 the cowboy hat and a little heart every
1:16:55 now and again but the uh oh
1:16:58 that's cool
1:16:59 uh
1:17:00 is gmaps plug it wait gmaps plugin is a
1:17:04 publisher AK marketer in my opinion for
1:17:06 now a lot of the plugins right now a lot
1:17:10 of the plugins are
1:17:12 um
1:17:13 just shitty ass ads for websites
1:17:19 come on focus focus you can do it you
1:17:22 can do it come on
1:17:24 there we go
1:17:26 all right here that generative AI could
1:17:28 add 2.6 to 4.4 trillion to the economy
1:17:31 yeah I think the I think the um the the
1:17:34 long-term prosperity
1:17:38 and and that's low like I that that was
1:17:41 a reporter it was an article that that
1:17:43 came out like what 10 days ago or
1:17:45 something like that Bruno damn dang your
1:17:48 name
1:17:48 um
1:17:50 and and what it was talking about it was
1:17:52 kind of comparing the early predictions
1:17:54 of what the World Wide Web would add to
1:17:56 the economy and what it actually added
1:17:58 which was you know significantly higher
1:18:00 and and then you know they're saying
1:18:02 yeah 2.6 to 4.4 trillion
1:18:05 um I think it's got the chance of of
1:18:07 contributing you know
1:18:09 see even more than that like a
1:18:11 significant amount like we could exceed
1:18:13 that
1:18:14 um again I you know I'm not an economist
1:18:17 but I I
1:18:18 just in in my simpleton mind
1:18:22 when I when I hear anecdotal evidence of
1:18:25 when uh when a tech firm goes all in on
1:18:29 code on uh GitHub co-pilot when they go
1:18:32 all in on AI pair programming they're
1:18:34 seeing a 30 to 40 percent increase in
1:18:37 productivity
1:18:38 so so if you take that as an average
1:18:40 take the low end of that let's say
1:18:42 there's a 25 increase in productivity
1:18:44 thanks to these tools and let's say
1:18:46 these tools you know get spread across
1:18:48 all the stuff that we do
1:18:50 so you've got this amount of people
1:18:51 working producing this amount of economy
1:18:54 and
1:18:56 with a 25 increase like that's a lot
1:19:00 that's a lot of increase when you know
1:19:02 when we're measuring things in you know
1:19:04 half a percent one percent two percent
1:19:06 if you see a 25 increase that's a big
1:19:08 gap that gets that gets created there
1:19:11 and then you know what do you do with
1:19:12 that Gap is interesting do you maintain
1:19:14 the economy and lower the workforce
1:19:17 right that's where you need Universal
1:19:18 basic income or do you maintain the
1:19:20 workforce and dramatically improve the
1:19:22 prosperity and if you can do that in a
1:19:26 way
1:19:26 that the billionaires don't just buy a
1:19:29 second and third yacht but do that in a
1:19:31 way where that money can go to deal with
1:19:34 the displaced and upskill the displaced
1:19:36 then then we potentially start to to
1:19:38 move into an economy that that is is
1:19:41 quite prosperous and that's that's a
1:19:43 very real possibility
1:19:45 um I want route optimization input
1:19:48 several addresses and it gives me gives
1:19:50 you the most efficient driving I think
1:19:53 there's some that do that I think um
1:19:57 is it Bing maps that does that I know
1:19:59 Waze does that
1:20:02 you can do it within Google Maps but
1:20:04 it's really uh
1:20:08 convoluted it's the interface for it's
1:20:10 really shitty
1:20:11 um
1:20:12 what would be worth trying is is go to
1:20:15 chatgpt and have it tell you which
1:20:18 mapping tools allow you to do that
1:20:20 because I I know there's I I oh what
1:20:23 it was uh
1:20:26 MapQuest MapQuest had a pretty good
1:20:28 version of that I don't know if MapQuest
1:20:30 is still around they were big for a
1:20:32 while
1:20:33 trading in llm is terrible CO2 emissions
1:20:35 equivalent to 60 or 70 cars yeah I I I
1:20:39 agree with that although the if you're
1:20:42 looking at what's going on with how
1:20:43 they're training the smaller large
1:20:45 language models in the open source
1:20:47 Community where it's costing them like
1:20:48 100 bucks in training to get it nearly
1:20:51 to GPT 3.5 performance
1:20:54 um I think I think the optimization of
1:20:56 these large language models for the for
1:20:58 the non-big ones right I think I think
1:21:01 open AI
1:21:03 inflection
1:21:06 um Google meta
1:21:09 you're gonna have big players that are
1:21:11 going to spend Millions on you know GPU
1:21:15 centers to train the massive data sets
1:21:17 but I think the efficiencies and the the
1:21:20 open source things are going to get
1:21:22 increasingly powerful in increasingly
1:21:24 smaller packages so I think there's
1:21:26 going to be a balance and equilibrium
1:21:28 that's hit there where I I don't think
1:21:30 it's going to be runaway energy costs I
1:21:33 hope but I don't know
1:21:35 did you see the uh the uh the Google
1:21:38 Sycamore
1:21:40 um
1:21:41 um quantum computer now now I don't
1:21:43 think this is related to AI because it's
1:21:45 a different kind of compute but maybe it
1:21:47 is I don't I don't [ __ ] know
1:21:49 um but the the Sycamore so so Google has
1:21:53 a 70 Cubit quantum computer and they
1:21:57 just did a benchmark that it it it
1:21:59 calculated in six seconds what it would
1:22:03 take today's state-of-the-art super
1:22:05 computer 47 years to compute
1:22:09 six seconds
1:22:10 versus 47 years it's like [ __ ]
1:22:14 so
1:22:16 um I have no idea what's going to happen
1:22:18 with that Quantum stuff I mean other
1:22:20 than you know all of your high security
1:22:22 passwords are now compromised that
1:22:26 shit's gone
1:22:28 anyone can get into any system at any
1:22:30 time uh when you have that level of
1:22:32 compute going on
1:22:36 luckily those things are still expensive
1:22:40 um oh it just jumped me around again
1:22:42 dang it dang dang it uh Portland is
1:22:46 homeless due to call center decline in
1:22:48 automation that's interesting
1:22:51 listen convert to text query response
1:22:54 text to speech play
1:22:57 response audio already running oh that's
1:23:00 interesting Sean yeah I'd be interested
1:23:02 to learn more about that because I I
1:23:03 mean I really do think like that's
1:23:05 that's where it's got to go like like
1:23:07 just just if you think in the end the
1:23:10 inefficiency of
1:23:13 the like the the incredible amount of
1:23:16 time that must go into programming phone
1:23:18 trees like I've got a small company and
1:23:21 programming our little voicemail thing
1:23:22 was an abject nightmare I can't even
1:23:24 imagine doing it for a bigger company so
1:23:27 so just just the labor alone to program
1:23:30 call center tree logic is
1:23:35 like that's a whole industry that that
1:23:37 you know
1:23:39 needs to go away it's it's a horrible
