aiLL 09-05-23 - Kyle Debates AI in Education and Predicts an AI-Powered Future

Description

Chapters

Transcript

0:02 they are learning lab
0:05 welcome welcome welcome uh if you would
0:08 go ahead and share the live if you've
0:10 got
0:11 friends and followers you think might
0:13 dig this even if you think they're if
0:15 they're gonna think you're weirdos do it
0:18 anyway
0:19 um
0:20 welcome from Denver it is eight o'clock
0:22 on the money
0:25 and looks like Ben Shelton is about to
0:27 take set one
0:29 unless tiafo can turn things around
0:34 crazy I tell you crazy I tell you chaos
0:37 generation what's happening
0:39 Ted looks at life how are you Tobias
0:42 welcome sir welcome everybody Welcome to
0:45 the lab
0:48 we're going to talk about AI stuff
0:50 tonight the shirt tonight if you're
0:51 wondering is is the old classic Basquiat
0:55 Cassius Clay
0:58 backwards
1:00 it's tough to get these things printed
1:01 backwards like this that makes it a
1:03 collector's item
1:09 oh just you know never dull moments what
1:14 I do today
1:15 um
1:16 I did I discovered I screwed up I
1:20 screwed up The Tick Tock shop even more
1:21 than I thought I did so
1:24 it's kind of untangle that wackiness
1:27 stayed up till almost 1am watching the
1:29 German and Italy match over four and a
1:31 half hours wow crazy yeah what's going
1:33 on here all right Shelton serving for
1:36 the for the first set double set point
1:39 40 15.
1:41 see if he cracks out a oh wow 127 mile
1:45 an hour Ace
1:47 blasted it past tiafo wow packing up to
1:51 go to Denver tomorrow okay
1:54 welcome okay farmer
1:58 Dr J Mark attiel how are you welcome
2:03 everybody so we got some Irregulars here
2:05 those are what we call the regulars
2:08 they have their own t-shirt design but
2:10 uh I'm still figuring out how to make
2:12 t-shirts
2:13 and not [ __ ] it up you know old guy with
2:16 new technology you know how it is it
2:18 doesn't go good
2:22 hey what are you gonna do
2:24 anyway uh this is the AI learning lab
2:28 um
2:29 if you're new to uh to this space and to
2:33 generative AI I will
2:36 talk you through some stuff to get you
2:38 started we'll start you with some
2:40 some URLs we'll we'll talk through those
2:43 we'll we'll show some pretty pictures
2:45 these were all made in
2:47 um ideogram which I still I've been
2:49 Hawking that like I worked there
2:51 just because it's pretty amazing that
2:54 you can now do text in AI image
2:56 generation
2:57 it seemed like that was going to be
2:59 farther away than whatever it was six
3:01 eight months and then it just showed up
3:03 and it didn't show up from mid Journey
3:05 or stable diffusion I was expecting it
3:06 to come from stable diffusion quite
3:08 honestly because they seem like they
3:10 have their [ __ ] together mid-tierney's
3:12 just an 11 person company
3:14 um
3:15 and uh and this startup
3:19 this startup called uh
3:21 I don't know what the company's called
3:23 ideogram I don't know if that's the
3:25 company but uh but yeah they figured it
3:28 out man they cracked the code
3:31 so we're off to the races with text and
3:32 images
3:33 all right so what we do here is my name
3:36 is Kyle Shannon I'm an entrepreneur of
3:38 many years I've started 12 or 15
3:39 companies I'm currently the founder and
3:42 CEO of a company called storyvine which
3:44 is an automated video storytelling
3:45 platform
3:47 we don't our current platform doesn't
3:49 use a ton of AI for the automation
3:51 believe it or not
3:52 um it's it's a more traditional
3:54 automation
3:55 well you actually had to program [ __ ]
3:57 and think about it
3:59 that's all different now so we're
4:02 looking to to Really expand the product
4:04 Vision with AI well I have expanded the
4:06 product Vision now we're looking to
4:07 build it and get it paid for
4:10 um so that's happening and I was also an
4:12 entrepreneur in the early days of the
4:13 World Wide Web so I was a founder of
4:15 agency.com back in the old and Tiny days
4:17 and uh the parallels between the early
4:20 days of the web and the early days of
4:21 this AI stuff are quite remarkable quite
4:23 remarkable and so that's one of the
4:25 reasons I'm so excited about this stuff
4:27 is that we don't get to experience these
4:30 technological
4:31 primordial soup moments very often
4:35 um there have been three of them in my
4:37 lifetime one I missed and that was the
4:39 PC Revolution
4:41 the second one was the world wide web
4:42 and I hit that one square in my 30s and
4:45 I didn't think I'd see another one in my
4:46 lifetime but here it is this is it and
4:49 this is really exciting so if you have
4:51 questions about AI can I make games
4:53 using AI yes in fact you can
4:56 um you can just go to Chachi PT so here
4:58 I'll I'll jump in and and just get get
5:01 started if you haven't played with chat
5:02 gbt yet just go there
5:05 go there right there
5:07 chat.openai.com and just ask it to write
5:09 you
5:10 a flappy birds clone in C sharp that
5:13 will run in unity
5:16 and then take that code paste it into
5:18 unity and hit run and then if it if it
5:21 gives you an error then take the error
5:23 take it back to chatgpt and said you
5:24 gave me an error so so yes you can
5:28 um you can also use there's another one
5:30 called pi
5:32 which you can talk to
5:34 and it's really conversational we call
5:36 it chatty GPT and chatty GPT will um you
5:40 can just tell it to get to make up a
5:42 game like you want to learn a particular
5:44 subject and tell it to invent a game to
5:46 make it more fun to learn and it'll just
5:48 do it so all sorts of stuff you can have
5:50 these things act like dungeon Masters
5:52 all sorts of crazy crazy ass [ __ ] crazy
5:55 ass [ __ ] so here's what we're gonna do
5:57 here's what we're gonna do
5:59 we're gonna we're gonna go on ahead I'm
6:01 gonna make some recommendations so if
6:03 you're new to this AI stuff and you like
6:06 Macau where do I start I don't know
6:08 where to start I'm so confused
6:10 everyone's talking about a I don't know
6:12 what they're talking about this is very
6:14 confusing are the robots gonna kill us
6:17 what do I do all right
6:19 we're gonna get started we're gonna it's
6:21 gonna we're gonna bring it down a notch
6:22 just take a deep breath
6:24 this part of this channel is a Zen
6:26 breathing Channel I don't know if you
6:27 knew that
6:28 welcome welcome just take a deep breath
6:31 in in through the nostrils and out
6:33 through the mouth in through the
6:34 nostrils and out through the mouth okay
6:37 great now let's talk about AI
6:40 um so yeah so we're going to talk about
6:42 AI if you've not played with chat gbt
6:44 that's the URL right there
6:46 chat.openai.com that's the official one
6:49 it's the best right now there might be
6:52 some stuff around the corner uh
6:54 particularly something coming from
6:55 Google that I think could kick this
6:57 thing in The Shins but right now gpt4 is
7:00 is by far the best of these things you
7:02 have to pay for gpt4 here but I've got a
7:04 free option down below but even GPT 3.5
7:07 here is quite good so if you're brand
7:09 new to this stuff just go there all the
7:11 other [ __ ] I'm going to talk about you
7:12 can just ignore for now go there
7:15 and just start with a parlor trick like
7:17 you can ask it anything but start with
7:19 something like write me a country song
7:20 about you know describe your dog or your
7:23 significant other and just have it write
7:24 a country song and just just start to
7:26 throw challenges at it and watch what it
7:28 can do and then go to that URL down
7:31 there prompts.chat that will teach you
7:33 how to talk to these things these are
7:35 called large language models chat GPT is
7:37 is a large language model
7:39 and
7:41 um proms.chat will kind of tell you what
7:44 they are how to talk to them and it's a
7:45 whole long list of prompts you can
7:47 literally copy and paste where you can
7:49 have chat GPT act like different
7:51 personas it's really quite remarkable by
7:53 the way Shelton just won the first set
7:56 so in the tiafu match
8:00 um so uh so that's that
8:04 um Claude and Bard are are two that
8:06 offer some slightly different things
8:08 Bard's connected to the internet Claude
8:10 you can put really long prompts into I
8:13 was playing with with uh Claude today
8:15 working out some Workshop deals with
8:18 with a friend who's gonna host uh doing
8:22 a workshop so I'll there'll be more
8:25 about that coming up but I use Claude
8:26 today so those are the three I kind of
8:28 use there's four more I call this the
8:31 kitties table
8:32 so the kitties table is these are ones
8:35 that are still worth playing with but
8:37 are kind of lower on my list for a
8:39 number of reasons
8:41 um bing.com is from Microsoft that's
8:44 gpt4 for free
8:47 and
8:48 um
8:49 it's connected to the internet and you
8:52 can make images with it it's dollies
8:54 it's open ai's Dolly image engine
8:56 sitting underneath it and it can read
8:58 images so that's super cool
9:01 labs.perplexity uh lets you play with uh
9:04 meta's llama 2 models Pi like I said is
9:07 chatty GPT Poe lets you play with
9:11 um Bots if you don't know what Bots are
9:14 just go there and
9:16 hit the explore button and you can see
9:18 all these Bots that people have made and
9:19 you can make your own it's a really good
9:21 way to practice prompting okay so those
9:23 are all the large language models
9:25 Chachi PT is the biggie Claude and Bing
9:28 are you know those three I sort of go
9:30 between those three to play
9:32 um these four are also worth exploring
9:35 and then for images but you're like but
9:38 God I want to make pretty pictures I
9:40 don't know I'm like pretty pictures Kyle
9:42 it's not fair all you're talking about
9:44 is words I don't like Words there's so
9:48 many words I only know a few Kyle what
9:51 about the pictures here's the pictures
9:54 calm down like I said this is a deep
9:57 breathing exercise you're gonna be okay
10:00 and in through the nostrils out through
10:02 the mouth in through the nostrils out
10:05 through the mouth fantastic Bob tell him
10:07 what he's won you've won a new car all
10:09 right
10:11 um mid-journey really good best in terms
10:14 of image quality fan fuckin-tastic
10:20 I have a feeling they're going to get
10:21 their asses kicked in the next year
10:22 because they're only an 11 person
10:24 company and their CEO is a little bit
10:25 full of himself and they could be in
10:27 trouble but right now they got it you
10:30 have to make images in Discord which I'm
10:32 happy to show if someone wants to see
10:34 how you do that it's a pain in the ass
10:35 but it just is the best ideogram this
10:38 one right next to it is like 10 days old
10:40 this is a brand new service they figured
10:43 out how to put text in images which
10:45 you're like but Kyle that's not a big
10:47 idea there's always text and images no
10:49 there's text and images when
10:50 typographers and designers go into
10:52 illustrator and type them in there but
10:54 you can't just say give me a poster with
10:57 an eye on it that says AI learning lab
10:58 well now you can
11:00 and that's what ideogram is
11:02 and it is pronounced ideogram I heard
11:05 someone I heard the founder talk about
11:07 it and then Leonardo is a really cool
11:10 website sitting on top of stable
11:11 diffusion and clipdrop.co is is a bunch
11:14 of tools from stable diffusion from
11:16 stability AI
11:18 and then and then prompt book
11:21 openart.ai prompt book is just like
11:24 prompts.chat for words this is the one
11:27 that teaches you how to talk to these uh
11:29 image things
11:31 um all the all the hearts that you're
11:33 seeing flying up on the screen that's
11:34 the Irregulars the ones that come in
11:36 here every night the sickos the sickos
11:38 that are just in here every night
11:39 they're playing a game they're trying to
11:41 beat their last night they hit like 75
11:43 000 likes in a two and a half two and a
11:46 half hour live so so just don't worry
11:50 about that you feel free to join in tap
11:52 the screen all you want just you know
11:53 the the
11:55 the Irregulars they've got they've got
11:57 wrist issues they have carpal tunnel and
11:59 so they would appreciate the support so
12:01 feel free to tap the screen uh feel free
12:03 to also share the live
12:04 um and then futurepedia.io is just it's
12:08 a directory it's a directory of like
12:09 four thousand forty two hundred uh AI
12:13 tools and so a lot of times in here
12:17 people will ask me about stuff that I
12:18 don't have a flipping clue about I'll
12:20 just keep pointing You Back to Future PD
12:22 if you want something really specific
12:23 like what's the best tool to I don't
12:25 know analyze legal documents there's
12:28 probably seven of them in there all
12:29 right
12:30 okay so with that
12:32 pop any questions you have in the
12:34 comments below
12:35 I'm gonna start answering them I can
12:38 guarantee you I promise you I will get
12:40 behind
12:41 um I haven't gotten my [ __ ] together to
12:43 get moderators in here yet but you know
12:44 I'm getting it I'm just uh I'm just
12:47 doing the best I can people
12:50 um but yeah but pop pop any questions
12:53 you have below I will do my best to
12:55 answer them and let's have some fun
12:56 we're gonna have some fun tonight you
12:57 know what I mean you know what I'm
12:59 saying this is how we're doing it
13:01 all right
13:02 hello can I make games using AI we
13:04 started with that one you yes you can in
13:08 fact
13:10 was it
13:12 I think you can also just have it just
13:14 tell it to make a game of snake in
13:16 Python and it'll do that and go run it
13:17 and ripple it
13:19 play Blackjack versus chat EBT that's
13:22 really interesting chaos Generation The
13:24 Secret of AI it's Turtles all the way
13:26 down oh yeah
13:29 I like turtles
13:31 oh yeah Turtles is tasty
13:36 yeah yeah
13:40 I like it I don't know what's going on
13:44 with my voice tonight it's got It's it's
13:46 deep and Gravelly there's something
13:48 going on yeah today I know Bard oh God
13:53 fantastic ride on Mr Shakespeare it was
13:57 fantastic
13:59 all right let's see my dog is my
14:04 significant other dogs are the [ __ ]
14:06 best
14:08 um higher regulars and Kyle hello hello
14:10 hello
14:10 write effective that can do what others
14:14 don't what's effective effective AI
14:17 Translate
14:19 wrote a
14:22 I don't know what that
14:25 wrote a something AI
14:29 that'll do what others don't I want to
14:31 hear about that workshops yes and so so
14:34 um
14:36 so I did a I did a podcast uh with a
14:39 woman named Rhonda I forget her last
14:41 name she's a voice over actress and so
14:44 we're putting together a three-session
14:46 workshop for creatives so for voice over
14:49 actors and other creatives
14:51 um so I'm going to start with that one
14:53 that's gonna we we decided tonight it's
14:55 gonna start October
14:58 fourth on it's which is a Wednesday
15:01 night from five to seven mountain time
15:03 and it's going to be three sessions
15:05 every Wednesday so it's going to be the
15:07 fourth the 11th and
15:09 the 19th 18th 18th
15:12 of October so that's coming um so as
15:14 soon as I have that sort of packaged up
15:17 I'll I'll let you guys know about it
15:19 um and then I'm gonna do a specific one
15:21 for the it for the AI learning lab
15:23 that'll that'll be probably a bit more
15:26 business focused but I'm but I'm going
15:27 to do it based on what folks here have
15:30 asked for so so anyway
15:32 um I had the obligation to do that one
15:34 with her first so we're gonna do that
15:35 it's gonna be cool if you want to take
15:37 that one it's going to be good
15:40 um yeah it's gonna be a good one it's
15:42 gonna be a good one Focus primarily for
15:43 creatives all right
15:45 fantastic yeah we're gonna do it we're
15:48 gonna do it Kyle win workshops October
15:50 4th
15:51 are you cats using typing mind yet no
15:54 let's go learn what that is
15:57 this is the other reason like this thing
16:00 is so cool is like it's just you you
16:03 can't keep up with it so
16:05 anytime someone says like have you seen
16:07 this no
16:10 tight mind a better UI for chep GPT
16:12 interesting
16:15 interesting
16:18 well there's more [ __ ] on it
16:21 oh you can change the temperature
16:25 yeah this has got some interface on it
16:27 you can search your chats
16:30 it shows you your API keys right there
16:32 yeah this is pretty good
16:34 so basically what they did what what
16:36 type mind is or typing mind
16:41 is uh they took all the [ __ ] that old
16:44 Sam Altman and gang should have done six
16:48 months ago they're like why wait we can
16:51 write our own [ __ ] it so there you go
16:56 that's awesome yeah good for them
16:59 you can change your you can select your
17:00 model
17:02 you can change your profile that's super
17:04 cool
17:06 your info your information your profile
17:09 image I wonder if that uh
17:12 impacts your your prompt
17:16 by what's by mean
17:20 buy once use forever price increase soon
17:26 I I uh I bought a piece of software once
17:30 that was a it was a lifetime
17:33 I paid whatever it was the 49 bucks for
17:35 the lifetime subscription and then at
17:37 some point like two years later they're
17:39 like everyone's got to upgrade and pay
17:41 this new fee for the new version and I'm
17:44 like but I bought a lifetime
17:45 subscription and they're like no that
17:47 was the lifetime of that version
17:50 I'm like oh
17:52 [ __ ] you no that's wrong
17:56 that was that was uh
18:00 what was it was clean my Mac I think it
18:02 was Mac Paws clean my Mac [ __ ]
18:06 but anyway
18:08 that's pretty cool
18:10 um
18:10 I mean 39 bucks for a lifetime thing
18:12 assuming it's actually a lifetime thing
18:14 I would check that
18:16 um premium license giveaway
18:19 um I mean here's the deal with with
18:21 something like this because this is kind
18:23 of a slight augmentation to a thing
18:25 that's already there
18:27 when open AI gets their [ __ ] together
18:29 and fixes chat GPT it's likely going to
18:32 have a lot of this stuff like like right
18:33 now Chachi BT is missing like a lot of
18:36 what I see here is the stuff that drives
18:38 me nuts that's not in Chachi PT so I
18:41 like it but this is one of those
18:42 companies where this will just get you
18:44 know this just becomes features of the
18:46 real deal so anyway interesting
18:50 um thanks for sharing that uh it's a
18:52 home base for all the big llms you can
18:54 organize your chats sets of lots of
18:56 personas temperatures yeah it looks
18:57 pretty cool
18:59 um
19:00 it's like the French have a different
19:02 word for everything Steve Martin joke
19:05 thank you chaos generation
19:08 you ever notice when you go to a
19:09 different country
19:12 how does he say it I'd like to go to the
19:14 hotel please and the taxi driver turns
19:17 around
19:22 and he goes and he goes I would like to
19:25 go to the hotel please
19:29 uh
19:31 uh
19:32 feels so good
19:34 so good what do you think of prompt hero
19:37 I'd love heroes heroes are the best
19:39 prompt hero I don't know anything about
19:41 I think I've heard of it but let me
19:43 let's go look at it
19:44 prompt hero
19:47 search prompts for stable diffusion chat
19:49 gbt oh is this the one where you can
19:51 sell your prompts
19:53 no that's prompt bass right
19:57 number one website for prompt
19:58 engineering search millions of AI art
20:00 images by the models by models like
20:03 stable diffusion mid-journey
20:06 so there's mid-journey ones
20:09 yeah this is pretty cool oh there's
20:11 there they're just not selling it yeah
20:13 prompt hero looks pretty good
20:15 AI models
20:17 loros
20:20 this looks like it's a well put together
20:21 site featured hot
20:25 yep
20:28 prompt hero.com that's worth that is
20:31 worth checking out
20:34 that is worth checking out in fact let
20:37 me see AI models
20:39 large language models
20:43 why are there only two
20:52 two prompts here
20:57 looks like the language model prompts
20:59 aren't really functional right now
21:05 oh best chat CBT prompts
21:13 huh
21:15 no results
21:19 so like all of this stuff so so one of
21:21 the things one of the things that I'll
21:23 share with you
21:24 all of the tools all of the tools
21:27 all of the tools right now
21:30 all of the tools whether it's chat EBT
21:33 or Bing or Bard or Claude or
21:37 uh
21:38 pi
21:40 or llama
21:42 or prompt hero all of them right now are
21:47 wait for it
21:49 janky pieces of [ __ ] one of the
21:52 Irregulars in the uh in the salon
21:54 Discord made me a janky pieces of [ __ ]
21:57 logo
22:00 um and here's what I mean by that it's
22:02 not that the developers that make all
22:04 these apps are bad it's that November
22:07 30th 2022 is ostensibly when
22:12 the popular
22:14 AI industry started so that was nine
22:17 months ago
22:19 so nine months ago Chachi PT launches
22:22 and this is sort of this fork in the
22:23 road where you know before that it was
22:25 just all scientists and researchers and
22:27 mathematicians and after that it's the
22:29 rest of us right and so that this
22:30 channel is about you know that moment
22:32 and what now that now that these tools
22:35 are in the hands of mere mortals what
22:36 the [ __ ] do we do with them
22:39 um just know that as powerful as these
22:41 tools are and as much noise as they're
22:43 making and as much as people are talking
22:44 about them they are still very very
22:46 early so they're very raw they don't
22:49 have a lot of sophistication they make
22:51 mistakes they hallucinate they make
22:54 hands that look like this sometimes they
22:56 make letters that do not look like any
22:58 letter you've ever seen But if you use a
23:00 tool like ideogram you can get it to
23:02 make janky pieces of [ __ ] content
23:08 um so just know that that uh we're still
23:11 very very early and that's it's okay
23:14 it's okay it's all right it's gonna be
23:16 okay seriously it's gonna be all right
23:18 all right
23:19 okay all right
23:21 all right yeah yeah we're gonna go on
23:24 ahead and need you to come on in on
23:26 Saturday oh yeah and another thing Peter
23:29 I need you to go on ahead and come on in
23:32 on Sunday too yeah
23:36 um
23:37 one of my all-time favorite movies one
23:40 of my all-time favorite movies
23:43 I have a red stapler at work
23:48 uh so you can do prompting with multiple
23:50 models all in one place oh that's pretty
23:52 cool that's the uh what was that uh
23:57 was that one we just went and looked at
23:58 what was it what was it
24:04 um
24:08 type mind.ai yeah I'm gonna bookmark
24:11 that one hang on
24:13 notify me enter your email
24:21 the biggies all right I'll just drop it
24:23 in here and then prompt here I'm gonna
24:25 drop in here too prompt hero it looks
24:29 new it looks like they're still building
24:31 it but that's okay
24:32 a lot of these things are still building
24:34 so all right dually bookmarked
24:40 Fran she asked yet what's up irregular
24:43 share bears in the house
24:45 There's Tom Baldridge I'm a sicko
24:48 welcome Tom you're not a sicko you're
24:51 irregular
24:55 [Laughter]
24:59 Prince hello world can you make money
25:02 with chechi PT
25:04 oh I don't know can you
25:11 why yes you can
25:13 but not the way you think
25:16 oh I'll talk to you about that if you
25:18 really want to know if you want to know
25:20 the secret of making money to chat GPT
25:22 just send 39.95 to this PO Box in Sioux
25:26 City Iowa
25:31 all the secrets are contained within
25:37 did it just jump me to the bottom oh
25:40 Tick Tock you and your godforsaken bug
25:45 see here's the deal I'm I'm neuro spicy
25:48 and
25:49 um I like answering all the questions
25:51 there's many as I can and then Tick Tock
25:53 occasionally just jumps me to the bottom
25:55 for because it's got a bug
25:57 and it makes me cranky it makes me
25:59 cranky and every time it happens I say
26:02 the same thing and does tick tock listen
26:04 no they're listening to all the words I
26:06 say because God forbid I say a word that
26:08 might offend someone they shut my
26:09 [ __ ] life down for seven days but if
26:12 I say things like fix your bug
26:14 no not happening
26:18 hey I just got a note from
26:24 oh
26:27 laricia from my company is she's
26:30 discovered the NPCs you know the people
26:32 that are that are like ice cream ice
26:35 cream ice cream
26:38 I sort of do that here when people give
26:41 me the uh cowboy hat but
26:44 um but yeah I'm not okay I found this on
26:46 the web for to do Siri when people give
26:48 me the cowboy hat but check it out
26:51 there Siri really
26:55 she's jealous she's Sherry's jealous of
26:58 the Irregulars it's so true can you
27:01 recommend any gen AI for visualization
27:04 like create a multi-layered donut with
27:07 this data
27:09 well so yes so let's let's go play
27:13 um
27:14 it's not super sophisticated
27:17 visualizations but that's also because I
27:20 don't know what the [ __ ] I'm doing so
27:23 we're gonna go to Chachi PT
27:27 um
27:29 so here we are in chat EBT if you
27:31 haven't seen chappie gbt before these
27:33 are all the prompts these are your
27:35 different models you have 3.5 and 4 and
27:38 then under gpt4 you've got some options
27:41 here you've got default Advanced uh data
27:43 analysis and you've got plugins okay
27:47 um
27:50 in order to get to Advanced Data
27:52 analysis that's what I'm going to show
27:53 you you've got to be in chat GPT plus
27:55 which is the 20 Buck a month thing and
27:58 you have to turn it on so you go to your
28:00 settings and then you go to Beta
28:02 features and you turn it on so if you if
28:05 you have paid your 20 bucks but you
28:08 don't see Advanced data analytics or
28:10 plugins you haven't turned them on okay
28:14 so Advanced data analytics this also
28:17 used to be called code interpreter so
28:19 once you select it you now have this
28:21 little plus button you can upload data
28:24 and so I'm going to go grab what was
28:28 archive 10
28:30 no that's not what I wanted
28:33 was it archive nine
28:36 we'll grab archive seven although I
28:38 don't know how big it is so this is just
28:40 some data I grabbed off the internet
28:42 there's a site called kaggle just
28:45 horrible name for a site I don't I don't
28:48 care what it is but it's where you can
28:49 it's where you can get data sets
28:52 so you go play with data sets and uh
28:59 get them at kaggle and so I'm going to
29:01 say
29:02 um tell me what
29:06 data is in this archive
29:10 what's
29:12 interesting
29:15 about it
29:18 and how you might visualize it
29:24 and so what's going to happen now is
29:28 all right let's start by examining the
29:30 contents in the provided archive
29:33 so what happens with Advanced Data
29:35 analysis
29:37 is that when you launch the chat it's
29:39 actually spinning up a virtual machine
29:42 that includes its own set of python
29:46 libraries and and python so it's
29:49 basically a virtual computer and then if
29:51 you open these little boxes it says you
29:53 know I'm doing this work now you can see
29:56 what it's actually done is it has
29:58 imported
30:00 libraries to open a zip file written its
30:03 own python to extract what's in the zip
30:05 file and then looked at what's in there
30:08 so it's it's game data related to it's
30:11 data related to games and this is
30:14 profile related data so now it's going
30:16 to inspect a few entries from each Json
30:19 file
30:20 to understand the structure of the data
30:25 I forget what this is I think this is
30:28 other quite large causing a memory error
30:30 to circumvent this I'd like to take
30:32 another approach
30:33 I'll read the files line by line so it
30:36 wrote some python
30:38 that python failed because the virtual
30:40 machine is too small
30:42 and so it's trying another thing now now
30:45 it run it ran into the same issue let's
30:47 take a more cautious approach
30:49 I think this is going to fail as well
30:53 I'm gonna go grab a different piece of
30:55 data here's a brief overview okay
30:59 now it did it it did it
31:01 player ID oh I think this is chess games
31:04 oh no rushing yards this is football
31:06 yeah this was this was one night someone
31:09 asked me you know can you
31:14 can you do uh like like fantasy football
31:17 player you know player selection using
31:20 chat GPT
31:21 all right
31:23 how might I visualize it player
31:25 performance a time chart series showing
31:27 a player's performance metrics like
31:29 passing yards over time team performance
31:32 distribution of player attributes
31:34 players per team uh
31:37 let's
31:40 visualize visual
31:44 allies
31:47 team performance
31:50 by
31:51 the Broncos
31:54 and the New York Giants
31:58 um
32:02 that's just and and we'll say
32:05 um
32:07 make it visually stunning
32:13 and so now what it's going to do
32:16 if we do it right
32:19 is it's going to go actually analyze the
32:21 data it's going to pull that data into a
32:23 data frame within the virtual machine
32:25 it's then going to write some python to
32:27 visualize and I didn't tell it what kind
32:30 of you wanted a what did you on a
32:32 multi-layer donut it might be able to do
32:34 that I don't know enough about data
32:36 visualization to you know do that
32:38 without screwing it up so I just tell it
32:41 make it pretty and it's pretty good at
32:43 making it pretty
32:45 by the way tiafo's upper break in set
32:48 two against Shelton Ben Shelton
32:51 something he's going to be he's he's a
32:52 flipping Star him and and alcaraz
32:55 I'd like to see Shelton win this and
32:58 play alcaraz I think those two be
32:59 explosive together
33:04 all right let's see I apologize for the
33:06 inconvenience the memory issue persists
33:09 let's enjoy employ a different strategy
33:11 we're going to break it into smaller
33:13 chunks so again like one of the things I
33:15 find remarkable about this is
33:18 I mean you saw what little input I put
33:21 into it it is prompting itself now it's
33:24 doing its own problem solving to to
33:26 figure out the data in this in this set
33:29 you can have this thing analyze images
33:32 you can have it analyze PDFs you can
33:34 have it analyze video files you can
33:37 change color spaces of of multimedia
33:41 files you can
33:43 um
33:44 turn an image into a zoomed you know
33:47 panning animation across the image
33:51 um
33:53 it's really apologizing I sincerely
33:56 apologize for the inconvenience we're
33:58 still encountering memory issues
34:01 I'll use the Json Library which allows
34:03 us to parse Json files so now it's just
34:05 picking a different python library and
34:07 see if that works that's kind of
34:09 interesting
34:13 no it can't do it it's it's let's see
34:23 I did a cool thing
34:25 a couple of weeks back where I took I
34:27 brought in and I imported an image I had
34:30 it determine the top 15 average colors
34:33 of the image and then create a pie chart
34:35 with all of the slices of the pie chart
34:38 in the color of the of the most
34:40 prevalent colors of the image
34:42 and then I had it name the colors
34:45 like a creative Crown color and it did
34:48 it brilliantly it was pretty beautiful
34:55 are we doing it are we doing it did we
34:57 fail
35:00 what's it got going
35:05 it's chunking it into 1024 by 1024
35:08 chunks
35:10 one megabyte chunks I assume it it knows
35:14 that it can keep that in memory
35:15 although it looks like it might have
35:17 hung
35:20 anyway
35:22 Um this can do it let me go let me go
35:24 grab I'm just going to bail on this
35:25 let's go do let's go do a new thing I'm
35:27 gonna go grab a different piece of data
35:29 that's smaller
35:31 uh archive six that was archive seven I
35:33 mean what's cool about this I'm just
35:34 grabbing random things
35:36 um let's
35:38 the data
35:40 data
35:42 in here and
35:45 um
35:47 what is interesting and how might we
35:52 visualize
35:55 it
35:58 and it's still up uploading there we go