1:23:42 experience for
1:23:44 the end User it's really expensive it's
1:23:47 really time consuming
1:23:49 um it's not necessarily better for the
1:23:51 company now is that bad for all of the
1:23:53 companies that have spent decades
1:23:57 building phone tree logic programming
1:23:59 companies yeah it's [ __ ] devastating
1:24:03 but which of those can pivot into this
1:24:06 new world because maybe they become
1:24:07 experts at creating the kind of system
1:24:10 that you're talking about Sean
1:24:14 oh Chris Bruno
1:24:17 largest private company in the world
1:24:19 interesting pretty big 155 000 employees
1:24:22 wow that's really uh I'm curious what
1:24:24 sector
1:24:26 um okay I'll call you Chris from now on
1:24:27 because crevuno is making my tongue hurt
1:24:33 but nice to meet you Chris
1:24:36 um it's really interesting that that
1:24:38 your company is leaning into this I'm
1:24:40 curious what sector it's in hey Kyle
1:24:42 hello Joy
1:24:45 have you ever asked chat gbt to write
1:24:48 some good jokes I have and it sucks
1:24:50 let's go try some now we're gonna try it
1:24:53 on a couple of different tools yeah
1:24:54 let's do this let's do this okay
1:24:58 so we will do this
1:25:01 well this actually this could be fun
1:25:03 um
1:25:05 so we're gonna go to so where I am is I
1:25:07 went to Poe
1:25:09 so it po
1:25:11 we we've got Sage we've got Claude 2
1:25:14 100K Claude instant 100K chat gbt4
1:25:18 Claude instant and Chachi BT
1:25:21 so chat EBT is 3.5 Claude instant is
1:25:25 anthropic's equivalent of that so those
1:25:26 are roughly equivalent Sage I think is
1:25:29 like those two
1:25:31 Claude 2 is like gpt4
1:25:35 yeah
1:25:37 all right so here's what we're gonna do
1:25:38 we're gonna go
1:25:41 we gotta pick a comedian oh we'll do Jim
1:25:43 Gaffigan okay I want you
1:25:47 uh like you are a stand-up comedian uh
1:25:53 wait stand up comedy writer
1:25:57 who writes
1:26:00 um for
1:26:01 let's see
1:26:03 um big name
1:26:05 comedians like Jim
1:26:09 Gaff again and
1:26:13 um John
1:26:14 Mulaney they write their own stuff I
1:26:16 know but that's all right
1:26:19 um
1:26:21 your expert
1:26:25 at writing short
1:26:29 jokes
1:26:31 about
1:26:33 the funny things in life
1:26:37 most people Overlook okay
1:26:42 so we've set up we've set up who this
1:26:45 person is now we say write me a joke
1:26:50 okay here we go about how fast this is a
1:26:54 good one
1:26:55 how fast
1:26:58 avocados
1:27:00 go from unripe
1:27:05 to rotten
1:27:06 which is that's a universal frustration
1:27:09 in all of the world okay
1:27:13 um
1:27:15 sure here's your joke for you avocados
1:27:18 are like gold the Goldilocks of fruit
1:27:20 you buy them unripe and they're too hard
1:27:23 then you wait a day
1:27:25 and they're just right but if you wait
1:27:26 one more day they turn into mushy green
1:27:29 sadness
1:27:30 it's like you're auditioning for a spot
1:27:32 on The Bachelor they go from not ready
1:27:35 to over emotional in just 24 hours
1:27:38 there there's the essence of a joke
1:27:40 there but that is
1:27:42 hugely not funny
1:27:44 okay so let's go let's go take our
1:27:48 little uh let's take our little prompt
1:27:50 here
1:27:52 and let's pop down to we'll do Claude
1:27:55 instance so this is anthropics
1:27:58 version of uh like GPT 3.5
1:28:09 about two days to use it before it turns
1:28:12 into a rotten stack of Goo Ah
1:28:15 I don't get it that's it okay that's a
1:28:19 good setup of a joke but the punch line
1:28:21 is disturbing is it like did you not
1:28:23 feed the puppy and now it's dead and
1:28:25 it's a rotten sack of Goo did you adopt
1:28:28 a dead puppy yeah that one's not good
1:28:30 okay
1:28:31 um
1:28:32 Jesus that was bad all right we're doing
1:28:36 no edits here this is first time joke
1:28:38 first time models okay so now we're on
1:28:40 chat GPT this is chat CPT 3.5
1:28:47 avocados are like Goldilocks of produce
1:28:50 interesting that it used that same thing
1:28:52 they're either too hard mushy or too
1:28:54 rotten it's like they're trying to tell
1:28:56 us to enjoy them at the exact moment
1:28:58 they hit Perfection or else they'll turn
1:29:00 into a mushy Brown mess faster than you
1:29:02 could say guacamole so that is
1:29:05 structurally much better than the sage
1:29:07 one but it's it's still a bad joke like
1:29:10 it's just bad it's not funny okay all
1:29:13 right let's go
1:29:15 we'll go Club two so Cloud 2 is
1:29:17 supposedly closer to gpt4
1:29:23 here's a short joke
1:29:26 my avocados are like an impatient
1:29:29 toddler one minute their heart is a rock
1:29:31 and then next they're a slimy mess
1:29:35 I swear I just bought them and blinked
1:29:37 now they're brown and mushy avocados age
1:29:39 faster than a child star in Hollywood
1:29:42 so again what's interesting here is it's
1:29:45 got an interesting setup it's
1:29:49 age faster than a child star in
1:29:51 Hollywood is that could be a punch line
1:29:54 but my avocados are like an impatient
1:29:57 toddler one minute their heart is a rock
1:30:00 okay that's just [ __ ] disturbing it's
1:30:03 like no no no
1:30:06 that's that's not how you do that okay
1:30:09 now we'll try gpt4
1:30:11 GPT before better not suck at this well
1:30:14 it can't get much worse like like I
1:30:16 don't even know which one Claude instant
1:30:18 was the the one that was was better so
1:30:22 far wasn't it or was it Claude two no it
1:30:24 was Claude instant
1:30:26 all right
1:30:28 this is gpt4 why do why do avocados have
1:30:31 the worst sense of timing I mean you buy
1:30:33 one and it's rock solid like a
1:30:36 green-handed grenade okay we're getting
1:30:38 Maybe
1:30:39 you check on it every day but it's still
1:30:41 not right then one day you go to the
1:30:43 bathroom and boom you come back and it's
1:30:46 turned into a mushy Brown guilt trip
1:30:48 it's like the avocado waits for you to
1:30:50 leave the room and goes quick let's get
1:30:52 rotten before they make us into
1:30:53 guacamole that's a joke
1:30:55 gpt4 wrote a joke
1:30:59 now is that like does that need to be
1:31:01 edited and tightened yeah but that's a
1:31:04 joke
1:31:05 right it's like a hand grenade just
1:31:07 waiting to go off and then you go to the
1:31:08 bathroom one day you know you keep
1:31:10 checking on it day by day that's totally
1:31:11 a joke and that's totally a Jim Gaffigan
1:31:13 joke all right
1:31:14 let's go try
1:31:17 um pie
1:31:20 there's pie and then there was another
1:31:21 one
1:31:23 so Pi this is the inflection company
1:31:29 um
1:31:32 start over
1:31:37 you're a tough crowd no problem let's
1:31:39 start over shall we start there
1:31:48 and avocado's lifespan is like a blink