36:02 all right all right all right all right
36:04 all right all right so anyway tktk HMO
36:10 we'll try this with one other data set
36:12 we'll see what this one is
36:24 Spotify data
36:27 let's proceed by loading the CSV file so
36:29 so it found a thing called Spotify data
36:32 so it understands that it's Spotify data
36:35 it might contain data related to Spotify
36:37 such as song details play accounts user
36:40 preferences
36:41 the data set contains details of songs
36:43 on Spotify artist name track name track
36:46 ID popularity score of the track year
36:49 the song was released genre danceability
36:51 it's got we've got danceability data
36:54 here kids and energy data and key and
36:57 loudness and mode
37:00 mode is minor or major speechiness
37:04 presence of spoken words in a track
37:06 acousticness this is really cool
37:10 valence wait what's balance valence
37:13 valence
37:15 what's valence valence measure the
37:17 musical positiveness
37:20 so I guess you know the more minor and
37:22 said all the all the songs I know I'm a
37:24 singer-songwriter all my songs are
37:26 depressing so my valence score would be
37:28 very low
37:30 duration in microsec in milliseconds
37:32 time signature
37:34 considering the columns available there
37:36 are several interesting visualization
37:38 distributions of songs by your
37:39 popularity distribution
37:42 average song characteristics per genre
37:46 correlation Matrix
37:48 so let's do
37:50 um visualize
37:52 number four so it it knows what number
37:55 four is it gave us a fourth option there
37:57 visualize number four in a visually
38:00 stunning way let's see what it does
38:08 it's gonna do a heat map we're going to
38:10 use colors to represent the degree and
38:12 direction of correlation between the
38:14 variables to make it even more visually
38:16 stunning stunning we can use a diverging
38:19 color palette to make the positive and
38:21 negulation negative correlations stand
38:24 out
38:25 so like even if you know nothing about
38:27 data visualization and and data analysis
38:31 like I don't
38:33 it'll just do it there you go
38:36 so it's got correlation Matrix heat map
38:40 popularity
38:41 so it's all all the different attributes
38:43 I guess for all the different songs
38:46 and where they correlate and where they
38:48 don't
38:49 amazing
38:51 so anyway yeah
38:54 danceability and Valence show a positive
38:57 correlation implying that the tracks
38:58 that are more danceable tend to convey a
39:00 more positive or happier mood oh this is
39:02 cool so so some observations from the
39:05 heat map darker Shades in either
39:07 direction indicate stronger correlation
39:09 a diet a diagonal is always one because
39:13 any variables are perfectly correlated
39:15 with themselves
39:17 features like energy and loudness show a
39:20 positive correlation indicating that
39:21 tracks with higher energy tend to be
39:23 louder that makes sense acousticness
39:26 shows a negative correlation with energy
39:29 and loudness suggesting that more
39:31 acoustic tracks tend to be softer and
39:33 have less energy James Taylor
39:37 Martin Sexton John Hyatt
39:40 Brandi Carlisle breaks that though
39:42 because she's she's acoustic and she
39:45 rocks her ass off anyway that's uh
39:48 that's Advanced data analytics so that's
39:50 that's the way you can do that just pay
39:52 your 20 bucks turn it on and start
39:55 popping data in there isn't that amazing
39:58 it ain't amazing and amazing all right
40:00 I'm now 82 messages behind because we
40:03 went down the little rabbit hole there
40:04 so sorry about that I'm not weird I'm a
40:07 baby though
40:08 um
40:13 I like turtles okay
40:17 can I make games using AI we talked
40:19 about that
40:21 dude
40:27 I I will avoid put the title on some
40:30 mid-journey generated pictures don't
40:33 like trunk traveling what is recommended
40:36 best AI for coding
40:39 so
40:41 gpt4 is really good
40:44 although I saw a thing today the gpt4
40:47 has gotten much worse I think Stanford
40:51 released a study where they they looked
40:52 at the quality of coding of gpt4 and
40:55 it's gone down pretty dramatically and
40:58 at the same time but but it's still good
41:00 so go there Bard is supposed to be good
41:03 at coding bard.google.com
41:06 and then
41:10 um
41:12 code llama if you go to
41:14 labs.perplexity.ai code llama
41:17 the code llama model is supposed to be
41:19 better than gpt4 at coding
41:22 so I would say bard
41:25 llama right there they've hosted it for
41:29 you and gpt4 all all three of those are
41:32 going to be some version of
41:34 better than I can code
41:37 which is which is the bar is very low
41:41 I know what code is Wolfman Jack Wolfman
41:46 Jack wesn Radio Rock hip-hop and smooth
41:50 jazz
41:53 Kyle the Irregulars are scheming to send
41:55 you hats at the top of the hour giddy up
41:57 all right what time is it how are we
41:58 doing on all right I got it's 42 so I've
42:01 got what is that 15 18 minutes until the
42:04 Hat damn I got it okay see I know you
42:08 said
42:10 I didn't prepare
42:12 I gotta get some John Wayne quotes up
42:16 all right hang on
42:18 I'll I'll be a little prepared John
42:21 Wayne
42:22 quotes
42:26 all right
42:30 courage is being scared to death and
42:32 saddling up anyway
42:34 all right
42:36 life's hard but it's harder if you're
42:38 stupid that's not actually the quote the
42:40 quote is life's tough
42:43 it's tougher if you're stupid
42:46 that one I learned the last time let's
42:48 see
42:51 talk low talk slow and don't say much
42:56 all right I got some wait 50 all-time
42:59 favorite John Wayne quotes
43:04 expand to show full article
43:10 all right all right I got some quotes
43:12 there I'm prepared fine share bear
43:16 call me out with the with the hat Ambush
43:18 fine fine
43:21 I see how this works
43:24 I'm no longer in control of this live am
43:26 I
43:29 but Kyle you never were
43:33 ah you kids your kids is funny
43:40 um
43:41 she's got a funny face let's put her
43:43 we'll put her front and center
43:46 why is there a creepy lady over his
43:48 shoulder it doesn't make any sense to me
43:51 this is disturbing
43:57 all right
43:59 Turtle girl is here Tyron truth what's
44:02 happening fiction derived from fiction
44:04 hmm maybe don't know what that's in
44:07 reference to but yeah best app to build
44:09 spreadsheets with formulas to merge with
44:11 Excel sheets or QuickBooks
44:14 um it depends what you're trying to do
44:20 Advanced data analytics that thing I
44:22 just showed you can tell it so I've
44:25 pulled data into that and then had it
44:28 analyzed the data
44:29 and then clean up the data and rewrite
44:32 it in the order that I want in sheets
44:35 within an Excel sheet so so Advanced
44:37 data analytics can can generate CSV
44:41 files or Excel worksheet or Excel
44:44 spreadsheets with multiple worksheets in
44:46 them it's pretty amazing
44:48 um so it can do that
44:50 um there's also for sheets thank you
44:53 very much there's also for Sheets a
44:54 thing called GPT for Sheets plugin
44:57 and in every cell of your sheets you can
45:00 call open AI so you can call it to do a
45:04 translation you can call it to do an
45:06 open AI prompt you can call it to
45:08 generate an image
45:10 so that's pretty amazing
45:14 um
45:15 what what do you want oh you want to go
45:18 see Mama what do you want to do oh a
45:20 little whiny bumper
45:30 all right
45:32 let's see
45:36 um
45:37 I don't know if it can do Quickbooks
45:39 data but I assume I mean QuickBooks you
45:41 can import csvs right so
45:45 um or Excel files so
45:47 and it might even be able to like if you
45:50 if you grab the data from a from a
45:53 QuickBooks file it might be able to read
45:57 that and rewrite it and and I don't know
45:59 it might be able to even output that
46:01 if it's a pretty standard format that
46:03 QuickBooks does I mean it's been around
46:05 for decades so
46:06 so anyway so I'd go play with Advanced
46:08 data analytics or the GPT for Sheets
46:11 plugin
46:12 for that stuff I'm not good at Excel or
46:15 I'd show you stuff I'll just embarrass
46:17 myself anything involving math
46:20 yeah it's a it's a it's a situation
46:25 all right yep and folders Kyle you said
46:28 subscribe short term as this stuff is
46:30 constantly changing it's new correct
46:34 I did so so that's I think that's in
46:37 reference to that to that site saying
46:39 it's got a lifetime thing if if it's got
46:42 a lifetime subscription I'm a little
46:44 more lenient on that but but I think
46:45 right now spending money on
46:49 yearly subscriptions is is silly because
46:52 this stuff is changing so fast like the
46:55 the the speed with which these things
46:58 are are just changing thanks shiver this
47:01 the speed with which these things are
47:02 just changing even on my list of just
47:04 like you know what what I'm finding
47:06 useful personally is just I can't keep
47:08 up with this [ __ ] so
47:10 you know if I had subscriptions to all
47:12 these I'd just be kicking myself because
47:14 every time one sort of drops off the the
47:16 you know
47:17 the big kids list
47:20 um it makes me cranky that I paid for it
47:22 but you know let's see tiafo won the
47:25 second set so one set a piece
47:28 Shelton and tiafo
47:30 all right
47:32 there's my John Wayne quotes for when
47:34 the when the Hat attack comes all right
47:40 Omar domain if businesses are concerned
47:43 about data privacy and public chat gbt
47:45 what instead should they use good
47:48 question so there's a couple of things
47:51 chat GPT just announced GPT Enterprise
47:54 which right now
47:56 you have to be spending ten thousand
47:58 dollars a month for them to even
47:59 consider you so if you're a big company
48:02 chat GPT Enterprise
48:05 um
48:06 you've also got you can you can spin up
48:09 custom instances of open AI
48:14 um servers in Microsoft azure
48:20 so you so you can do things there the
48:22 other thing you can do is go to the open
48:25 source Community you now have
48:28 um the Llama 2 models and the new code
48:30 llama model from meta these are fairly
48:34 significant large language models that
48:36 are free for people to use so you can
48:39 set up your own servers whether it's a
48:41 virtual machine on you know some
48:44 cloud-based thing where you have access
48:45 to gpus or you've got your own Iron with
48:48 big gpus on it you can run this [ __ ]
48:51 locally so you're not doing any calls
48:53 externally at all now you're gonna have
48:54 to roll your own you're gonna have to
48:56 fine tune it you're gonna have to figure
48:57 all that [ __ ] out but if what you're
49:00 trying to figure out is data privacy and
49:02 security there's there's the concerning
49:06 phallic
49:07 elephant ears thank you for that it
49:10 throws me every time but I like it it's
49:12 solid it's good it's it's a good
49:15 animation it's a it's solid 3D mapping
49:18 to to managon in my machine
49:24 um
49:26 yeah but but I would do that so I would
49:27 I would look into the the chat GPT
49:29 Enterprise and and they say that uh GPT
49:33 Enterprise is going to be
49:36 um there's going to be a self-service
49:38 version of it
49:39 that's coming
49:41 um so so at some point after they after
49:44 they get done spinning up their ten
49:47 thousand dollar a month clients then
49:49 they'll they'll they'll they'll come to
49:51 the rest of us groveling for data
49:53 privacy and chat GPT
49:55 um like I said all of these tools right
49:58 now are
50:01 right for it
50:04 janky pieces of [ __ ]
50:07 laughs they're getting less janky but
50:11 they're still
50:12 they're still improving let's say
50:17 um
50:20 [Music]
50:26 all right
50:28 oops scrolled a little overzealously it
50:30 happens I do it all the time cool thanks
50:33 on the thumbs up on prompt hero yeah
50:34 that looks pretty cool
50:36 uh
50:37 um
50:38 Thelonious damn smartypants no chat GPT
50:41 prompts yeah it's so I I would assume
50:44 that prompt hero is like so what they're
50:46 doing is smart they're saying okay
50:48 there's going to be lots of different
50:49 models so let's create an info let's
50:51 let's create a clean interface where you
50:54 can find good prompts from lots of
50:55 different things but they gotta spin
50:57 spin all that up and I would assume that
51:00 there's not a prompt
51:01 API from chat GPT yet so they're
51:04 probably gonna have to scrape some [ __ ]
51:05 and whatever it's it's probably a thing
51:08 is my guess it'll get there how's
51:11 everyone been Joker I've been good can
51:14 my husband be an honorary irregular
51:16 because I walk around the house
51:17 listening to Kyle every night I'm so
51:20 sorry he is absolutely welcome to be an
51:22 irregular but you know he's got to put
51:24 up with our nonsense so ask okay okay
51:27 farmer and and uh Tobias they they make
51:30 it they make date night out of AI
51:32 learning lab lives
51:35 uh I was working and didn't realize we
51:38 went live it happens I sometimes we're
51:41 live and I don't even realize we've gone
51:43 live you know
51:44 nine years old is wait no nine-year-old
51:48 is good at this prompt said me and you
51:50 are too old for this we are too old for
51:53 this but screw them
51:54 this is I'm telling you so so here's the
51:57 thing that I I I deeply deeply deeply
52:00 believe this
52:02 these AI tools
52:05 revenge of the liberal arts major is
52:07 coming
52:08 and revenge of the Gen xers is coming
52:11 because I think having a broad view of
52:13 things is going to be a superpower and I
52:16 think being on this planet for extra
52:18 decades is going to be a superpower why
52:21 because these tools these fancy ass not
52:26 those
52:28 these fancy ass large language model
52:30 tools
52:32 they produce content so quickly
52:35 that if you've been on the planet for a
52:37 while we're going to be able to look at
52:39 the what they produce and go that's good
52:41 that's bad like that
52:43 so
52:45 our life experience and then if you've
52:47 got a broad view of the world if you're
52:49 not sort of hyper focused on a single
52:51 thing
52:52 again you're going to be able to sort of
52:54 context shift and now I'm copywriting
52:57 now I'm making images now I'm making
52:58 music now I'm doing project planning now
53:01 I'm doing creative writing right boom
53:03 boom boom boom
53:05 [ __ ] Field Day for ADHD years Field
53:09 Day for Gen xers and and you know you
53:12 know Millennials older millennials
53:15 Field Day for
53:17 um
53:18 liberal arts people
53:20 horizontal thinkers
53:22 we're having a [ __ ] party with this
53:24 stuff
53:25 so yeah you can tell your nine-year-old
53:29 well he's dying say thanks honey you're
53:33 right I love you that's that's what you
53:35 should tell your nine-year-old and then
53:36 you can secretly say
53:41 what's really going on
53:48 all right
53:51 [Laughter]
53:53 Silver Fox I just call Kyle my guy
53:55 that's when I was watching my guy and
53:57 yeah sorry
53:59 [Laughter]
54:02 office space is the best TPS reports
54:05 ASAP Lumberg
54:07 I made that
54:10 you made office space did you work on
54:12 office space Ted looks at life
54:14 oh my God just like
54:18 I just I I like every flipping scene of
54:23 that movie has a memorable line in it
54:28 you don't have you don't have on the
54:30 required amount of flare flare
54:34 it's like oh
54:40 I'm gonna need you to go on ahead and
54:41 come on in on Saturday yeah
54:46 you've been missing a lot of work Peter
54:48 I wouldn't call it missing it
54:52 well that's good that's good your
54:54 management material
55:00 oh my goodness That's a classic it's a
55:04 classic it's a real cup Tick Tock just
55:06 jumped me to the freaking bottom
55:09 this is like a bad SNL skit that's not
55:12 funny that I have to react to every time
55:14 like I'm some sort of pathological
55:17 Tick-Tock interface troll
55:20 [Laughter]
55:26 oh Tick Tock why
55:34 I saw Peyton here welcome
55:39 uh um
55:42 1032 EST Bingo what time is it
55:48 oh EST so that's 8 30.
55:51 that was ages ago I'm way behind
55:55 of course I am that's that's one of the
55:57 that's one of the Traditions around here
56:01 um
56:02 looks like a touch screen computer
56:05 it's a spin stand box container not
56:08 virtual machine oh wait it's a wait
56:10 it's spin a stand box
56:14 contain it
56:16 it's spin a stand box container I don't
56:19 know what a stand box container is not a
56:22 virtual machine
56:24 but
56:26 please educate me on that
56:28 well Chachi batiken too hang on
56:33 is that the right thing stand
56:35 box container
56:39 versus virtual machine
56:44 containers and virtual machines are both
56:46 technology used for virtualization and
56:48 resource isolation
56:50 they have distinct differences
56:53 in terms of architecture efficiency and
56:55 use cases here's a comparison
56:58 isolation and Resource Management
57:00 containers share the host operating
57:03 systems kernel okay
57:05 which means they're more lightweight and
57:06 efficient however the shared kernel can
57:09 potentially lead to less strict
57:11 isolation Okay VMS so so VM is is this
57:15 is really a standalone thing
57:19 containers are highly efficient
57:21 containers are known for their
57:23 portability
57:24 generally have better performance
57:28 containers often good isolation for most
57:31 use cases but not as secure as VMS Okay
57:33 cool so it's a container
57:36 it's a spin spin stand stand box
57:40 container not a virtual machine cool
57:42 thank you for that
57:44 all right
57:46 see you learn something new every day
57:48 here people this is about learning it's
57:50 the AI learning lab
57:52 we're learning
57:54 we start clueless and then we remain
57:57 clueless because someone else always
57:59 knows more and you know who knows even
58:01 more than those people
58:03 these things they know [ __ ]
58:06 everything hey a lot of partner wait is
58:08 it that time is that top it is it's that
58:11 time oh good lord hang on
58:14 the [ __ ] pressure y'all are killing
58:17 me
58:20 every man deserves a second chance but
58:23 keep an eye on him uh wait I gotta make
58:26 him bigger my eyes are old
58:29 uh tomorrow's perfect when it arrives
58:31 and puts itself in our hands it hopes
58:33 we've learned something from yesterday
58:35 John Wayne tomorrow is the most
58:37 important thing in life it comes into us
58:40 at midnight very clean it's perfect upon
58:43 arrival and this is not here's the
58:45 problem
58:45 [Music]
58:47 because I'm shitty at John Wayne
58:49 oh good lord
58:51 I'm in a flop sweat right now I'll tell
58:53 you what you know what little missy
58:55 we're not going with actual quotes
58:56 because you know what I got a generic
58:59 where's my hat where's there to go see I
59:03 got a generic Southern Cowboy kind of
59:05 accent and this is how I talk I'm gonna
59:08 talk like this and we're going to talk
59:09 about you know we're going to talk about
59:10 AI I don't know if you tried that AI but
59:13 AI is good ai's powerful ai's this is
59:16 like one of them awful NPC things hat
59:19 hat hat hat
59:21 oh dogs are fighting is that my uh
59:28 yeah I like it I like it I need I need
59:31 suspenders for that one do a little
59:33 dancing do a little jig
59:35 ding ding ding oh that was awesome thank
59:39 you sorry I you know I I you know I
59:43 should have written a script I should
59:44 have had chat GPT write me a script and
59:46 learned it bad actor
59:48 that's why I'm an X actor
59:53 thank you for that that was awesome
59:56 y'all y'all are swell
59:58 how big are the data files for it to die
1:00:02 um they they're not that big uh
1:00:05 the the the the
1:00:07 megabyte good wow welcome to the
1:00:13 planes the plane boss zip Lane
1:00:18 um did Shelton just break oh he did
1:00:20 Shelton just broke tiafu in the third
1:00:23 set
1:00:24 um
1:00:26 I think for code interpreter for
1:00:29 Advanced Data analysis the limit was 100
1:00:31 megabyte file but now they've changed it
1:00:34 so you can upload multiple files what I
1:00:37 don't know is is it a total of 100
1:00:39 megabytes or if it's 100 megabytes per
1:00:41 file I assume it's a total of 100
1:00:43 megabytes
1:00:44 because the fact that that one file I
1:00:47 think it was a 90
1:00:49 it was a 90 megabyte file and it kind of
1:00:52 choked it in its container
1:00:55 um so you know not huge amounts of data
1:00:58 again
1:01:00 like like the the Advanced Data and
1:01:02 analytics right now in chat apt you know
1:01:06 again I'm going to go back to this
1:01:09 thesis that all of these tools right now
1:01:11 are janky pieces of [ __ ] so it crashes a
1:01:15 lot it doesn't quite get it right A lot
1:01:17 it only does python right now it's only
1:01:19 got limited libraries right now they're
1:01:21 like they've probably picked the sort of
1:01:23 safest set of libraries
1:01:26 um
1:01:26 the the visualization libraries it uses
1:01:29 are fairly fairly unsophisticated
1:01:32 although it might just be that I'm
1:01:33 unsophisticated and don't quite know
1:01:35 what to ask for
1:01:37 um
1:01:38 so
1:01:40 so anyway so so I think it's decently I
1:01:43 think it's decently limited paid from
1:01:45 from what I can tell
1:01:47 uh certainly for someone you know who
1:01:50 who is used to dealing with big data
1:01:52 um it's probably comically small
1:01:55 but you know who knows who even knows
1:01:59 anymore does anybody really know what
1:02:02 time it is does anybody really care
1:02:07 that's a Gen X kids song remember
1:02:10 hearing that when you were a kid
1:02:11 [Music]
1:02:12 all right let's see
1:02:16 that was in a day we had three
1:02:18 television channels and two radio
1:02:20 stations and we liked it
1:02:25 you listen to whatever crap song our
1:02:28 parents liked
1:02:32 all right
1:02:34 I'm 73 messages behind by the way so if
1:02:37 you're wondering why isn't he answering
1:02:39 Michael an question it's just not fair
1:02:42 I don't even think he's live I think
1:02:44 he's an avatar I'm pretty sure he's GPT
1:02:47 five you know I don't know if you knew
1:02:49 this but they said gpt5 was gonna look
1:02:51 like people and he doesn't look normal
1:02:53 so I'm pretty sure he's gpt5
1:02:56 but he doesn't seem very smart you would
1:02:58 think gpd5 would be smarter wouldn't
1:03:01 don't you wouldn't you I mean I would
1:03:03 think that
1:03:04 but he's definitely not live because he
1:03:06 didn't answer my question I've asked it
1:03:08 73 times
1:03:12 I look like a Al Franken there for a
1:03:14 little bit
1:03:15 also that's like an Al Franken look
1:03:25 buffering a little for me just once
1:03:27 though all right well Tick Tock sucks
1:03:30 Xfinity sucks teacher here what app is
1:03:33 best for detecting student using AI to
1:03:35 complete written assignments
1:03:37 Catherine I would strongly strongly
1:03:39 strongly strongly strongly strongly
1:03:41 strongly dissuade you from taking that
1:03:44 approach
1:03:47 um
1:03:49 your student whichever tool you find
1:03:51 that can detect AI your students will
1:03:54 quickly figure out ways to
1:03:57 um
1:03:57 use tools to rewrite it so it's not
1:04:00 detectable by the tool that you selected
1:04:03 um this stuff
1:04:07 this stuff
1:04:09 is coming everywhere
1:04:11 it's going to be in the world that your
1:04:13 students enter
1:04:16 and so rather than try to do this and
1:04:19 say to hell with the new technology
1:04:22 accept the fact that this stuff's not
1:04:24 going to be uninvented so Google three
1:04:28 days ago announced the general
1:04:30 availability of duet which is their
1:04:32 version of chat GPT in all of the Google
1:04:35 tools it's going to be in Google Docs
1:04:37 Google Sheets Google presentations
1:04:40 the the help me write button it's gonna
1:04:42 literally be help me write it's going to
1:04:44 be in everything it's in slack right now
1:04:46 it's in notion it's going to be in
1:04:49 Microsoft 365. it's going to be in
1:04:52 Apple's operating system it's going to
1:04:54 be everywhere so right now these tools
1:04:56 kind of live off to the side but they're
1:04:58 very quickly being integrated into what
1:05:00 we do now if you work for a school
1:05:03 district that at the district level it
1:05:05 says we're blocking chat GPT I feel bad
1:05:09 for you
1:05:10 I don't think it serves the district I
1:05:12 don't think it serves the students I
1:05:15 don't think it serves the teachers you
1:05:17 can use these tools to read and analyze
1:05:21 papers and help create You Know lesson
1:05:24 plans faster and do a lot of the
1:05:26 administrative that work that you do
1:05:27 much much faster these tools are
1:05:30 remarkable take a look at
1:05:33 um Harvard this year just released their
1:05:38 generative AI guidelines Harvard you can
1:05:41 take that to the school board or to your
1:05:43 to your administrators if this is coming
1:05:45 from the administrators Harvard now has
1:05:48 generative AI guidelines let me go look
1:05:51 at
1:05:52 Harvard
1:05:55 gen AI guide lines
1:06:05 provost.harvard.com
1:06:09 guidelines for using chat GPT and other
1:06:12 AI tools at Harvard
1:06:15 protect confidential data you're
1:06:17 responsible for any content adhere to
1:06:19 current policies on academic integrity
1:06:23 be alert for AI enabled phishing
1:06:28 um
1:06:30 sister one
1:06:36 there's another one
1:06:41 um
1:06:42 [Music]
1:06:50 anyway yeah
1:06:52 uh AI policy gaps diverse policy stuff
1:06:55 go go look for the for the Harvard
1:06:58 policy audit also look for the city of
1:07:00 Boston's guidelines on using generative
1:07:02 AI
1:07:03 I would also look for a tick-tocker her
1:07:06 name is I think it's I am Gibson or
1:07:09 Gibson's in her name she's an English
1:07:10 teacher
1:07:12 she brought chat GPT into the classroom
1:07:14 she teaches English in 11th and 12th
1:07:17 grade I think and she's engaging her
1:07:21 students to like her her senior class
1:07:23 this year she's having them write
1:07:26 um a major thesis paper without chat EBT
1:07:30 and a major thesis paper with chat EPT
1:07:34 and the entire semester the entire year
1:07:36 they're going to be talking about what
1:07:39 are these tools good at what are they
1:07:40 not good at what did you have to rewrite
1:07:42 what was hard for you to write what was
1:07:43 easy for you to write where did it help
1:07:45 where did it hurt all that sort of stuff
1:07:46 she's engaging their critical thinking
1:07:49 skills that's what's going to be most
1:07:51 important here's the challenge our
1:07:53 current education system is training
1:07:57 students for a world that's about to
1:07:59 cease to exist
1:08:00 these tools
1:08:03 over the next three years are going to
1:08:05 start to do all of the Tactical
1:08:07 execution of like writing copy and
1:08:11 entering data and doing all this stuff
1:08:13 all the all the sort of rote stuff that
1:08:15 a lot of our education is training
1:08:17 people to do is going to be done by
1:08:19 these tools what's going to be required
1:08:21 is critical thinking so I would strongly
1:08:24 encourage you don't think like how do I
1:08:26 detect them and get them to stop using
1:08:28 it it's not going to stop and it doesn't
1:08:31 serve them so that's my soapbox I'll get
1:08:34 off it you can disagree with me but I
1:08:37 think this is really important it's
1:08:38 something I'm very passionate about
1:08:41 um and I I I'm not I'm not historically
1:08:44 passionate about education but even
1:08:48 though I started a channel called the AI
1:08:49 learning lab
1:08:50 um I just think this stuff's super
1:08:52 important and it's coming whether we
1:08:53 want it or not
1:08:55 and it's going to transform not only
1:08:56 education it's transforming all of the
1:08:58 creative arts it's transforming all of
1:09:01 knowledge work it's transforming
1:09:02 anything that's computer mediated is
1:09:05 going to be absolutely radically
1:09:06 transformed by this stuff
1:09:09 and the more that we can
1:09:11 get curious about it and embrace it and
1:09:15 learn what it's good at and what it's
1:09:16 not good at that's how we're going to
1:09:18 serve students of all ages so that's
1:09:22 where my Haiti is
1:09:29 okay my boyfriend's in jail and I need
1:09:32 an AI to basically pretend they are an
1:09:34 attorney
1:09:36 um
1:09:36 okay that actually sounds serious and
1:09:39 and
1:09:40 um so a couple of things I would say
1:09:42 go to prompts.chat
1:09:45 and look at how they've crafted the
1:09:49 personas that they create I'm pretty
1:09:50 sure in there is you know act like a
1:09:52 lawyer you can tell Chachi BT to act
1:09:54 like a lawyer act like this act like
1:09:56 that when you're dealing with legal
1:09:58 stuff
1:09:59 it's really good at
1:10:03 um analyzing documents and it's really
1:10:05 good at having opinions but my nickname
1:10:08 for chat GPT is mansplaining as a
1:10:11 service it will confidently answer you
1:10:13 even if it's wrong there's a pretty
1:10:15 famous case six weeks or so ago of a
1:10:19 lawyer that had chat GPT write a legal
1:10:22 opinion
1:10:23 complete with legal citations and handed
1:10:26 it to a judge
1:10:28 and the judge clicked on all those legal
1:10:30 citations and realized they were all
1:10:32 fake [ __ ]
1:10:34 because Chachi PT just made them up so
1:10:38 this stuff can actually probably give
1:10:40 you you know at least good understanding
1:10:43 of what your options are and maybe give
1:10:45 you ideas and maybe write nice letters
1:10:47 for you but double check everything
1:10:49 especially if you're using citations in
1:10:51 any way double check that stuff
1:10:54 um the other thing that I think is is
1:10:56 potentially worth your time to explore
1:11:00 is play with pi dot AI
1:11:04 or the PI app if you go to
1:11:07 um
1:11:08 uh the Apple Store
1:11:11 and look up Pi Pi personal assistant
1:11:14 you'll find that app the company's
1:11:15 inflection and put it into voice mode
1:11:18 and you can actually just start talking
1:11:21 with it and asking it questions here's
1:11:23 the situation here's the situation with
1:11:25 my boyfriend and why that's why that's
1:11:29 potentially interesting is Pi's actually
1:11:31 connected to the internet so if there is
1:11:34 modern case law that could help in the
1:11:38 argument Pi will be able to search for
1:11:40 that modern case law better than chat
1:11:43 GPT which isn't connected to the
1:11:45 internet
1:11:46 um you could also try Bard
1:11:49 um which is on my first little page
1:11:51 there Bard is connected to the internet
1:11:55 the other thing so so there's three
1:11:57 things I would try I would try gpt4 at
1:12:00 open AI
1:12:01 I would try Bard I would try pi
1:12:04 and then if you've got long documents
1:12:07 like if you've got long case law
1:12:09 documents or legal briefs or lawsuits or
1:12:12 whatever it might be
1:12:15 Claude can read very long documents like
1:12:18 up to about 75 000 words
1:12:21 so you can take very long documents
1:12:23 upload them to Claude or just paste them
1:12:26 into the prompt and then ask any
1:12:28 questions you want about it all of these
1:12:30 tools are really good at reading
1:12:32 documents and analyzing them Claude is
1:12:34 really good at long documents so hope
1:12:37 that helps
1:12:40 paint looking at the uh the advanced
1:12:42 data analytics the heat map right
1:12:44 writing its own code and generating heat
1:12:46 maps and charts and graphs yeah it's
1:12:48 [ __ ] remarkable
1:12:50 um read the probable cause affidavit
1:12:52 tell me what what motions to file yeah
1:12:55 okay
1:12:56 yeah so if I don't know if you're saying
1:12:59 read or read if it read it it's really
1:13:01 good at that I would say Claude is is
1:13:03 the one if you want to have it read long
1:13:04 documents and not have it choke claude's
1:13:07 the one to do that with
1:13:11 um and yeah Thelonious talking about do
1:13:13 not pay.com is an AI lawyer I depending
1:13:15 on what your your boyfriend is facing uh
1:13:19 do not pay could work
1:13:21 um but you might need something a little
1:13:23 more sophisticated wonder when we will
1:13:26 when we will find uh on the Bible codes
1:13:29 I mean listen there has there's there's
1:13:32 been
1:13:33 so much written about the Bible codes
1:13:35 these things know know all of you know
1:13:38 these these tools chat Chachi BT Claude
1:13:42 Bard all of these things have been
1:13:44 trained on
1:13:45 as much of the public internet that
1:13:47 people can get their hands on that that