1:31:52 of an eye you look at it one day and
1:31:54 it's rock hard the next day you look at
1:31:56 it again and it's a it's mushy and brown
1:31:58 it's like
1:31:59 they've gotten the lifespan of a fruit
1:32:02 fly how did I do
1:32:04 horrible
1:32:06 okay so Pi is not a joke writer
1:32:10 um and then there was one more that what
1:32:12 was the other one
1:32:14 there's pie
1:32:17 oh Claude we did cloudpo.com okay so
1:32:21 there's your verdict
1:32:26 for Jim Gaffigan style jokes gpt4 is the
1:32:29 only game in town oh you know what'd be
1:32:31 interesting to try let's go try Bing
1:32:32 because Bing is gpt4 but Bing has also
1:32:35 been neutered
1:32:37 so I won't flip the camera but I'll just
1:32:39 I'll just pop it in here and let's see
1:32:40 what it does
1:32:48 why did the avocado hire a personal
1:32:50 trainer because it wanted to go from
1:32:52 unripe to overwrite in 0.2 seconds
1:32:59 that's a very Microsoft
1:33:03 joke structure
1:33:05 and joke writing
1:33:07 that's awful all right gpt4 at
1:33:10 openai.com is the only place you're
1:33:13 allowed if you're going to use AI to
1:33:14 write jokes you can only use gpt4 from
1:33:17 open AI That's the new law
1:33:21 if I hear anyone in here say oh I used
1:33:25 Claude to write a joke for my stand-up
1:33:27 act I'll kick you out you'll be banned
1:33:29 forever that's it get out
1:33:31 get out
1:33:32 gpt4 is the only one that wrote a joke
1:33:37 it's just like children they start out
1:33:39 rock hard what
1:33:41 no
1:33:44 no you can't say that you go to jail for
1:33:47 that
1:33:51 where are the safety guard rails
1:33:57 yes to Sycamore six seconds down to 45
1:34:00 years yeah 47 years
1:34:02 six seconds I know it's Insanity
1:34:05 Insanity Insanity by the way if you're
1:34:07 new here my name is Kyle Shannon this is
1:34:09 the AI learning lab we like to have fun
1:34:10 here we write jokes We compare different
1:34:12 large language models and see which
1:34:14 one's the funniest guess what none of
1:34:16 them are funny but gpt4 at least
1:34:18 understands what a joke is
1:34:21 um what else do we do if you got
1:34:22 questions about AF i a i pop up in the
1:34:25 questions below I'll do my best to
1:34:26 answer them uh if you're wondering who
1:34:28 the hell I am and what are my
1:34:29 qualifications I have none
1:34:32 uh yeah I'm an entrepreneur I'm the CEO
1:34:36 of a company called storyvine which is
1:34:37 an 11 year old automated video
1:34:39 production platform a video storytelling
1:34:41 platform uh I drink coffee at night
1:34:46 I live in Denver Colorado I'm a Taurus
1:34:48 and I got some unfortunate things going
1:34:51 on on the top of my head all right oh
1:34:54 yeah and the shirts are always a topic
1:34:56 of conversation whether or not they're
1:34:58 appropriate for tick-tocking because of
1:35:00 reading text on top of them
1:35:02 you answer after I'd already hit send oh
1:35:05 that's that's not a problem you can you
1:35:07 can put things in here tree logic is a
1:35:09 heavy task now very heavy Portland lost
1:35:12 most of its big centers years ago that's
1:35:14 interesting they lost their jobs due to
1:35:16 automation years ago that's the thing
1:35:18 about this AI stuff is the the AI stuff
1:35:20 has been around for what two or three
1:35:22 decades and it's it's really been being
1:35:24 incorporated into almost everything we
1:35:27 interact with for the past decade
1:35:29 the new the new stuff about these
1:35:31 generative tools is just this idea of
1:35:34 you know the the speed with which you
1:35:36 can process everything and the the scale
1:35:39 of what these things are trained on
1:35:42 um
1:35:43 so yeah so I think it I you know
1:35:46 I think the acceleration continues
1:35:49 but yeah that that logic
1:35:52 for phone trees is it's just it's got to
1:35:55 be excruciating to work on on the build
1:35:57 side Bill Burr
1:35:59 yeah that would be interesting to
1:36:01 let's see
1:36:04 I'm gonna go do a something here
1:36:08 was it Poe
1:36:09 yeah
1:36:11 I'm gonna say
1:36:14 Bill Bill Burr and Tom Segura are my
1:36:16 favorite
1:36:18 um
1:36:18 two of my favorites
1:36:21 um so I'm gonna say you're a stand-up
1:36:23 comedy writer who writes for big name
1:36:24 comedians like
1:36:27 let's change this Bill Burr and to Tom
1:36:31 Segura
1:36:32 [Music]
1:36:34 you're an expert at writing short jokes
1:36:36 about the funny things in life that most
1:36:38 people overlook
1:36:41 that can really piss you off
1:36:47 write me a joke about how how okay let's
1:36:50 see if the joke gets any better so this
1:36:52 is gpt4 but we switched out the comedian
1:36:54 from Jim Gaffigan
1:36:56 to Bill Burr and Tom Segura
1:36:59 you ever feel like avocados are just
1:37:01 messing with us I mean you buy an
1:37:02 avocado and it's rock hard I'm waiting
1:37:04 for days hoping it'll finally be ripe
1:37:06 enough to use
1:37:07 but the second I turn my back do a
1:37:10 little Bill Burr second I turn my back
1:37:12 the avocado goes from unripe to rotten
1:37:14 faster than politicians promise it's
1:37:16 like they're sitting there smirking
1:37:17 thinking hey you want to eat healthy
1:37:19 right well here's a challenge for you
1:37:21 get me at the right time or enjoy the
1:37:23 taste of disappointment
1:37:25 that's not bad gpt4 can write jokes
1:37:28 GPT for walking right jokes jokes jokes
1:37:32 are they really funny no
1:37:35 but I'll tell you what if you were
1:37:39 you know you know you could do a gpt4
1:37:41 you could do this in a spreadsheet with
1:37:43 the chat with the GPT for Sheets plug-in
1:37:47 you could you could automatically
1:37:49 generate a whole bunch of premises like
1:37:51 why do avocados go
1:37:53 you know you know go rotten so fast
1:37:57 whatever just have have it right like 50
1:38:00 joke premises and then just have it just
1:38:03 keep iterate just keep writing jokes
1:38:05 just keep writing jokes I you know you
1:38:08 spit out 300 jokes in an hour there's
1:38:10 got to be five in there that are worth
1:38:12 worth
1:38:14 saying
1:38:15 Bots aren't overly new the logic
1:38:17 displaced a lot of Entry levels CC work
1:38:19 on the high end yeah well I I get the
1:38:23 Bots aren't new but like what what I
1:38:25 haven't seen I mean maybe it exists it's
1:38:27 just not any company I've ever [ __ ]
1:38:29 called is that there should be no reason
1:38:31 that I'm on hold like I shouldn't need
1:38:35 to go into a phone tree waiting to talk
1:38:37 to a person I should just immediately be
1:38:39 able to start talking to a bot
1:38:42 and then that bot should be dynamically
1:38:45 like it should just have enough access
1:38:47 to all of the information that it can