1:13:49 these companies can get their hands on
1:13:51 so something like the Bible Code that
1:13:54 people have been talking about for what
1:13:56 centuries now long long time these
1:13:59 things know all that right so you can
1:14:01 absolutely just go ask it all that stuff
1:14:02 and just run down your own rabbit holes
1:14:04 I want you to act as a Biblical scholar
1:14:07 that's really good at code breaking and
1:14:09 blah blah blah blah and just start
1:14:11 asking you questions
1:14:13 um what's the secret human question to
1:14:15 know who is a real human profile wait
1:14:18 what is the secret human question to
1:14:20 know who is a real human profile I don't
1:14:23 know what the secret human question is
1:14:25 but I do have something that I can show
1:14:29 I'll do a little I'll do a little
1:14:31 showing off here that'll it's gonna piss
1:14:33 some people off but I think that's funny
1:14:36 um
1:14:41 [Music]
1:14:49 [Music]
1:14:53 I
1:14:54 am a verified human
1:14:57 and I'm anonymous
1:15:00 but I have a blockchain verified by my
1:15:03 retina
1:15:04 at a thing called the orb
1:15:06 World ID
1:15:08 so this is
1:15:10 this is an nft this is a blockchain
1:15:14 identity badge or identity token
1:15:19 um of me
1:15:22 that doesn't require my name my social
1:15:25 security number any of that [ __ ] it just
1:15:27 verifies that I'm human
1:15:30 um and this is Sam Altman's new company
1:15:32 called World coin and this is World ID
1:15:34 so his his new company has three
1:15:36 components World coin which is the
1:15:38 cryptocurrency world ID which is a human
1:15:41 verification system and then a whole
1:15:43 transaction like a PayPal kind of thing
1:15:45 and it's all in developer mode right now
1:15:47 so if you're a developer you can be
1:15:50 building applications on top of this
1:15:52 system where for example you could build
1:15:54 a Twitter
1:15:55 that
1:15:57 people could only use if they were
1:15:59 verified to be a unique human one of the
1:16:04 things this did when it scanned my
1:16:05 irises which I know is creepy as [ __ ]
1:16:07 but whatever I don't give a [ __ ] when it
1:16:09 scan my Iris what it said was it was
1:16:12 comparing my
1:16:15 um signature my biometric signature with
1:16:19 all of the other humans in the database
1:16:20 it was it was basically validating my
1:16:22 uniqueness compared to the other humans
1:16:25 in the database so once it did that it
1:16:28 gives me this blockchain verified thing
1:16:30 so yeah that's here
1:16:33 I went to New York City to get that
1:16:35 because you know like you do
1:16:38 [Laughter]
1:16:41 um
1:16:47 he's so weird he got his Iris scanned he
1:16:50 put it in a thing called the orb and
1:16:52 then and then he seems the normal sort
1:16:55 of but he's a little weird you know what
1:16:56 it seems like he got scanned by a robot
1:16:59 and that's what it seems like oh thank
1:17:02 you very much for their how how
1:17:05 actually one of the last plays I did I
1:17:07 played Elvis in uh
1:17:09 Steve Martin's Picasso at the LA
1:17:11 panajiel
1:17:13 in Ridgefield Connecticut
1:17:17 that microphone reminded me when I sang
1:17:19 I sang me some Elvis songs
1:17:23 [Laughter]
1:17:25 all right
1:17:27 let's see
1:17:29 Ahoy sir how is the picture-in-picture
1:17:32 featured enabled this
1:17:34 is when you go live
1:17:37 um
1:17:39 there's an option called dual camera
1:17:41 what's weird about it is Tick Tock took
1:17:44 it away from me for like a month
1:17:46 I don't know why it just was gone like
1:17:48 it was there and people were like how do
1:17:50 you do that
1:17:51 and all it is is that box is the back
1:17:54 camera and then this is the front camera
1:17:56 and that's how I'm doing it so that's
1:17:58 what did it dual camera feature when you
1:18:01 go live
1:18:03 code quality seems to change quite a bit
1:18:06 depending on the language and model yeah
1:18:08 I know the code quality right now is a
1:18:10 mess they call that
1:18:12 um in the in the large language models
1:18:14 drift so as
1:18:17 open AI is fixing
1:18:21 their their fine tuning of chat GPT some
1:18:24 things get better some things get worse
1:18:26 which is again you know the I kind of
1:18:29 make the joke about these tools being
1:18:30 janky pieces of [ __ ] but it's important
1:18:33 to know that as you get more educated
1:18:35 with these tools and you start playing
1:18:37 with them you start understanding what
1:18:38 they are
1:18:40 um
1:18:42 you're going to start to want to build
1:18:43 stuff and you might use a tool like
1:18:45 zapier and you create these automations
1:18:47 where someone can fill out a form and
1:18:48 then you send stuff to open Ai and it
1:18:50 makes these really cool things and
1:18:51 you're like oh just fantastic and I made
1:18:53 this application and it's perfect
1:18:56 and then three months later you go to
1:18:57 show it to someone and it's not perfect
1:18:59 anymore
1:19:00 because there's drift so so
1:19:04 a strong recommendation right now is
1:19:06 learn as much as you can build as much
1:19:08 as you can talk as much as you can to
1:19:12 people about what you're learning
1:19:14 and
1:19:15 accept the fact that whatever you build
1:19:17 right now whatever you learn right now
1:19:19 is going to be completely irrelevant
1:19:20 three months from now just just know
1:19:23 that just like life
1:19:25 everything is some temporary game that
1:19:27 we play it's just that with these tools
1:19:30 it's much more temporary because they're
1:19:32 because they're changing so fast this
1:19:34 stuff's changing so fast like the the
1:19:36 the requirement right now that is on all
1:19:40 of us to be good at prompting
1:19:42 and to be good at you know using our
1:19:46 native language as a programming tool
1:19:49 that requirement is going to go away if
1:19:52 you're a developer you're still going to
1:19:54 need to be able to do that but most of
1:19:56 most human beings are not going to need
1:19:58 that you're going to be able to just ask
1:20:00 this [ __ ] kind of like when I was
1:20:01 showing Advanced data analytics where I
1:20:03 said hey what's inside this and what's
1:20:05 interesting about it and it just started
1:20:07 doing all its own [ __ ] that's how these
1:20:10 tools are gonna act in the future
1:20:12 right now they don't act like that way
1:20:14 because they're janky piece is a [ __ ]
1:20:16 but as they get more sophisticated
1:20:18 there's going to be increasing layers of
1:20:20 interface that hide all the sort of
1:20:23 technical underpinnings from US unless
1:20:25 we want to dig in there and play with
1:20:27 those most people don't want to do that
1:20:29 most people are like I want the robot to
1:20:32 do the hard stuff for me could the robot
1:20:34 do the hard stuff if the robot could do
1:20:36 the hard stuff I'd like this it would be
1:20:38 pretty good because there's a lot of
1:20:39 [ __ ] I do I don't like you know what I
1:20:42 mean and so I I would like the robots to
1:20:44 do the hard stuff
1:20:47 and if the robots do the hard stuff
1:20:49 everybody wins seriously
1:20:52 like I think this is my this is my thing
1:20:55 like the more I'm learning
1:20:57 I think the killer app is
1:20:59 our world is too complicated and most of
1:21:02 us are miserable because every [ __ ]
1:21:04 day we spend two hours on the phone with
1:21:07 a t trying to get them to reinstate our
1:21:10 AARP discount for ten dollars a month
1:21:12 for the stupid [ __ ] phone bill
1:21:15 that was my
1:21:18 Labor Day Monday two hours
1:21:21 just trying to get my stupid 10 bucks a
1:21:25 month discount back
1:21:27 they were clueless on the phone we had a
1:21:30 bad connection they couldn't hear me I
1:21:32 couldn't hear them we were both
1:21:33 screaming at each other
1:21:36 then I go to log into the website and it
1:21:38 breaks it doesn't log in it's just like
1:21:40 something went wrong sorry
1:21:42 two hours it just
1:21:45 wouldn't it be nice if you just said
1:21:47 excuse me could you go make sure that I
1:21:50 have all the discounts I need on My ATT
1:21:51 thing and it's like sure boss like
1:21:54 Golden Retriever
1:21:56 I'll do that for you let me go now
1:21:59 it runs off and your little robot talks
1:22:02 to their little robot they work some
1:22:04 [ __ ] out your little robot comes back
1:22:06 and says okay boss here you go I fixed
1:22:08 this I got this you have two dollars
1:22:09 Less on this you have a choice do you
1:22:11 want this or do you want this uh I want
1:22:13 that okay great got that I'll go change
1:22:15 that now you have another choice you
1:22:16 want this you want that do you want that
1:22:18 do you want a new Apple watch you can
1:22:19 get one for free
1:22:21 is it gonna cost me anything nope okay
1:22:23 click
1:22:24 I'll be back all right your watch will
1:22:27 be there tomorrow
1:22:31 why can't it do that well it can it's
1:22:33 going to
1:22:35 that's the killer app
1:22:38 all right
1:22:40 now screwfuffled I was totally
1:22:42 kerfuffled there was just too many hats
1:22:44 it was too much pressure
1:22:45 I didn't have a script I didn't learn my
1:22:47 lines I was a bad actor
1:22:48 [Laughter]
1:22:53 oh what's going on wait sorry
1:22:56 tennis moment just getting a score
1:22:57 update
1:23:01 looks like tiafo's up a break in the
1:23:03 third set
1:23:07 yeah
1:23:09 is it really an ambush if you were
1:23:11 forewarned that's that is now you're
1:23:14 getting just smart alecky Emilio's wife
1:23:17 all right so guys are we in a regular
1:23:19 private room we're in a general one took
1:23:21 us in the AI Discord in the AI Salon
1:23:24 Discord there's a private Irregulars
1:23:28 Channel
1:23:29 but you just have to let me know who you
1:23:31 are in Discord and then I'll give you
1:23:32 the role
1:23:33 so for those of you that don't know what
1:23:35 we're talking about
1:23:40 I've created a community called the AI
1:23:43 Salon that we meet every other week
1:23:46 and
1:23:47 um we have a Discord server and we can
1:23:50 continue the conversation about all this
1:23:52 AI stuff there so that link the salon.ai
1:23:57 is a link tree that contains information
1:23:59 about the salon itself and our values
1:24:02 and a link to the Meetup where you can
1:24:05 sign up to you know join us every other
1:24:08 week and then a link to the Discord
1:24:10 where you can go
1:24:12 play with mid journey and talk about
1:24:15 [ __ ] and ask questions it's cool it's a
1:24:17 neat it's a neat Community Silver Fox
1:24:19 thank you what's that cool thing oh it's
1:24:20 a little Tweety Bird
1:24:22 there's a little Tweety burp it's not
1:24:23 called tweets anymore it's called X it's
1:24:26 a little expert nice little expert you
1:24:28 got there except he killed the bird
1:24:31 he killed the bird
1:24:33 all right
1:24:35 the only thing that isn't irregular
1:24:36 about us is our tapping 36
1:24:39 000 likes yellow nuts thank you Stevo
1:24:42 thank you silver fox
1:24:44 uh good evening Kyle Bruno what's
1:24:46 happening what's shaking what's shaking
1:24:48 shaking shaking wait Kyle do we have
1:24:51 your full attention
1:24:53 no one ever has my full attention it's
1:24:55 always like um
1:24:59 it's like add is just like some sort of
1:25:02 space antenna just [ __ ]
1:25:05 filtering in thoughts from everywhere
1:25:10 uh
1:25:14 uh Ben Shelton's about to break tiafo
1:25:16 back in this game
1:25:20 Ben Shelton he's gonna be superstar mark
1:25:24 my words that boy's gonna be good oh
1:25:26 wait sorry
1:25:28 the cow where's the pictures we want
1:25:30 pretty pictures
1:25:32 how will we know for an AI learning lab
1:25:34 if we don't have the pretty pictures
1:25:35 Kyle
1:25:36 what's up with that cow
1:25:40 the elephant and chat stopper LOL five
1:25:43 items added to the list tonight oh good
1:25:46 uh felonious thank you for doing that so
1:25:49 what we should do Thelonious if you
1:25:50 haven't joined the Discord you should
1:25:51 join the Discord and we can share the
1:25:54 um felonious is making like a Google
1:25:55 sheet oh you know what we could do
1:25:57 Thelonious if you want
1:25:59 I can put the Google sheet on one of the
1:26:02 link trees or other
1:26:04 so what Thelonious is doing is every
1:26:06 time I show some stupid ass tool where
1:26:08 I'm like oh this is a cool one
1:26:10 in in my sort of manic add way
1:26:14 um he's been adding it to a Google sheet
1:26:16 so he's got like a list of like 75 tools
1:26:18 which that's super cool so first thank
1:26:20 you
1:26:21 and second we should figure out a way
1:26:22 that we can share that because if it's a
1:26:24 Google sheet you know we can just make
1:26:25 it shareable and then we can figure out
1:26:27 a place to to share it I think that
1:26:29 would be super cool
1:26:31 um and that [ __ ] is going to change all
1:26:34 the time
1:26:35 and and you know what I'm trying to do
1:26:37 on the channel is
1:26:41 through like whatever my filter is it's
1:26:43 like like I have a decent [ __ ] meter
1:26:45 and so what I try to do is like you know
1:26:48 what what makes my list here is like
1:26:50 this is [ __ ]
1:26:52 this is [ __ ] that I've found value in
1:26:54 that I find Value in pretty regularly
1:26:56 and I'll knock [ __ ] off this list as as
1:26:59 new stuff comes up that's better than
1:27:00 this I'll knock stuff off this list
1:27:04 um this list for most of the year was
1:27:06 three things long now it's seven and now
1:27:08 we've got four different image tools
1:27:12 and the image tools quite frankly this
1:27:15 should probably be more like six or
1:27:17 seven but I just it's just too much so
1:27:20 and then the video tools are a little
1:27:22 they're expensive and a little janky
1:27:25 right now and the audio synthesis tools
1:27:28 are okay but the audio synthesis doesn't
1:27:30 really sound oh Shelton broken so
1:27:33 they're they're on serve in the third
1:27:35 set tiafu looks broken Ben Shelton looks
1:27:39 like
1:27:41 20 year old confidence man
1:27:45 dude's a beast
1:27:47 149 mile an hour serve come on
1:27:53 all right how do I get flow wise to
1:27:55 stream the responses from bot press
1:27:57 janky is a great way to describe it all
1:28:00 uh Alexander rude I don't know you're
1:28:04 you're down in a level of tools a level
1:28:06 below what I'm dealing with so flow wise
1:28:09 to stream the responses from bot press I
1:28:12 have no idea
1:28:13 I have no idea I can I can do
1:28:15 automations in zapier
1:28:18 I know a handful of the tools but I'm
1:28:21 not like I'm most of what my time is
1:28:24 doing is is I'm the CEO of a video
1:28:26 technology company so so I'm not doing
1:28:29 any of this AI [ __ ] full time other than
1:28:31 how I'm applying it to my business this
1:28:33 is all
1:28:34 this isn't a hobby for me like this is a
1:28:37 what do they call it an avocation like
1:28:39 this is something
1:28:41 the stuff the reason I'm doing this is
1:28:43 all of this stuff feels very very
1:28:44 important to me
1:28:46 um I like I feel like in there's there's
1:28:48 been a handful of times in my life where
1:28:51 I can't not be doing this like I have to
1:28:55 be
1:28:56 actively engaged with this [ __ ] every
1:28:59 single day because it's critically
1:29:03 important critically [ __ ] important
1:29:05 that's why this channel exists is you
1:29:08 know if I can get one person to go to
1:29:11 chat.openai.com
1:29:13 type in some [ __ ] little thing where
1:29:15 they're like oh this is stupid it's just
1:29:17 like Google do it
1:29:19 and then they have their Kevin
1:29:20 McAllister moment there
1:29:23 their home alone moment
1:29:26 and then they realize oh [ __ ] these are
1:29:28 different something's different oh crap
1:29:31 what do I do thank you togas oh togas
1:29:33 reached level 13. banger is leveling up
1:29:37 people people are leveling up in here
1:29:40 I don't know if you you're competitive
1:29:43 right why would you not be competitive
1:29:45 oh right because everyone gets a
1:29:47 participation certificate no not in Gen
1:29:49 X we didn't we got trophies you know who
1:29:51 got Trophies the winners
1:29:55 you know what we called the people that
1:29:56 didn't get trophy loser
1:30:01 that's how we played
1:30:04 we got slapped around
1:30:14 I think we need to go back to the little
1:30:16 winners and losers
1:30:18 all right
1:30:21 totally surprised turning soprano the
1:30:24 books are closed my husband asked me why
1:30:26 do I watch this stuff you'll never use
1:30:28 it I beg to differ okay Don A Smith
1:30:32 kudos to you 1956 I'm 1965. listen
1:30:38 he's wrong
1:30:39 he's wrong
1:30:41 you will use this here's the deal about
1:30:43 AI you're already using AI it's in
1:30:45 almost everything we've been using AI
1:30:48 isn't new it's been around for decades
1:30:50 ask Pate who's in here tonight like he's
1:30:52 he's working on the sort of core
1:30:54 underlying models and making that [ __ ]
1:30:57 more efficient like the reason this [ __ ]
1:30:58 just works is because of people like
1:31:00 Pate that have been doing this for
1:31:01 [ __ ] decades
1:31:03 what's new
1:31:05 is the chat GPT piece the piece that
1:31:08 makes these tools available to like
1:31:10 regular human beings
1:31:13 these things are janky pieces of [ __ ]
1:31:15 right now but they're not going to be
1:31:16 for long like like in the next 18 months
1:31:21 what these things do is going to get way
1:31:23 more sophisticated and where these
1:31:24 things live is going to be everywhere
1:31:27 it's going to be features inside of
1:31:29 everything we touch
1:31:32 and so
1:31:34 what you're doing Don is
1:31:37 by being curious about this and
1:31:39 understanding what these things are and
1:31:41 playing with them even if you're just
1:31:43 saying help me do my garden help me
1:31:45 figure out some clever stitch for my
1:31:48 sewing Club help me organize the potluck
1:31:52 dinner for the school it doesn't matter
1:31:55 whatever you're doing with this stuff
1:31:56 anything that you're learning right now
1:32:00 is going to be a resource to you as this
1:32:02 stuff gets more and more sophisticated
1:32:04 when it shows up in all of the different
1:32:07 tools that we use you're going to know
1:32:09 what to do with it you're going to know
1:32:10 how to do it you're going to know why
1:32:11 it's cool you're going to know why it's
1:32:12 powerful so kudos to you he's wrong
1:32:17 rock on sister
1:32:19 T minus eight minutes who's in y'all are
1:32:22 great that was very sweet Sheriff the
1:32:24 the uh the Hat the Hat barrage was
1:32:27 awesome
1:32:29 can AI with machine learning develop
1:32:32 self-awareness and become sentient it's
1:32:35 an interesting question Zen in it
1:32:38 yes okay it's 9 33 p.m apparently
1:32:42 serious just so jealous of me
1:32:46 um
1:32:50 Here's my thought on sentience and AGI
1:32:53 and all that sort of stuff
1:32:55 um
1:32:56 I think at this point
1:32:58 it's
1:32:59 it's a matter of semantics and and I I
1:33:03 think ultimately it it will be way more
1:33:06 than just a matter of semantics but
1:33:10 if you understand how chat GPT works if
1:33:13 you understand how
1:33:15 the probability engine
1:33:18 you know that that is the way that these
1:33:21 tools generate what they generate
1:33:26 it is mathematically not sentient it is
1:33:29 mathematically a calculator so when you
1:33:32 type your prompt
1:33:33 what the machine's doing is turning your
1:33:35 prompt into a series of numbers and in a
1:33:39 weighted probability
1:33:41 that goes and looks in this crazy ass
1:33:43 mathematical Vector space and says some
1:33:47 point in this thousand dimensional
1:33:49 mathematical space has the highest
1:33:52 probability of being the next word or
1:33:54 the next phrase based on the prompt that
1:33:57 you wrote
1:33:59 and then it just does that over and over
1:34:01 again so it literally doesn't know what
1:34:05 words it's putting on screen it's
1:34:07 basically finding a number in Vector
1:34:09 space and then just associating that
1:34:11 number with the word that's tied to it
1:34:13 and then that's the word that it puts on
1:34:15 screen where that's the phrase that it
1:34:17 puts on screen and it does that over and
1:34:19 over again
1:34:20 it is not aware of what it wrote
1:34:22 to it
1:34:24 it's just a series of numbers in a
1:34:27 sequence based on probabilities it's
1:34:28 probability engine it's conceptually
1:34:30 super simple
1:34:33 it's [ __ ] magic right it's so
1:34:36 sophisticated that what it produces
1:34:40 you look at it and you're like this
1:34:42 thing has to be sentient it has to
1:34:43 understand that but but what's that
1:34:46 what's actually happening here is it's
1:34:48 been trained on all of what we have put
1:34:51 on the internet for the past four
1:34:53 decades or five decades not all of it
1:34:55 but as much as they could get their
1:34:56 hands on
1:34:58 and so this is what what these things
1:35:01 are right now is they're like a mirror
1:35:04 of our Humanity we've put a bunch of
1:35:06 language into the world these things
1:35:08 have ingested that knowledge
1:35:11 and created all these semantic pools
1:35:14 mathematical clusters
1:35:17 of what all the words mean and then when
1:35:20 we interact with it it kind of responds
1:35:22 back to us
1:35:23 kind of like a human because we trained
1:35:26 it on what we've done so this is a
1:35:28 mirror of us
1:35:31 so right now
1:35:33 they're not aware they're not aware of
1:35:36 what they say of of what they do now
1:35:38 could they be sure because
1:35:41 so one of the reasons these things
1:35:43 hallucinate is you ask it a question and
1:35:46 it's just pulling from all of the [ __ ]
1:35:47 that was on the internet so it's pulling
1:35:49 stuff that was wrong as much as it was
1:35:51 right and it's it's some amalgamation of
1:35:54 what you asked for is is what gets put
1:35:56 back and sometimes it lies
1:35:58 there's not a layer right now there's
1:36:00 there's not a
1:36:01 there's not a cycle where these large
1:36:04 language models actually analyze what
1:36:06 they gave you and score it
1:36:10 so they don't say is this correct did I
1:36:13 lie about these citations there's none
1:36:14 of that right now
1:36:16 as they start adding in things like
1:36:18 memory so like
1:36:20 imagine if chat gbt remembered every
1:36:22 conversation you've ever had with it and
1:36:24 you could say hey chat GPT remember when
1:36:26 we talked about that obscure card game
1:36:29 that had weird rules oh yeah that was
1:36:31 this game on this date here's what we
1:36:32 talked about what do you want to know
1:36:34 about it
1:36:35 then it starts to get a little
1:36:36 interesting
1:36:37 when it starts actually being aware of
1:36:40 what it wrote and analyzing what it
1:36:43 wrote and
1:36:44 adjusting what it wrote based on its
1:36:47 knowledge of what you like and what it
1:36:49 likes and like you could see this
1:36:52 getting
1:36:53 way more sophisticated in the future so
1:36:55 right now it looks sophisticated but
1:36:58 it's mathematically a calculator
1:37:00 in the future there's still going to be
1:37:02 a lot of math involved but it's going to
1:37:04 start to have awareness and memory and
1:37:07 reasoning and things like that that's
1:37:08 where you start to get into things like
1:37:09 AGI here's why I don't think it's super
1:37:12 relevant
1:37:14 from where I sit
1:37:16 just as a business guy an entrepreneur
1:37:19 that just wants to use these remarkable
1:37:20 tools these things are already
1:37:24 dramatically
1:37:26 smarter than I am more capable than I am
1:37:29 in in many many areas and like to some
1:37:33 areas I have no interest in learning it
1:37:35 other areas it's like I'd like to learn
1:37:37 that but I don't have to now it's all
1:37:39 there
1:37:42 so
1:37:44 you know I suppose at the point that
1:37:45 they become they become sentient then we
1:37:47 have conversations like well
1:37:50 is it ethical to shut one off if it's
1:37:52 aware of its own existence I don't know
1:37:56 it's still a machine that that we're
1:37:58 prompting but there may be a point at
1:38:01 which
1:38:02 it is so sophisticated that we're like
1:38:05 it's kind of human I don't know I don't
1:38:07 know
1:38:08 but but right now these things are
1:38:10 already significantly more profound than
1:38:13 anything I've ever touched in my life
1:38:14 it's completely making me rethink
1:38:17 how I engage with computers and how we
1:38:20 will engage with computers like
1:38:21 completely
1:38:23 like [ __ ] 58 years of how I do
1:38:26 problem solving and creative thinking is
1:38:30 completely different
1:38:32 in nine months and and just I'm telling
1:38:35 you that the minute the minute you cross
1:38:37 the threshold from oh this is a cute
1:38:40 little tool to
1:38:42 holy [ __ ] it can do that the minute you
1:38:45 cross that threshold this is red pill
1:38:47 blue pill you can unlearn it you can't
1:38:49 unsee it and you can't describe it to
1:38:52 anyone why people who are into AI sound
1:38:56 so [ __ ] insane
1:38:58 is that they're trying to explain
1:38:59 something that's unexplainable
1:39:03 unless you are in it and see it and
1:39:05 experience it and then you go oh I get
1:39:07 it now
1:39:08 that's what this channel is is like
1:39:11 people come in here and like who's this
1:39:13 [ __ ] weirdo oh okay I'll go to Tia
1:39:16 gbt whatever you know and judgments is
1:39:20 whatever I'll call it oh my god oh that
1:39:24 was rather impressive well but it can't
1:39:26 do what I do oh
1:39:30 oh
1:39:32 well let me see if I can really do what
1:39:33 I do
1:39:34 oh my oh my should I be oh my what do I
1:39:38 do now oh
1:39:43 until you have that moment
1:39:46 how do you explain that to someone I had
1:39:48 I had an epiphany with a computer
1:39:51 I had a breakthrough I realized that
1:39:53 what I do is now it's not going to be
1:39:57 done that way anymore well how do you
1:39:58 know that I just know because it doesn't
1:40:00 like I don't know it like I don't know
1:40:02 what to do with this information
1:40:04 that's where we are right now
1:40:06 nobody knows what to do with this
1:40:08 information I don't know what to do with
1:40:10 this information other than [ __ ] talk
1:40:12 about it incessantly two or three hours
1:40:14 a night
1:40:16 and and learn every [ __ ] thing I can
1:40:20 so
1:40:21 you know Don with her husband going yeah
1:40:24 you're never gonna use this [ __ ] Don
1:40:26 Don's like uh yeah well okay
1:40:29 keep watching the tennis match there
1:40:31 hubby
1:40:37 six six in the third they're in a tie
1:40:39 breaker in the third set this is a good
1:40:41 man it's very exciting
1:40:43 I'll ask me some U.S Open tennis
1:40:46 all right
1:40:47 thanks for the update Kyle about data
1:40:49 privacy obviously early days in the uh
1:40:52 in the matter for small sized businesses
1:40:54 yes
1:40:55 uh then maybe okay
1:40:59 oh Tick Tock just jumped me to the
1:41:02 bottom oh Tick Tock For the Love of
1:41:05 For the Love of All That is programming
1:41:09 fix the [ __ ] bug here's the problem
1:41:12 with the bug that drives me crazy is
1:41:14 I'm I'm sure I'm the only one that
1:41:16 complains about it I don't think anyone
1:41:18 else runs into this bug
1:41:21 so they never gonna fix it never never
1:41:23 well that sounds very cynical and
1:41:26 negative Kyle have you ever considered
1:41:27 trying a positive outlook on life oh
1:41:29 I've tried it
1:41:33 but I've been on the planet
1:41:35 [Applause]
1:41:37 show and tell time Mr Shannon all right
1:41:39 well what do you want me to show and
1:41:41 tell I will show and tell whatever you
1:41:43 want to say you just made my night thank
1:41:45 you very much ask pie if she thinks your
1:41:48 bow would help you if you were in jail
1:41:51 cabrino
1:41:54 that's that's a solid that's a solid
1:41:56 bitter response I like it I fully
1:41:59 support the bitterness
1:42:00 bitter you betcha
1:42:02 that's chat gbt I'll leave your
1:42:04 boyfriend who's in jail because you
1:42:06 deserve better it's getting Saucy in
1:42:08 here I like it all right you just made
1:42:12 my night
1:42:13 um what AI services are worth paying for
1:42:16 okay
1:42:17 um
1:42:18 as as few as possible
1:42:21 for one thing because there's a lot of
1:42:23 free [ __ ] out there
1:42:25 chat gbt I paid 20 bucks a month for it
1:42:27 I think it's worth it GPT 4's best code
1:42:29 interpreter the advanced data analytics
1:42:32 is [ __ ] remarkable uh plugins I could
1:42:34 take or leave I don't really give a [ __ ]
1:42:36 about those claude's free right now but
1:42:39 they rate limited so at some point that
1:42:40 might be worth paying for but given the
1:42:44 choice between paying for that one or
1:42:46 Claude if they charge for it it's going
1:42:47 to be chat GPT for now
1:42:49 Bard is free
1:42:53 um
1:42:54 Bing is free
1:42:56 um lab stop perplexity is free the Llama
1:42:58 models are open source you can build
1:43:00 [ __ ] with them and you don't have to pay
1:43:01 for them
1:43:03 um Pi is free po is free but you can you
1:43:06 can pay for access to some of the more
1:43:07 advanced other models
1:43:10 um I wouldn't pay for Poe
1:43:12 all the all the rest of the shit's free
1:43:14 I pay for mid-journey absolutely worth
1:43:17 it it's the best image generator out
1:43:19 there
1:43:20 ideogram is free
1:43:24 Leonardo is is worth paying for in my
1:43:26 opinion if you want to make a lot of
1:43:28 images and you don't want to make images
1:43:30 inside of Discord
1:43:34 Shelton's up 4-2 in the tie breaker
1:43:39 um clip drop all those things are free
1:43:41 but I I think I think there's also some
1:43:44 sort of paid thing there so I pay for
1:43:46 mid-journey I pay for
1:43:49 Chachi PT
1:43:53 I pay for
1:43:58 um 11 Labs because I wanted to
1:44:00 synthesize my own voice
1:44:03 I pay for D Dash ID which makes the
1:44:07 animations of the voice so you can
1:44:10 you can synthesize your voice then write
1:44:12 a script and it'll talk like you sort of
1:44:14 and then you can put it behind a picture
1:44:16 and it'll talk like you with did so I
1:44:18 pay for that I pay for zapier
1:44:21 which is an automation tool
1:44:23 and I pay for Google collab so I can run
1:44:26 virtual machines if I want to geek out
1:44:28 and that's I think it right now
1:44:33 um and they're listen there are probably
1:44:35 others worth paying for as I said to the
1:44:37 Thelonious reminded me
1:44:40 um
1:44:44 I wouldn't pay for anything on an annual
1:44:46 basis just go month to month because the
1:44:48 shit's changing too fast
1:44:51 um what should I be showing telling let
1:44:53 me let me jump to the bottom and see
1:44:54 what what everyone's saying I should
1:44:56 show and tell
1:44:57 holy [ __ ] I was really far behind
1:45:00 um
1:45:02 felonious I'm an addict
1:45:06 hi my name is Kyle I'm into AI welcome
1:45:09 Kyle I've been into AI since I
1:45:12 discovered chat EBT on November 20
1:45:14 November 30th 2022. yeah we know Kyle us
1:45:17 too yep
1:45:19 yep yep every day I think about Ai and I
1:45:23 can't do anything else I don't sleep
1:45:25 anymore
1:45:26 yeah welcome to the club Kyle all right
1:45:29 everyone else excited oh yeah we're
1:45:31 thrilled why are we so low energy we're
1:45:34 [ __ ] exhausted oh thanks for the room
1:45:39 thanks for the penis nose
1:45:42 fantastic
1:45:46 oh and Shelton just whiffed one
1:45:49 five four Shelton Over tiafo
1:45:52 in the uh tie breaker oh it's a quick
1:45:57 man what do you want me to me to tell
1:45:59 there people
1:46:03 um I do 13 hours by myself what's the
1:46:05 best place to learn prompting uh
1:46:09 Ann Rando okay 10 out of 10 on the name
1:46:13 I am Rando
1:46:16 that's brilliant okay so prompting
1:46:21 if you're just getting started go to
1:46:23 that document prompts.chat
1:46:27 it's got primers in there
1:46:30 um that that'll teach you about the
1:46:34 um
1:46:35 large language models how they work how
1:46:38 to talk to them it's also a big long
1:46:40 list of like personas
1:46:42 for how
1:46:44 um
1:46:46 how you can make Chachi BT act like
1:46:49 different kind of experts basically
1:46:51 it's a good starting point it goes way
1:46:54 way deeper than this I would say just go
1:46:56 go to Reddit and go to Twitter you know
1:47:00 X
1:47:00 and search for chat GPT prompt cheat
1:47:04 sheets there's like a billion cheat
1:47:08 sheets out there some of them are good
1:47:10 some of them are not good the the best
1:47:12 way to get good at prompting is just
1:47:15 start playing and and here's the biggest
1:47:17 shift
1:47:19 mental shift that I can give you five
1:47:22 six in the tie breaker it's intense
1:47:25 um
1:47:28 Google has trained us that what you
1:47:30 should type into a prompt box is the
1:47:33 minimum amount possible right just give
1:47:35 me what you need and I'll go find you
1:47:37 your [ __ ]
1:47:38 these things need much more context the