1:38:49 answer all the questions and if it can't
1:38:51 answer a question then it escalates hang
1:38:53 on Champs gotta go pee
1:38:55 you want to see Mama
1:38:59 yes
1:39:07 your boy needs to feed me
1:39:11 Jeffy needs to be
1:39:15 [Music]
1:39:21 [Music]
1:39:23 um
1:39:29 all right
1:39:33 um
1:39:38 Sean I hit enter before I was done again
1:39:41 I'll just I'll just listen hands are too
1:39:44 sweaty exercise bike I'm glad I'm glad
1:39:47 you clarified why your hands are sweaty
1:39:53 oh food egg by bio industrial trading
1:39:56 shipping
1:39:58 so not funny but you can feel a tinge of
1:40:00 it yeah exactly like those at least GP
1:40:04 T4 understood
1:40:07 joke structure it was a little wordy
1:40:11 and it was a little not funny but
1:40:16 like I feel like the GPT 4 joke was it
1:40:18 was at least close enough that if that
1:40:21 if I just were if I were just stuck with
1:40:23 writer's block like I can't figure out
1:40:24 how to make this premise work I can't
1:40:26 figure out how to make this premise work
1:40:28 and I told chat gbt4 to give me you know
1:40:32 10 possible jokes for that and and I and
1:40:35 I seated it with some of my work or if
1:40:37 it knew my work
1:40:38 um
1:40:39 I it it would it would at least get me
1:40:42 unstuck
1:40:45 I've been live for non 90 minutes that's
1:40:47 dark comedy that's disturbing exactly
1:40:49 that was the kid one this is a great
1:40:51 experiment that's funny best one
1:40:54 interesting snooze yeah I know exactly
1:40:56 is bring is being free with Microsoft
1:40:59 365 it's just free
1:41:03 um time for bed on the East Coast peace
1:41:05 Jen
1:41:06 thanks for being uh entertaining me
1:41:08 tonight you're welcome
1:41:11 um Kyle I'm too humble thank you Becky I
1:41:13 appreciate that
1:41:17 uh
1:41:18 it was great fun tonight thank you
1:41:24 disturbing their conceptualization yeah
1:41:26 the conceptualization of puppies and
1:41:28 babies
1:41:29 yeah it's just like a puppy you bring it
1:41:31 home and you have two days before it's a
1:41:33 gooey pile of you know mess
1:41:37 hmm
1:41:43 avocados aren't funny anyway try a rabbi
1:41:46 and a priest yeah exactly
1:41:49 all right
1:41:52 you have become my total nightly
1:41:54 entertainment education I certainly
1:41:55 appreciate you teaching thank you
1:41:57 appreciate that
1:41:59 bummed out yesterday when chat GPT 4
1:42:01 servers went down yeah I know that was
1:42:03 weird
1:42:05 because of code interpreter Maybe
1:42:08 yeah it's possible because because
1:42:11 code interpreter is spinning up a bunch
1:42:14 of virtual machines and that could be
1:42:17 the way that
1:42:18 um chat gbt works but I I have a
1:42:22 sneaking suspicion that's different so
1:42:24 it it might be
1:42:26 code interpreter might have might have
1:42:28 hosed might have hosed the servers
1:42:30 but I was it was acting really weird
1:42:32 last night so
1:42:35 I ran a
1:42:38 I ran a melt Mount malware scan after my
1:42:41 my live last night I was having weird
1:42:43 malware like captcha things show up and
1:42:47 I didn't I didn't have any I'm on a Mac
1:42:49 so so generally we don't have a ton of
1:42:51 ton of malware stuff on Macs but
1:42:54 I don't know I don't know what's going
1:42:56 on T just gave me this joke why does why
1:42:59 don't skeletons fight each other because
1:43:01 they don't have guts
1:43:03 it's if you're writing a joke for sixth
1:43:06 graders
1:43:08 and not even these days doesn't feel
1:43:10 like a joke for fourth graders these
1:43:12 days
1:43:13 third graders
1:43:16 um someone in the chat just made a great
1:43:18 Point avocados they're not funny
1:43:21 it can be funny
1:43:24 but you just need to make fun of
1:43:25 hipsters with them on toast right isn't
1:43:28 that the way to do it someone uh
1:43:35 oh Sonia so
1:43:37 so bing.com requires you to be an edge
1:43:42 for it to be free is that true wait let
1:43:44 me go look I'm gonna go run
1:43:47 Chrome if that's the case I'd like to
1:43:50 know that because I don't want to steer
1:43:51 people wrong because I've been telling
1:43:53 everyone it's free just go there you can
1:43:55 use it
1:43:57 you can do it
1:44:02 in
1:44:04 calm
1:44:16 chat
1:44:22 oh yeah look at that
1:44:27 Bing chat is only available in Microsoft
1:44:30 Edge
1:44:32 look at Microsoft getting all
1:44:34 monopolistic and all
1:44:37 they're like oh yeah that Chrome thing
1:44:40 from that Google company you can't use
1:44:43 that no no you're gonna have to go on
1:44:46 ahead and come on over to the dark side
1:44:48 we got your free gpt4
1:44:52 all you can eat it's here for the taking
1:44:56 just download Edge mcping your default
1:45:00 it'll be fine
1:45:02 that's pretty [ __ ] evil all right
1:45:05 I'm glad you said that thank you for
1:45:07 saying that
1:45:08 so Bing is free salon
1:45:11 because you sign your soul to the devil
1:45:16 oh Microsoft
1:45:18 your ways are dark and gnarly no
1:45:26 it gives me a perverse thrill
1:45:28 to see how you run your business
1:45:31 leverage people extract money from them
1:45:35 it's delicious
1:45:39 avocados are like children they start
1:45:42 out rock hard what what no
1:45:46 that's not the joke
1:45:51 what was the big one the Bing joke was
1:45:54 even worse
1:46:06 why did the avocado hire a personal
1:46:08 trainer because it wanted to go from
1:46:09 unripe to over ripe in 0.2 seconds
1:46:17 yeah Microsoft I know
1:46:20 it's listen they're doing they're
1:46:22 they're doing the good work they're
1:46:23 they're getting this AI stuff out in the
1:46:25 world
1:46:26 we're gonna we're gonna give them we're
1:46:28 gonna give them
1:46:29 some benefit of the doubt but you know I
1:46:32 do have 35 or 40 Years of Microsoft
1:46:34 bashing to get over
1:46:38 code interpreter wasn't working for me
1:46:40 today mid-afternoon EST but maybe
1:46:42 because
1:46:44 I was waist deep in H2O are you in
1:46:47 Vermont hopefully not good Lord that
1:46:50 looks like a mess up there
1:46:55 my MacBook Air made me download Edge to
1:46:58 use Bing chat feature yep
1:47:01 looks like that's what they're doing it
1:47:02 won't write dirty jokes yeah they won't
1:47:04 you're gonna have to go to the open
1:47:05 source models for that and then I bet I
1:47:07 will write some jokes that are so dirty
1:47:09 that you're like oh good God geez
1:47:13 is this connected to the internet is the
1:47:16 FBI watching
1:47:19 yeah
1:47:21 don't go to those unsupervised models if
1:47:23 you you know if you don't want to get
1:47:25 smacked upside the head with some
1:47:28 some freedom
1:47:33 wait a few weeks
1:47:35 wait wait a few weeks ago you said they
1:47:37 were ahead in the AI space what's
1:47:39 changed I said who was ahead in the AI