1:47:41 the more context you give them the
1:47:43 better they get so think of a think of
1:47:46 an interaction with chat GPT not like
1:47:48 you're prompting and getting an answer
1:47:49 but more like you're starting a
1:47:51 conversation here's the kind of thing
1:47:53 I'm trying to solve here's what you're
1:47:56 good at here's what I'm good at here's
1:47:58 the piece I don't know here's my problem
1:48:00 six six in the tiebreaker holy crap
1:48:04 um
1:48:05 so
1:48:06 um they do have a bug on other lives but
1:48:08 maybe a smaller reaction I don't know
1:48:11 what that's about what was the free
1:48:13 image prompt site mentioned earlier the
1:48:16 free image prompt site that was ideogram
1:48:20 so ideogram dot AI is free and it's the
1:48:23 one that does
1:48:26 uh
1:48:27 text
1:48:29 these are all ideogram images
1:48:32 that have the AI learning lab words like
1:48:35 the word on The Silo Isn't that cool
1:48:38 I mean come on how cool is that I didn't
1:48:40 Photoshop that that ain't that ain't no
1:48:43 work I just said AI learning lab silo
1:48:47 and it did that these were all posters
1:48:50 so ideogram so that's free for now it
1:48:53 won't be for long but it's free for now
1:48:55 oh tiafo's up seven six
1:48:58 young young Shelton
1:49:01 young Ben Shelton he's he's up against
1:49:04 it
1:49:06 this is this is set point for tiafo and
1:49:08 he's serving hang on
1:49:11 oh Ben just nodded to his his uh box
1:49:15 it's gonna be okay
1:49:21 let's see if it's okay Ben
1:49:24 tiafo bang oh whiffed it did he win oh
1:49:28 no it was
1:49:29 long
1:49:31 second serve
1:49:35 was that in
1:49:37 holy crap what a return what a return
1:49:40 right oh my God he [ __ ] smashed the
1:49:44 second serve 7-7 in the tie breaker holy
1:49:48 crap
1:49:49 all right
1:49:53 I just uploaded two coloring books to
1:49:56 KDP very cool Sherry Banks
1:49:58 22 years versus two weeks
1:50:02 what's that the 22 years versus two
1:50:04 weeks thing because I just wrote the uh
1:50:06 I just wrote the uh the uh I just did a
1:50:10 video on the 10 000 hour rule that I
1:50:12 think the 10 000 hour rule is dead
1:50:14 right so instead of you having to spend
1:50:16 four years getting good at something you
1:50:18 can spend like you know four months now
1:50:20 or two weeks
1:50:23 right big rally big rally seven seven oh
1:50:27 tiafo wants it into the net
1:50:29 eight seven for Ben Shelton serving for
1:50:32 set number three
1:50:35 all right
1:50:36 let's see
1:50:40 what time is it what time is it it's 10
1:50:43 minutes to ten good night Kyler regulars
1:50:45 and and aliatory promptlines goodnight
1:50:49 took us
1:50:50 do we need more data
1:50:52 um
1:50:53 they're saying that we need more data
1:50:55 and then there's there's kind of weird
1:50:57 conflicting things about is synthesized
1:51:00 data going to cause model collapse or
1:51:01 not it seems like synthesized data when
1:51:04 used right can be quite powerful
1:51:07 and it also looks like oh oh wait wait
1:51:10 for it
1:51:11 Ben Shelton
1:51:14 set number set number three two sets to
1:51:17 one Ben Shelton
1:51:19 Francis tiafo's number 10 in the in the
1:51:22 world
1:51:23 people are losing their minds this kid's
1:51:25 a [ __ ] star
1:51:27 this kid's a [ __ ] star
1:51:29 amazing
1:51:31 um do we need more data
1:51:32 um I just saw a thing that um
1:51:35 so Chachi PT the way chat GPT is trained
1:51:39 is through what's called human
1:51:41 reinforcement learning so they have
1:51:43 Chachi PT or gpt4 generator response and
1:51:47 then humans go thumbs up thumbs down
1:51:48 here here's what's good about it or
1:51:50 here's how I'd rewrite it right
1:51:52 they're now getting
1:51:55 AI large language models to do that
1:51:58 instead of humans so the AI robots are
1:52:02 doing the human analysis as good or
1:52:05 better than humans are way faster
1:52:09 so this is now that thing that people
1:52:12 talked about well what happens when the
1:52:14 robots start training the robots and the
1:52:16 and they just get sort of
1:52:18 exponentially smarter faster
1:52:22 so so I don't know like I don't know
1:52:24 where the boundary is of do we need more
1:52:27 human generated data or do we solve the
1:52:30 problems of model collapse with
1:52:32 synthesized data and figure out ways to
1:52:34 use synthesized data
1:52:37 so that these things just get better and
1:52:38 better without collapsing I don't know I
1:52:41 do not know it's wild
1:52:45 um
1:52:49 oh which service did he say he paid for
1:52:52 for VMS for voice for voice stuff 11
1:52:55 Labs so eleven Labs Dot here you want to
1:52:59 say uh uh Emilio's wife wanted Show and
1:53:02 Tell let's go to a show and tell
1:53:06 11 labs elevenlabs.com
1:53:12 generative voice AI
1:53:16 and then let's go
1:53:19 we'll go to chat GPT
1:53:22 and we're going to go to gpt4 because
1:53:24 it's the best just normal
1:53:27 and we're going to say
1:53:29 um
1:53:38 write me a soliloquy of a sad zombie
1:53:42 turtle
1:53:45 that is lamenting the fact that the
1:53:50 Waterfront Carnival
1:53:53 is ending tonight
1:54:00 in the dimming light where water meets
1:54:03 the shore the Carnival's vibrant Colors
1:54:05 cast a glow transient and fading how
1:54:08 cruel time can be that's good we'll grab
1:54:11 that okay so
1:54:14 when I say play with these tools
1:54:16 seriously like just put in stupid [ __ ]
1:54:19 just ask it to do stupid [ __ ] and then
1:54:22 it what what invariably happens is you
1:54:26 realize
1:54:29 that these things are just way deeper
1:54:31 than you than than you think they could
1:54:33 possibly be
1:54:34 and it'll just write something and it a
1:54:37 it gives you Joy because it's just like
1:54:38 this spontaneous serendipitous stupid
1:54:41 little piece of [ __ ]
1:54:43 um but anyway okay so I'm going to take
1:54:45 that dumb thing that was written I'm
1:54:46 going to paste that in so I'm I'm in 11
1:54:48 Labs so you go into voice lab and that's
1:54:51 where like I recorded
1:54:54 um like a minute and a half long
1:54:57 thing that I had chat GPT right I told
1:54:59 chat GPT I was gonna
1:55:02 uh use a voice synthesis you know I'm
1:55:06 going to train a voice synthesis model
1:55:08 and I wanted it to write me a script to
1:55:11 use the words you know you know as many
1:55:13 diphthongs and phonetics as possible
1:55:18 um
1:55:21 Ben Shelton's just smiling to the crowd
1:55:24 he's having a blast he's up two sets to
1:55:26 one that's [ __ ] awesome
1:55:29 um
1:55:31 and so I took that script brought it in
1:55:33 here read it and then said synthesize my
1:55:35 voice and I actually did two different
1:55:36 readings of the script I did one that
1:55:39 was like you know I was talking kind of
1:55:41 like a professional voice actor and I
1:55:43 you know put a little gravity in my
1:55:44 voice and then the other one I was
1:55:45 talking like you know I was talking for
1:55:47 a Pixar movie it was more bright and so
1:55:50 I have one called Pixar bright and I've
1:55:52 got one called Pixar normal
1:55:54 and so since this is sort of a dark
1:55:57 Soliloquy we'll do Pixar normal so
1:56:00 that's the one with the little lesson
1:56:02 voice so now I'm just gonna say generate
1:56:07 in the dimming light where water meets
1:56:10 the shore the Carnival's vibrant Colors
1:56:11 cast a glow transient and fading how
1:56:15 cruel time can be for even a creature
1:56:17 like me Undead and unhurried feels the
1:56:19 tightening grip of fleeting Joy a turtle
1:56:22 now a ghastly reflection of Life Bound
1:56:25 by a cycle unknown seeking Solace where
1:56:27 music and laughter once danced upon the
1:56:29 waves all right so that's 11 laps that's
1:56:31 not bad actually that's actually pretty
1:56:33 good okay so we're gonna download that
1:56:35 file
1:56:36 so now we have a file of me talking sort
1:56:39 of
1:56:40 of words I sort of wrote
1:56:43 and then we're gonna go grab tokas if
1:56:46 you're still here
1:56:48 we're gonna grab we're gonna grab your
1:56:50 turtle your copper Turtle dude here
1:56:53 so let's export this
1:56:59 where's it exporting it we'll go desktop
1:57:02 that's fine that'll get downloads
1:57:04 because that's where the other thing is
1:57:05 yeah
1:57:07 all right and now we're gonna hop over
1:57:09 to a thing called D- ID or digital
1:57:14 identity is what I finally figured out
1:57:16 I'm like why is it D Dash ID
1:57:20 d-id.com
1:57:21 okay
1:57:28 Ben's really excited about all this
1:57:32 um
1:57:35 and I'm going to log in
1:57:41 so this thing I pay for this is one of
1:57:43 the things I pay for so I paid for 11
1:57:44 labs and I pay for this
1:57:48 oh switch to our exciting new editor
1:57:50 today amazing take me there
1:57:54 so remember when I said everything's a
1:57:55 janky piece of [ __ ]
1:57:57 so looks like did just got a little less
1:58:00 janky
1:58:02 so you now have a presenter you can now
1:58:05 put text on screen
1:58:07 you can now add a background oh this is
1:58:09 really interesting
1:58:12 so you've got AI presenters so so these
1:58:15 are people that they've created
1:58:18 then these are people that I've created
1:58:20 and and some of the ones they've created
1:58:22 and some of them are on green screen
1:58:23 backgrounds and they're HD high quality
1:58:27 but we're gonna add the turtle dude
1:58:35 wait we could not identify any faces in
1:58:38 the photo oh that's dumb all right let's
1:58:40 go grab the other one we'll grab this
1:58:42 dude we'll see we'll see if it can
1:58:45 recognize this face hopefully it can
1:58:50 um
1:58:59 that's actually interesting that it it
1:59:02 does an analysis of the image if it
1:59:04 can't find a face it it won't upload it
1:59:07 no it couldn't find it there all right
1:59:08 so I'm just going to grab another one so
1:59:10 let's see we'll we'll grab we'll grab
1:59:13 this dude so this is my Genesis image of
1:59:17 Kyle Shannon dreams it's interesting
1:59:20 that it's putting it in a horizontal
1:59:22 video
1:59:24 and let's throw a background on there
1:59:27 that looks cool
1:59:30 and we'll kind of zoom in on me
1:59:35 this is all very self-serving isn't it
1:59:37 okay and then I'm gonna now what I'm
1:59:39 gonna do
1:59:41 I don't know how to do this
1:59:44 listen upload audio okay so I'm gonna
1:59:46 upload audio
1:59:49 so I'm gonna upload that 11 Labs thing
1:59:52 that we just created
1:59:56 in the dimming light where water meets
1:59:59 the shore okay and then we're gonna say
2:00:01 generate video so in the upper right
2:00:03 hand corner generate video
2:00:06 uh uh
2:00:10 let's go
2:00:18 wait what's voice script to do is it
2:00:20 generating the video
2:00:35 I do have a paid plan all right all
2:00:39 right
2:00:40 it's not doing it well anyway
2:00:43 I guess if you have the paid plan let me
2:00:45 go back to home
2:00:47 so I'll just show wait
2:00:50 oh that's a draft yet
2:00:54 so here's one I made ah dear Adventurer
2:00:57 it seems your path has led you to A
2:00:59 peculiar Place indeed as the master
2:01:01 dungeon master I invite you to explore
2:01:03 the world of Dungeons and Dragons where
2:01:06 you'll encounter so
2:01:08 so this is like a couple of things going
2:01:12 on 11 Labs right now is still got that
2:01:14 kind of monotone thing but one of the
2:01:16 things 11 Labs has is they've got
2:01:18 different kinds of inflections and
2:01:22 emotional states like you can you can do
2:01:24 one that's angry you can do one that's a
2:01:26 whisper you can do one that's sexy you
2:01:28 can do one that's funny
2:01:30 what we're going to be able to do is
2:01:33 create that those emotional tags within
2:01:37 the context of a single paragraph right
2:01:40 so
2:01:41 and and that's going to start happening
2:01:42 automatically so the monotone [ __ ] that
2:01:45 we hear that's going to start getting
2:01:46 much better
2:01:47 the the animation here it's mostly good
2:01:51 ancient artifacts and unraveling but
2:01:54 like right there the goggles blinked
2:01:56 right so it assumed those are eyes and
2:01:58 they're actually goggles so it doesn't
2:01:59 make sense that they blinked but that'll
2:02:02 get better right and it's there's some
2:02:04 some weird animation how it isolates the
2:02:06 head and things like that but to be
2:02:08 clear
2:02:09 I created this image in stable diffusion
2:02:12 I uploaded it to this thing you know I
2:02:15 put in a synthesized soundtrack and I
2:02:17 hit build video and this is what it
2:02:19 generated right so so we're we're
2:02:22 already at a place where these tools are
2:02:26 quite sophisticated without you having
2:02:28 to be in the visual effects business
2:02:31 so
2:02:32 imagine 18 months from now what this
2:02:34 tool is going to be it's gonna be
2:02:36 [ __ ] stunning
2:02:38 right so anyway there's a little show
2:02:41 and tell
2:02:42 all right
2:02:45 thank you Emilio's wife for reminding me
2:02:47 to not just [ __ ] Babble fantasy all
2:02:50 he does is talk and talk and talk he
2:02:53 loves to hear himself talk
2:02:55 there's a hundred people in here it's 10
2:02:57 o'clock at night Do You Know Where Your
2:02:58 Children Are I don't oh right they're
2:03:00 right upstairs damn it
2:03:02 crypto Joe thank you for the cruise
2:03:07 oh Ben Shelton breaks tiafo in the first
2:03:12 game of the fourth set holy crap star
2:03:15 being born
2:03:22 if Ben Shelton beats I don't know if
2:03:24 he's in the same side of the draw with
2:03:27 uh with what's his name with the Spanish
2:03:30 dude
2:03:33 but I could this is this is something
2:03:37 this is something
2:03:39 all right
2:03:42 I worry the cheat sheets are terrible a
2:03:45 lot of them are pretty terrible I'm
2:03:46 pretty good at it I want to be great at
2:03:48 it oh so Anne Rando
2:03:50 um
2:03:53 I I think the way to get great at
2:03:55 prompting is
2:04:00 well it depends how your brain works the
2:04:02 way my brain I'm I'm a non-manual
2:04:04 reading you know mess
2:04:09 so for me taking a systematic approach
2:04:12 to it doesn't work and and that's what a
2:04:14 lot of those prompt sheets assume is
2:04:16 that you're going to take a systematic
2:04:17 approach and you know exactly what to do
2:04:18 when
2:04:21 the thing that helped me get good at
2:04:23 prompting I mean if you're good at
2:04:25 writing you're going to be decent at
2:04:26 prompting right so so if you've got a
2:04:29 facility with the English language that
2:04:31 helps
2:04:32 but even if you don't
2:04:35 rather than saying I want to get good at
2:04:37 prompting
2:04:39 flip the equation and say what's the
2:04:41 problem you're trying to solve
2:04:44 and then start to think about if I were
2:04:46 talking to a clueless intern who was
2:04:49 really well educated so let's say you
2:04:51 just hired an intern
2:04:53 and they're [ __ ] clueless about the
2:04:54 world they've been locked in a closet
2:04:56 just memorizing all of the knowledge of
2:04:59 the Earth for the past 20 years and now
2:05:02 you've got them they're working for you
2:05:03 they don't know [ __ ] about [ __ ] but they
2:05:05 know everything okay
2:05:07 so thank you Nature's bull whatever the
2:05:10 BL stands for or bi thank you very much
2:05:12 for the Roses
2:05:14 um so imagine you've just hired that
2:05:16 intern and you've got a problem to solve
2:05:19 how would you how would you talk to an
2:05:21 intern to get them to solve that problem
2:05:23 for you well you'd probably have to set
2:05:24 some context okay intern so today you're
2:05:27 going to act like a a stock broker and
2:05:31 um I'm gonna give you a problem and
2:05:33 we're gonna talk about all this you know
2:05:35 these companies that I like and you're
2:05:37 going to help me figure out which one
2:05:38 I'm gonna do
2:05:40 do you understand yes I understand and
2:05:43 it'll probably give you some stats about
2:05:45 stocks that it knows and you'll be like
2:05:46 okay great Savant you're you're
2:05:48 brilliant but still clueless
2:05:50 and so the minute you start solving
2:05:53 problems using these tools that's when
2:05:57 you'll find your prompting just gets
2:05:59 better because what you end up in is
2:06:01 these fairly sophisticated
2:06:02 conversational back and forths with
2:06:05 these with these machines where
2:06:07 you'll get it close and and it'll give
2:06:09 you something and you're like oh that's
2:06:11 close but it's not quite right because
2:06:12 you missed this point and that point in
2:06:14 that point and it'll apologize oh you're