1:47:41 space
1:47:43 um open AI
1:47:46 what's changed is in the past two days
1:47:50 anthropic came out with Claude
1:47:55 um inflections uh
1:47:59 Pi is out there
1:48:03 um
1:48:03 Poe now has gpt4 in it
1:48:09 they got rid of browse with Bing
1:48:14 um and it was just wonky yesterday so
1:48:16 but I don't know if that's what you're
1:48:17 asking
1:48:18 a few weeks ago said they were ahead in
1:48:20 the AI space I don't know who you're
1:48:21 talking about Becky if you if you let me
1:48:23 know I'll say it yeah Bing no yes oh yes
1:48:27 Vermont oh I'm so sorry
1:48:32 sending all of my best wishes and
1:48:35 thoughts to Vermont Vermont I used to I
1:48:37 did a summer theater in Warren Vermont
1:48:39 for like
1:48:41 four or five years
1:48:42 and uh damn it's beautiful up there but
1:48:45 holy crap that flooding is insane
1:48:50 like really sad really sad
1:48:53 can I just hear AI do you get it do you
1:48:56 get it
1:49:00 is that something I say you get it you
1:49:02 get it I don't know I don't know if
1:49:05 that's something I say it's one of the
1:49:06 joys of add I've got all these you know
1:49:09 verbal tics that I have no [ __ ] clue
1:49:11 what I'm saying half the time so
1:49:14 um is it true that chat gbt cannot spell
1:49:16 lollipop backwards I don't think it can
1:49:19 do things like spell it backwards or
1:49:21 give me exactly this number of things or
1:49:24 tell me how many things that you've
1:49:25 created it all of that stuff it does
1:49:27 poorly
1:49:28 um we can try it
1:49:32 um spell
1:49:35 Lolly
1:49:39 pop back words
1:49:52 pillow Paul no it did not come close
1:49:58 nope
1:50:00 it can't but it makes sense it
1:50:03 if if you understand how it does what it
1:50:06 does it has no awareness of
1:50:10 of what it's actually created it only
1:50:12 has the awareness of what's the
1:50:14 probability what's the highest
1:50:16 probability of the the next word in the
1:50:18 sequence based on your prompt
1:50:21 and so until it gets better at
1:50:25 self-reflection essentially it's going
1:50:28 to be shitty at stuff like that because
1:50:30 that's bad
1:50:32 you know it would be interesting
1:50:34 um am I in code interpreter
1:50:37 no let me try it in code interpreter
1:50:44 um right a python
1:50:48 right
1:50:50 let's see right
1:50:53 and run a python
1:50:59 script
1:51:03 that
1:51:04 um
1:51:07 spells
1:51:09 lollipop
1:51:13 backwards
1:51:19 let me run it for you
1:51:23 yeah it spelled it wrong again
1:51:26 oh no it didn't
1:51:27 pop it law
1:51:31 so so it
1:51:34 if if so I asked it to write a python
1:51:37 script
1:51:38 so because it's writing a python script
1:51:41 it can actually write the logic and and
1:51:43 it it knows the logic of the script and
1:51:46 it can run the script and it ran it and
1:51:47 it got it right so if you have it if you
1:51:50 have it right code to do that it can do
1:51:51 it right but just the way the large
1:51:53 language model Works things like
1:51:55 counting predicting like predicting
1:51:57 length of things
1:51:59 um getting an exact number like give me
1:52:02 13 bullet points and it'll give you 12
1:52:04 or 14 or 16 and you're like well you
1:52:06 give me 14. it's like oh my apologies
1:52:08 here's 13 bullet points and then it
1:52:10 gives you 16.
1:52:12 that's just the nature of how the thing
1:52:13 works but if you use code interpreter
1:52:16 and have it write scripty scripts for
1:52:19 you then then you can start to do things
1:52:20 like actual math and actual prediction
1:52:22 and actual
1:52:23 word jumbles
1:52:27 or or you could just go down to the
1:52:29 grocery store and get the word jumble in
1:52:31 checkout in the checkout line do they
1:52:33 still sell those is that a Gen X thing
1:52:34 do those still exist the word jumbles no
1:52:38 that's just because I'm old right
1:52:47 that sent that one through the nose is
1:52:49 it true yeah cannot spell lollipop
1:52:51 backwards that is correct
1:52:54 oh Microsoft so so I think Microsoft
1:52:57 Becky is
1:52:59 um
1:53:00 I think for a company
1:53:03 the size and nature of Microsoft they
1:53:06 are acting more like a startup than they
1:53:08 have in a long long time
1:53:10 I think the fact that they are
1:53:12 so aggressively
1:53:15 um incorporating
1:53:17 open AI chat GPT into all of what they
1:53:20 do they're placing a big Bet On It
1:53:23 um I found out today they're one of the
1:53:26 major investors in
1:53:28 um inflection
1:53:30 which is the which is a competitor to
1:53:33 open AI
1:53:34 um so Microsoft's a big company they're
1:53:36 going to do what they're going to do
1:53:38 um
1:53:42 you know they will always tend toward
1:53:45 the more conservative to protect the
1:53:47 revenue of what they've got
1:53:48 um but yeah I think I think listen for a
1:53:51 company as big as they are
1:53:53 what they've done in the past six months
1:53:56 is nothing short of remarkable and you
1:53:59 know whether or not they can sustain
1:54:00 that
1:54:02 um and really take it to Google long
1:54:04 term is going to be really interesting
1:54:06 to see my my gut tells me yes they're
1:54:10 going to be able to do that
1:54:12 because I don't I just don't
1:54:15 I don't see goo see Google has the
1:54:17 engineering skill to do the science of
1:54:20 this
1:54:21 really well
1:54:22 I don't know that they have the
1:54:24 usability and human factors expertise
1:54:27 like an apple does you know Microsoft
1:54:29 isn't great there but they've gotten a
1:54:31 lot better in the past six months
1:54:35 um and maybe before that like frankly I
1:54:37 just I haven't interacted with Microsoft
1:54:39 products in so long that you know it's
1:54:41 it's possible they've been improving
1:54:43 over the years but the stuff they've
1:54:44 been doing with the AI integration is is
1:54:46 really solid
1:54:51 act like a spelling teacher and spell
1:54:53 lollipop in reverse order did it right
1:54:55 oh that's interesting
1:54:57 yeah I mean that's the thing the the the
1:54:59 code that Microsoft used to program Bing
1:55:03 chat which is sitting on top of a gpt4
1:55:06 large language model and what openai did
1:55:10 with theirs like it's two very different
1:55:12 code bases sitting on top of the same
1:55:14 model so it's fascinating to compare
1:55:15 those two
1:55:18 and and I get you know my my sense is
1:55:21 that
1:55:22 long-term
1:55:25 Microsoft's Bing chat is just gonna get
1:55:27 better and better and better like it's
1:55:29 kind of like this like
1:55:32 GPT three 3.5 and 4 for for open AI did
1:55:38 this and now they've kind of plateaued
1:55:40 and it's even starting to come down a
1:55:42 bit I feel I feel like Microsoft went up
1:55:45 to maybe like three quarters of where
1:55:47 they were like you know open AOS here
1:55:49 Microsoft came to here and then
1:55:51 Microsoft had to bring it down to like
1:55:53 50 because of the New York Times thing
1:55:56 and I just feel like Microsoft is going
1:55:58 to be on this slow improvement over the
1:56:00 next two years or three years it's just
1:56:01 going to get better and better and
1:56:02 better slowly steadily
1:56:05 um
1:56:06 and like when they roll out
1:56:09 co-pilot or chat GPT integrated into
1:56:12 office 365. it's going to be very well
1:56:15 thought out
1:56:16 it's not going to be some janky ass
1:56:18 add-on
1:56:19 so I think I think Microsoft's a really
1:56:22 interesting one to to pay attention to
1:56:31 do
1:56:32 when little kids tell ridiculously
1:56:35 toward jokes they always say if you get
1:56:37 it oh do you get it do you get it I get
1:56:39 it I get it
1:56:41 that's the that was the Bing joke
1:56:46 that's so bad
1:56:48 why did the avocado have a personal
1:56:50 trainer because I wanted to go from
1:56:51 unripe to ripe from 0.2 seconds you get
1:56:53 it get it
1:56:56 oh that's bad it's so bad
1:57:01 Celebrity Jeopardy was a rerun last
1:57:03 night so why are you here tonight
1:57:07 about last night about last night and
1:57:09 failed miserably wait what I don't know
1:57:12 if that's a reference to wait Claude
1:57:14 allows PDFs as input just like that yes
1:57:17 I just had to summarize my employee
1:57:19 handbook yes so look this is
1:57:23 so a couple of things to show overnight
1:57:27 from from when we talked last night
1:57:28 there's new things so right now I'm in
1:57:30 po po.com
1:57:35 um
1:57:37 you used to have access to Claude
1:57:39 instant 100k
1:57:42 now you have access to Claude 2 100K so
1:57:46 Claude 2 is like the equivalent of gpt4
1:57:49 but they now have a version with a
1:57:51 hundred thousand token
1:57:53 um context window so both of these clods
1:57:57 in Poe have 75 000 word context windows
1:58:01 and gpt4 right now has four
1:58:04 maybe eight I don't know what it is here
1:58:06 but I think it's four
1:58:08 um the other thing that's going on at
1:58:10 Poe I got this email today this morning
1:58:13 see that little paper clip you can now
1:58:15 upload files into Poe as well
1:58:18 so that's pretty remarkable
1:58:31 um
1:58:33 so that's new
1:58:35 um
1:58:37 what else is I going to show you if you
1:58:39 go to claude.ai
1:58:46 um hair growth geek if if you want to
1:58:48 hop up on stage your thing just popped
1:58:49 up here if that was you doing that just
1:58:51 do it again I'll bring you up on stage
1:58:53 that might have just been Tick Tock
1:58:55 going hey do you want to do this but if
1:58:57 you do if you have a question or want to
1:58:59 hop up here let me know
1:59:00 um
1:59:03 okay this is pi so if you go to
1:59:04 claude.ai
1:59:06 foreign
1:59:11 this now also has
1:59:14 um a little paper clip here you can
1:59:16 upload up to five files so five PDFs or
1:59:19 I I don't think this will do zip files
1:59:22 let me let me see if this does zip files
1:59:25 I don't think it does but you can do
1:59:26 five PDFs
1:59:31 um yeah you can't do zip files but you
1:59:34 can do csvs and PDFs
1:59:40 well dogs singing Goodwill Hunting
1:59:42 screenplay
1:59:44 so let's open that
1:59:46 so now it's got the Goodwill Hunting
1:59:47 screenplay in there so I'll say
1:59:50 um who are the main
1:59:53 characters
1:59:58 and who likes apples
2:00:04 I don't have enough content to know so
2:00:06 it didn't do that well all right
2:00:10 failed there anyway Claude and Claude
2:00:14 dot AI also has a hundred thousand token
2:00:17 context window so
2:00:19 um so a lot of things changed overnight
2:00:21 like I I feel like there are legitimate
2:00:25 chat gbt competitors in the market now
2:00:29 um and there haven't been there really
2:00:31 haven't been I mean being in being an
2:00:33 open AI we're kind of in cahoots
2:00:36 Bard just wasn't anywhere close meta
2:00:39 didn't all of their stuff was crap
2:00:42 um
2:00:43 the Claude stuff was okay but not great
2:00:46 but then Claude 100K now Claude 2 with
2:00:49 100K plus you can upload files
2:00:52 thank you for that Sean I appreciate
2:00:54 that
2:00:57 um
2:01:01 all right why do things
2:01:04 that will take longer to input
2:01:07 the prom oh yeah but why do things that
2:01:10 will take longer to input The Prompt
2:01:12 than do the thing manually that's that's
2:01:14 just one of those things with computers
2:01:16 you know how you spend two hours trying
2:01:18 to get you know some piece of software
2:01:21 working you download it try to run the
2:01:24 demo then you find out the demo doesn't
2:01:26 actually two hours later
2:01:28 you realize oh I could have just written
2:01:30 that thing in that amount of time
2:01:32 because it's new technology and we're
2:01:34 geeks and Geeks Geeks are gonna geek so
2:01:39 you know and you know quite frankly the
2:01:43 the reason to invest the time into
2:01:46 overwriting a prompt
2:01:49 is that when you get it right it's
2:01:51 scalable
2:01:54 like if I if I really [ __ ] around with
2:01:56 that joke writing prompt like there's I
2:01:59 guarantee you there's a way to write
2:02:01 a joke joke writing prompt with a few
2:02:04 example structures in there
2:02:07 um where you could get it to
2:02:09 consistently write the kind of jokes
2:02:11 that you tell
2:02:13 like it might take you
2:02:17 two hours a session for ten sessions
2:02:20 five sessions
2:02:22 where you could have been writing jokes
2:02:24 but if you get that [ __ ] dialed in
2:02:26 you could you could really start
2:02:29 cranking out some interesting stuff and
2:02:30 and change up your act and you know so
2:02:32 so so I I think there's I think there's
2:02:36 there's an investment in learning how
2:02:38 these things do what they do that's
2:02:40 worth worth
2:02:41 that waste of time but it's a it's a
2:02:43 great point
2:02:47 funny funny jumbles I know it
2:02:50 uh have you have you ever asked it to
2:02:53 list U.S presidents who died in office I
2:02:55 have not
2:02:58 I'm I'm wait I'm going to get a raise
2:03:00 with this so much of what I do is
2:03:03 hyperspecific so the PDF helps
2:03:06 definitely definitely go get a raise
2:03:08 with this I'm telling you we have a
2:03:11 short window where if you've got Senior
2:03:14 Management that's slightly curious about
2:03:17 this but doesn't really feel like
2:03:18 learning it if you can pick on the
2:03:21 expert and make them look like
2:03:22 Superstars
2:03:26 there's there's going to be a whole lot
2:03:27 of new VP of AI generation in companies
2:03:31 that were like you know
2:03:33 had some sort of like you know
2:03:35 entry-level associate level project
2:03:37 management all of a sudden wait why are
2:03:40 they the VP of AI prompting integration
2:03:43 I don't know she went into Bob's office
2:03:46 and came out with a raise and a new
2:03:48 office
2:03:51 go make Bob look like a superstar
2:03:57 I mean Bob's a dick but he can give you
2:03:59 a raise
2:04:00 foreign
2:04:04 thank you all enjoyed it bye Joy thank
2:04:07 you iced coffee man oh yeah Charles
2:04:11 oh
2:04:12 ah it's a lovely bouquet
2:04:19 do you think Chad GPT is bad no I think
2:04:22 chat GPT is going to fundamentally
2:04:23 change the world
2:04:25 I think chat GPT is the inflection point
2:04:28 after which nothing will be the same
2:04:32 I think Chachi PT is remarkable I think
2:04:34 it's the most powerful tool put in the
2:04:37 hands of regular everyday people
2:04:40 probably in history
2:04:42 I think it's remarkable
2:04:44 um
2:04:45 and it's not perfect it hallucinates
2:04:47 meaning it lies it's
2:04:50 overly wordy sometimes underly wordy
2:04:53 sometimes
2:04:54 um
2:04:55 will confidently produce shitty work and
2:04:58 think it's good but doesn't think
2:04:59 anything it'll confidently produce
2:05:01 shitty work
2:05:03 and then if you really push it hard then
2:05:05 it starts to get better and then if you
2:05:07 really push it hard after that you can
2:05:11 start you realize how remarkable it is
2:05:13 it's it is truly remarkable technology
2:05:16 the thing that makes it weird right now
2:05:19 is that it's so raw and early
2:05:23 and
2:05:24 it was tossed together chat GPT was
2:05:27 written in two weeks
2:05:29 two weeks
2:05:31 now the underlying model underneath it
2:05:34 that was in development for seven years
2:05:38 one gpt2 gpt3 GPT 3.5 chat gbt
2:05:45 just a little
2:05:47 shitty thin application layer sitting on
2:05:50 top of this powerful model
2:05:52 and then that little thin application
2:05:54 layer
2:05:55 100 million people started using it in
2:05:58 the first six weeks fastest adoption of
2:06:02 technology in history
2:06:04 so I think chat EPT is [ __ ]
2:06:07 remarkable
2:06:08 and it's imperfect and new and duct
2:06:11 tapey and spit and string are holding it
2:06:14 together
2:06:17 it's just the nature of being
2:06:19 so it's the price you pay for being
2:06:22 early if you want all this stuff to be
2:06:24 slick and perfect wait for a year and a
2:06:25 half and it'll be
2:06:27 you won't even the whole idea of having
2:06:29 to prompt stuff I think that drifts very
2:06:32 quickly into the background
2:06:35 you'll just say hey I want this and
2:06:36 it'll go figure out how to do it right
2:06:38 let's see I'll do a few more questions
2:06:41 here and then I gotta go
2:06:44 isn't even on building a competitor as
2:06:47 well in AI never bet against Elon
2:06:51 yeah he just hasn't come out with
2:06:52 anything yet he's a little distracted
2:06:54 with he he's got a cage match to prepare
2:06:57 for and and Zuck you know Zuck decided
2:07:01 to to clone Twitter just to piss him off
2:07:04 so I think I think musk is is going to
2:07:07 be a little distracted I mean don't get
2:07:10 me wrong the man could
2:07:12 he shouldn't be able to run two
2:07:14 companies and he's running what seven
2:07:15 right now and you know many of them are
2:07:18 changing the world so I'm not going to
2:07:20 put anything past the guy but he's a
2:07:22 little busy right now so I assume that
2:07:25 we're not going to see anything from X a
2:07:27 i for probably till early next year
2:07:31 and then they'll probably come to Tesla
2:07:35 so all he has to do is
2:07:37 pluck three of them out of that and say
2:07:40 hey go take all that [ __ ] you did here
2:07:42 and go spin it up on these 10 000 gpus
2:07:45 over here and
2:07:47 make it you know 100x what open ai's
2:07:51 doing and tell me when you're done that
2:07:52 I mean that conversation's probably
2:07:54 already happened
2:07:56 how long before AI will replace news
2:07:58 anchors and other types of Information
2:07:59 Services I think
2:08:07 I think that AI May reintroduce news
2:08:10 anchors for local news that no longer
2:08:13 exist
2:08:17 so I I don't see it as an as a net loss
2:08:20 I don't I I think big networks and
2:08:23 things like that you're going to have
2:08:25 you're going to have um
2:08:29 you're gonna have news anchors you're
2:08:31 gonna have regular news anchors I think
2:08:32 though what happens in the smaller
2:08:34 markets
2:08:35 where maybe the the local newspaper the
2:08:39 local TV station has been nationalized
2:08:41 because it got bought up by you know
2:08:43 Gannett or whatever the [ __ ] it's called
2:08:45 now and
2:08:46 uh tegna I think it's called now you
2:08:49 know one of those big companies or one
2:08:50 of the big newspaper companies that buy
2:08:52 up all the local markets and then the
2:08:53 local
2:08:55 the local news economy literally just
2:08:58 goes away so so you only have national
2:09:00 news being masqueraded as as local news
2:09:04 I think we could see independent
2:09:07 journalism come back
2:09:09 with newspapers that are generated from
2:09:12 Ai and with news broadcasts that are
2:09:15 generated from AI in an economically
2:09:17 feasible way that that
2:09:20 brings
2:09:22 local journalism back
2:09:25 and I think that would be good there's a
2:09:27 business to start
2:09:28 although I gave that one out two nights
2:09:30 ago
2:09:31 all right
2:09:33 I wanted to make jokes from thousands of
2:09:36 joke structures that I can input how to
2:09:38 force it to use the structure
2:09:40 you can just so go to claude.ai
2:09:45 go there or go to poe.com and pay the
2:09:48 subscription but you can do it for free
2:09:49 at claude.ai claude.ai has a context
2:09:52 window of a hundred thousand tokens
2:09:54 that's 75 000 words so put
2:09:59 categorize make a spreadsheet of your
2:10:02 thousands of jokes in your different
2:10:04 categories
2:10:06 upload that or copy and paste it into
2:10:10 the prompt and say based on all of these
2:10:12 jokes in these categories
2:10:16 I want you to ask me for a topic and a
2:10:19 premise
2:10:21 like a subject and a premise
2:10:24 and or or a a premise and a and a joke
2:10:28 category right I want a knock knock joke
2:10:31 I want a story I want to this I want to
2:10:33 that and you've got them all categorized
2:10:35 up above
2:10:36 so ask me you know for a premise and
2:10:40 then I want you to give me a joke per
2:10:42 category
2:10:44 whenever I give you an answer
2:10:48 and then and then I want you to and I
2:10:51 want you to ask me which of the jokes
2:10:53 you wrote is the funniest
2:10:55 and then I want you to give me 10 more
2:10:57 like that you could do that right now
2:10:58 you could just like I would go right now
2:11:00 to claude.ai and do that Mr AI a i o oh
2:11:05 fantastic Name by the way kudos to the
2:11:08 Mr AI a i o come on
2:11:10 huh
2:11:12 do you see people how it's done do you
2:11:14 see how it's done
2:11:17 Elon
2:11:18 never have a generative AI project
2:11:20 unless he buys it we'll see you know
2:11:23 we'll see
2:11:25 I mean he's not a he's not a dumb man
2:11:27 he's he's actually a smart guy
2:11:34 I'm up to over 9 000 likes whoa [ __ ] tap
2:11:39 that like button let's give me over ten
2:11:41 thousand thank you for the for the heads
2:11:43 up I don't pay attention to any of this
2:11:45 [ __ ] I don't know where are you seeing
2:11:46 this
2:11:49 total views diamonds good connection I
2:11:53 don't know oh here's my duration I've
2:11:55 been at this for two hours and 12
2:11:56 minutes right as my butt's getting numb
2:12:00 it's like clockwork
2:12:03 guess who used chat GPT to work on their
2:12:06 resume and job app today fantastic
2:12:09 yes
2:12:10 like chess GPS great for that kid's
2:12:14 bedtime thank you thanks Sean appreciate
2:12:15 it
2:12:17 all right lol Kyle pause when you switch
2:12:20 from front to back camera yep we miss
2:12:22 what you say sorry about that
2:12:24 what Bill Gates Affliction with
2:12:26 Microsoft now what's Bill Gates
2:12:27 Affliction with Microsoft now I don't
2:12:30 know what you mean by that did something
2:12:31 happen that I missed
2:12:33 or like what's his oh what's his
2:12:35 affiliation is that what you're asking
2:12:37 uh made more sense to chat GPT than
2:12:41 backwards
2:12:43 have a good night apple apple it's a
2:12:45 cult it is a cult I'm part of that cult
2:12:47 I'm an apple Fanboy from way back
2:12:51 gotta go pleasure as always thanks
2:12:52 silver fox
2:12:54 s b okay there's all that stuff
2:12:58 I'm gonna wind this puppy down
2:13:00 Claude is great but they get but they
2:13:03 better get their act together on scale
2:13:05 it was super slow today I know I mean
2:13:07 listen this is
2:13:11 you know Elon can be pissed at Sam
2:13:13 Altman all he wants for taking that
2:13:16 thing
2:13:16 for profit but
2:13:20 the fact that open AI partnered with
2:13:23 Microsoft and Microsoft could throw
2:13:25 their entire Cloud infrastructure
2:13:27 at
2:13:29 making this thing work like if think
2:13:32 about
2:13:33 if you threw together
2:13:38 a little piece of software
2:13:41 in two weeks let's say you did it
2:13:43 in a month it took you a month and you
2:13:46 launched that piece of software and that
2:13:48 piece of software even went to a million
2:13:50 users in two months
2:13:53 could you handle it no if it went to a
2:13:56 hundred million users in six weeks like
2:13:58 chat GPT did
2:13:59 like that thing shouldn't be functioning
2:14:02 right why is it functioning because they
2:14:04 cut a deal with Microsoft and part of
2:14:06 that deal was we're going to launch this
2:14:09 thing and if if by some stretch of the
2:14:11 imagination it [ __ ] takes off
2:14:13 we want to be able to scale up on your
2:14:16 Azure Cloud infrastructure like that
2:14:19 that's why that thing's still alive so
2:14:21 yeah I totally agree that claude's gonna
2:14:23 have to get their [ __ ] together and if
2:14:25 you look at the
2:14:27 um the the investment uh you know what
2:14:30 the what the folks at inflection are
2:14:32 talking about what they're going to
2:14:33 spend the bulk of their 1.3 billion
2:14:36 dollars on it's building
2:14:38 a massive data center to be able to
2:14:41 scale so I think you're right I think
2:14:43 anthropic's got to figure that [ __ ] out
2:14:45 pretty quick
2:14:47 like they could have
2:14:49 could have here your
2:14:51 answers
2:14:53 but if they're too slow and don't work
2:14:54 like now that we've got
2:14:57 thank you
2:14:59 at least three kind of major competitors
2:15:03 and pretty soon we'll probably have five
2:15:05 or six
2:15:07 like we as humans we can only like we
2:15:11 can do only do things in three we can
2:15:12 only do things in three
2:15:14 so let's say there's six major
2:15:16 competitors with these large language
2:15:17 models Plus open source
2:15:20 only three emerge out of that and that's
2:15:22 gonna it's gonna the the winnowing is
2:15:24 gonna happen quick
2:15:26 so who survives
2:15:29 it's gonna be someone who's got some
2:15:30 [ __ ] infrastructure that's for sure
2:15:33 um easily
2:15:35 how long before AI webcoast news anchors
2:15:38 we did that the Thousand jokes when we
2:15:40 did that
2:15:41 that's a good idea
2:15:43 regarding local journalism Revival I
2:15:45 think it's actually a really important
2:15:47 thing
2:15:48 that could be a good business that could
2:15:51 be absolutely solved with this AI stuff
2:15:54 probably pretty simply I was even
2:15:56 thinking about I think I might do a
2:15:58 zapier automation
2:15:59 um
2:16:00 to um to mock it up because I think you
2:16:03 could probably pull local public records
2:16:06 and with a zapier automation create a
2:16:09 daily newspaper a daily local newspaper
2:16:11 for your neighborhood
2:16:13 why not
2:16:16 prompting wouldn't be that tough it's
2:16:18 just sourcing the public records and
2:16:21 automating it
2:16:24 not that tough now what would be tough
2:16:26 is how do you turn that into a business
2:16:27 but that would be an interesting
2:16:29 exercise to play with
2:16:31 thanks for answering my questions you're
2:16:32 welcome joke sir joke sir Jack sir the
2:16:35 mighty jokes are the mighty oh jokes are
2:16:37 the mighty
2:16:38 over ten thousand likes now peace thank
2:16:40 you people
2:16:42 yes sir affiliation I'm assuming he's
2:16:45 still on the board uh Bill Gates
2:16:47 affiliation with Microsoft I think he's
2:16:50 still on the board but he's definitely
2:16:51 out of day to day I and he doesn't it's
2:16:53 funny he doesn't he doesn't seem to have
2:16:55 a horse in the game although I think
2:16:58 go look at the investor
2:17:01 um
2:17:02 the the investor list for inflection I
2:17:05 think Bill Gates and Microsoft both
2:17:07 invested in uh
2:17:09 in inflection
2:17:12 thanks Kyle you and your hair made
2:17:14 another fun evening I know oh it's
2:17:16 getting worse if you see it's [ __ ]
2:17:18 losing it it's
2:17:20 I'm ready to take it's like a helicopter
2:17:21 that's like if you said to Mid journey I
2:17:24 want you to show spinning helicopter
2:17:26 blades I think that's what it would make
2:17:31 all right
2:17:33 um
2:17:35 I'll leave that one as the final comment
2:17:37 thank you all so much for hanging out
2:17:40 peace good night Chris
2:17:42 peace out everybody I'm gonna go do
2:17:44 something with this hair
2:17:46 act like Kyle's here
2:17:50 peace everyone