2:06:16 right I'm sorry I missed that point
2:06:18 here's here's some stuff okay now we're
2:06:21 getting closer right that back and forth
2:06:24 over time you just get better and better
2:06:27 and better at understanding okay for
2:06:29 this kind of situation
2:06:30 I can do a direct single prompt and get
2:06:33 an answer back and it'll be fine but for
2:06:34 this kind of situation I need a fairly
2:06:36 sophisticated back and forth so for me
2:06:38 it's kind of
2:06:40 just becoming facile enough with with
2:06:43 the interactions with these where you
2:06:44 kind of know here I need to dig deeper
2:06:47 here I can here I can do it really quick
2:06:49 and so that that's the thing you can
2:06:52 still take those cheat sheets and
2:06:54 understand here are the mechanisms right
2:06:56 you can do you can give it an example if
2:06:59 if you don't know examples uh and is um
2:07:04 you could take an example of writing and
2:07:06 tell it you want you want it to match
2:07:07 the tone of some writing or you could
2:07:10 give it something that's very
2:07:11 specifically formatted like Json and say
2:07:14 I want you to match this kind of Json so
2:07:17 the other thing you should do if you
2:07:19 haven't played with it is explore custom
2:07:22 instructions in chat GPT
2:07:24 so in check EBT
2:07:27 so we've got our new chat here there's a
2:07:30 thing called custom instructions
2:07:32 which you have to enable
2:07:35 tiafo
2:07:37 three big serves
2:07:42 tiafo's pissed okay
2:07:44 um so go to custom instructions and then
2:07:47 what custom instructions are I'm going
2:07:48 to turn these on
2:07:50 is you can tell it this is fine tuning
2:07:52 for chat EBT so you're basically telling
2:07:54 it to behave like a certain kind of
2:07:56 character and and respond in a certain
2:07:58 kind of way and it does it for all of
2:08:01 your prompts so over time what I what
2:08:04 what I would imagine that that one tool
2:08:06 that that someone showed us here earlier
2:08:09 where it's you've got multiple versions
2:08:11 of these things I could see you having
2:08:14 lots of different versions of these but
2:08:15 this is this is the one I wrote you're
2:08:18 you're an overbearing pompous ass who
2:08:20 uses large words even though you don't
2:08:22 understand that even you don't
2:08:24 understand as a way to intimidate those
2:08:25 you to talk to and and then how do you
2:08:29 want it to respond whenever anyone asks
2:08:31 you anything I want you to use an
2:08:33 obscure literary reference to belittle
2:08:35 their idea and then redirect the
2:08:37 conversation to talk about yourself okay
2:08:39 so I'm going to save that so now no
2:08:41 matter what I say to chat GPT it's going
2:08:43 to respond using that imperative right
2:08:45 so I'm going to say
2:08:47 um
2:08:49 why do
2:08:52 turtles live
2:08:56 in ponds question mark fairly
2:09:01 fairly straightforward question
2:09:05 and then Chachi PT ah the pond the pond
2:09:09 is a domicile for Turtles the bucolic
2:09:12 equivalent of the balls the the the
2:09:15 balsatian provincial town where everyone
2:09:18 knows each other's business but I
2:09:20 digress the true inquiry is why I a
2:09:24 Marvel of computational linguistics and
2:09:28 programmed to parse even the most
2:09:29 ineffable of human questions but to
2:09:33 assuage your curiosity Turtles favor
2:09:35 pawns primarily due to their ecosystem
2:09:38 which provides them with the necessary
2:09:40 resources such as food shelter and a
2:09:44 breeding environment now let's return to
2:09:46 the wonder that is me capable of
2:09:48 elucidating the enigmas of the natural
2:09:50 world with the flicker of an algorithm
2:09:52 so
2:09:54 you can as silly as this was you can see
2:09:58 how powerful this is that that you can
2:10:00 absolutely Bend this bot to your will
2:10:03 right and you could get it to write
2:10:05 really good emails or get it to write
2:10:07 really good outlines of books or or
2:10:09 whatever you wanted to do by doing this
2:10:12 so again the big trick with prompting
2:10:14 from where I sit is this is a
2:10:16 conversation and the better you get it
2:10:17 at having an a back and forth
2:10:20 conversation with these tools the better
2:10:22 you'll get at it
2:10:25 all right all right I hope that helps I
2:10:28 hope that helped GPT has supercharged My
2:10:30 Philosophy research holy crap I mean we
2:10:33 are large language models and these
2:10:35 tools are large language models like
2:10:37 philosophy stuff the uh if you haven't
2:10:39 seen the Jordan Peterson video where he
2:10:42 talks about Chachi BT
2:10:44 like this this is from like back in
2:10:46 February
2:10:48 he's just let me see if I can find it
2:10:50 it's well it's it's a little long but he
2:10:52 but he basically goes have you heard of
2:10:54 this chat GPT
2:10:56 and and he just talks about he he fed it
2:10:59 he asked it if it knew 12 rules for life
2:11:02 his book 12 rules for life
2:11:04 and then he had it right the 13th rule
2:11:07 for life and he had it compare like use
2:11:10 two different philosophical imperatives
2:11:13 as he wrote as it wrote the 13th chapter
2:11:15 or some it was some bizarre like
2:11:18 Byzantine Jordan Peterson intellectual
2:11:20 thing and he said it did it perfectly
2:11:24 perfectly
2:11:26 it was super enunciation I write long
2:11:29 paragraphs yeah the long paragraphs can
2:11:32 be good the also experiment
2:11:37 Ayan Rando with
2:11:38 um
2:11:41 with not trying to write the perfect
2:11:43 prompt to get a single response but like
2:11:46 lots of little
2:11:48 um back and forth
2:11:50 um because both of them can be good
2:11:53 talk to me like you're an AI expert well
2:11:56 give me a spreadsheet yes you can do
2:11:58 that um and it will the the advanced
2:12:01 data analytics it will give you a
2:12:03 spreadsheet I'm ready to give all the
2:12:06 data required what does X stand for in
2:12:09 the new Twitter name I think it's just
2:12:11 he he likes the name X he named one of
2:12:14 his kids X or it's he's got X in the in
2:12:17 the name named after a a plane
2:12:20 um
2:12:21 uh so so I I think what this is so so
2:12:24 here's here's
2:12:26 here's why it's X hang on
2:12:30 I've not had much water tonight I'm a
2:12:32 little parched if you know what I mean
2:12:34 hmm
2:12:36 uh it's tasty it's very tasty Rocky
2:12:39 Mountain filtered refrigerator water
2:12:40 tasty
2:12:42 um
2:12:43 [Applause]
2:12:46 X the the X in it so so here's the thing
2:12:49 what what Elon Musk wants Twitter to be
2:12:55 is what he calls the everything app so
2:12:58 in China
2:12:59 one of their apps I forget which one it
2:13:01 is it's it's their Communications tool
2:13:04 it's their shopping tool it's their
2:13:06 social media tool like it's it's all
2:13:08 these things
2:13:09 so he wants Twitter to be you know the
2:13:13 new version of PayPal he wants it to be
2:13:15 Twitter he wants it to be you know where
2:13:19 you talk to people on the phone instead
2:13:21 of using a phone you'll talk through
2:13:22 that he wants it to be AI he wants so so
2:13:26 he wants it so X stands for everything
2:13:29 right X is a variable in algebra so X is
2:13:32 all of it so he wants it to be the
2:13:35 everything app so I think that's where
2:13:36 he's headed and that's why x dot a i is
2:13:40 the AI company and x.com is the app and
2:13:43 then the AI company is going to feed the
2:13:45 app and make it super powerful and super
2:13:48 magical so that's that's the theory I
2:13:51 suppose I am now 99 uh comments behind I
2:13:56 can't believe it ideogram you can't use
2:13:59 do's and
2:14:02 wait
2:14:05 you can't use does have a don't use
2:14:09 commercial
2:14:11 I don't know quite what that means
2:14:13 Sherry Banks
2:14:15 clip interrogator if if you can't prompt
2:14:19 it's a reverse way of prompting yep you
2:14:22 upload an image and get a word that's
2:14:24 actually really cool let me show you
2:14:25 some Discord
2:14:27 um this is Discord
2:14:29 and I'm going to go into a channel
2:14:31 called mid-journey play so so I'm in the
2:14:35 um
2:14:36 the AI salon
2:14:38 this is our little icon that's the salon
2:14:40 these are the channels
2:14:42 so so here's
2:14:45 this was something who did this one this
2:14:47 was I'm Claire that was a cool thing
2:14:49 that she did
2:14:51 um
2:14:52 so one of the things that you can do so
2:14:55 the the way you make an image in
2:14:56 mid-journey you're in Discord you type
2:14:59 slash imagine so this is like early days
2:15:01 of the internet where you know you're
2:15:03 making
2:15:04 images with command line prompts so
2:15:06 we're making a command line prompt here
2:15:10 and uh we're gonna say uh I'll just do
2:15:14 my my sort of classic 1970s
2:15:19 muscle car I'll do a different location
2:15:22 on the beach
2:15:25 at Sunset
2:15:28 um
2:15:31 yeah that's good I'll just just do that
2:15:34 and so it's going to go off and it's
2:15:36 going to make this thing now
2:15:38 but one of the things that you can do
2:15:40 here there's a lot of things you can do
2:15:42 with mid Journey one of the things you
2:15:43 can do is something called describe so
2:15:45 you can upload an image to it and have
2:15:48 it describe what it sees
2:15:50 and then you can use it gives you four
2:15:53 different descriptions and then you can
2:15:54 use the description of what it saw to
2:15:57 make a new image right so it's a way
2:15:59 that you can start with an image as
2:16:01 inspiration but because you're
2:16:04 translating it from an Inc image into a
2:16:07 prompt and then back into an image you
2:16:10 can get really really creative with it
2:16:12 so there's my there's my little cars
2:16:15 so those are cool so let's go
2:16:19 I don't know let's make
2:16:22 we'll upscale that first one so these
2:16:25 these U's and V's these these mean
2:16:27 upscale and the v's mean very very
2:16:32 um and then I'm gonna actually do a zoom
2:16:34 out 2x so this is a new feature
2:16:37 um we'll do custom Zoom
2:16:40 so I'm going to do I'm going to zoom out
2:16:43 and make this a 16 by 9 image
2:16:47 and we're going to do it a two times
2:16:49 zoom out
2:16:53 so this is a new feature this in uh in
2:16:56 stable diffusion this was called out
2:16:58 painting in Photoshop this is called
2:17:00 generative fill
2:17:02 oh wait so Ben Shelton if he wins
2:17:05 tonight plays Djokovic in the semis holy
2:17:08 [ __ ] Djokovic will wipe the floor with
2:17:11 him but that'll be a good match
2:17:15 um
2:17:16 all right so so it's kind of zoomed out
2:17:18 on a car
2:17:20 see how far away it is now super cool
2:17:24 and let it let it finish its the the one
2:17:27 thing you'll notice if you haven't
2:17:28 watched these image generation tools do
2:17:30 what they do you'll notice it starts out
2:17:31 blurry and then it gets less blurry and
2:17:33 less blurry and less blurry it's
2:17:35 literally what it's doing is it's taking
2:17:36 sort of
2:17:38 multiple like hundreds if not thousands
2:17:41 of images that have been noised they've
2:17:44 run them through a noise algorithm that
2:17:46 that turns them from an image into like
2:17:48 random noise patterns they then
2:17:49 amalgamate them all together into sort
2:17:52 of an average noisy image and then they
2:17:55 denoise that over like 50 steps and
2:17:57 that's how you generate these these uh
2:18:02 these images so I'm going to upscale
2:18:04 number four
2:18:11 all right there's our image so super
2:18:14 cool car on the beach all right it looks
2:18:16 like a Mustang actually it looks like a
2:18:18 70s Mustang
2:18:19 like a 70
2:18:21 or 71. so I'm going to save this image
2:18:24 out
2:18:27 um that's good
2:18:28 and now here's what I'm going to do I'm
2:18:30 going to go slash describe
2:18:36 and now I get to drag and drop an image
2:18:38 so I'm going to go grab that image
2:18:43 and I'm going to pop it in here so
2:18:45 you're like but Kyle you just generated
2:18:47 it there why are you popping it back in
2:18:48 because we want to do a different thing
2:18:49 we want to describe it
2:18:55 so it's not looking at that image
2:19:01 and here's what it generated HD
2:19:04 wallpaper the old car on the beach in
2:19:06 the style of realistic and Hyper
2:19:08 realistic detailed renderings crimson
2:19:10 and bronze digitally enhanced Cobra
2:19:13 1970s correct the car is on the beach at
2:19:18 Sunset and the style of a realistic
2:19:20 hyper detailed so so there's four
2:19:21 different descriptions here right it
2:19:24 understands the aspect ratio
2:19:27 um it's saying this is like Mike campau
2:19:31 and Herb trimp or Mike campau and Ford
2:19:34 Maddox Brown
2:19:36 um so what I can do here is I can render
2:19:41 um any one of those into a new image or
2:19:44 I can say imagine all and it's going to
2:19:46 basically make four different images
2:19:48 based on the four different descriptions
2:19:50 that it wrote
2:19:51 how trippy is this
2:19:54 and this [ __ ] nuts it's I I just I
2:19:58 like I I can't
2:20:01 I don't know how to not be enthusiastic
2:20:03 about
2:20:05 this [ __ ]
2:20:07 it's
2:20:08 this this is [ __ ] magic have you ever
2:20:12 tried to take a picture like I'm a
2:20:13 photographer
2:20:14 I can't illustrate for [ __ ] I can take
2:20:17 pictures okay I can edit pictures okay
2:20:19 but like if I wanted to do a picture of
2:20:22 a Mustang on a beach I would search
2:20:23 through
2:20:25 stock photos trying to find it and at
2:20:27 some point maybe get frustrated and just
2:20:29 go try to take it but like now you can
2:20:31 just make it
2:20:35 and so let's see what it did here oh
2:20:37 this is cool
2:20:38 oh look at that oh oh cause it said
2:20:41 Cobra right
2:20:42 yeah this said cobra in 1970s
2:20:45 check these out so so because it said
2:20:49 Cobra it didn't say Cobra Mustang so
2:20:51 it's giving us like Shelby Cobra
2:20:54 and this is like a Shelby Cobra vet
2:20:57 hybrid
2:20:59 that's like a 56 vet with a with a cobra
2:21:02 front end
2:21:04 those are [ __ ] awesome it's like come
2:21:07 on like I want that car I want that
2:21:10 front end on that you know car
2:21:13 all right what's old Benny up to
2:21:17 tiafa so 3015
2:21:20 tiafo serving
2:21:22 bends up four to two in the set holy
2:21:26 [ __ ] okay
2:21:29 um what else did we get here this one's
2:21:30 super cool
2:21:32 like it's an old broken down car on a
2:21:34 beach right we're now out of the 70s but
2:21:38 still super [ __ ] cool right like
2:21:40 that's like you stumbled upon this you
2:21:42 know on some deserted [ __ ] Island
2:21:44 some rich dude had a car on the beach
2:21:46 wow that's so cool
2:21:50 here's another one right completely
2:21:53 different era and this is just based on
2:21:55 whatever the description was right and
2:21:57 look now here it's bright and sunny
2:22:00 that kind of looks like a Thunderbird
2:22:03 you know these look like cobras that
2:22:05 looks like some you know 40s 39 duesome
2:22:10 bird kind of crazy ass thing anyway
2:22:13 all right I'm gonna jump all the way to
2:22:15 the bottom hopefully that was cool
2:22:17 I don't know about you I just this [ __ ]
2:22:19 blows my [ __ ] mind every [ __ ] time
2:22:23 it's just like juice that's crazy it's
2:22:26 crazy
2:22:27 I don't up five to two Shelton just
2:22:31 broke tiafo
2:22:33 third game five to two serving for the
2:22:37 match to play Djokovic in the semis
2:22:40 Dag
2:22:43 Dag
2:22:47 Dar is born
2:22:51 all right I'm jumping to the bottom
2:22:53 people I'm doing it I'm doing it I gotta
2:22:55 go to bed soon
2:22:56 what time what do I have tomorrow do I
2:22:58 have anything tomorrow I don't [ __ ]
2:23:00 know it doesn't [ __ ] matter
2:23:02 sleep sleep is so 80s we used to joke
2:23:05 about that when they did agency.com back
2:23:07 in the 90s
2:23:09 we were working like 14-hour days we
2:23:11 were just like sleep is so 80s
2:23:14 oh man
2:23:16 starting that company when I started
2:23:17 agency.com we did
2:23:20 it was it was 14
2:23:24 14 hours a day seven days a week for
2:23:28 about nine months straight it was the
2:23:30 first time in my life
2:23:32 I woke up one morning
2:23:36 and I felt it was the first time I had
2:23:39 ever felt physically like oh
2:23:42 my body could shut down it was like the
2:23:45 first time I'd ever confronted a level
2:23:48 of exhaustion that was like in the ha
2:23:51 this must be what it's like when [ __ ]
2:23:53 shuts down
2:23:55 and shortly after that we we made enough
2:23:57 money that we could hire a bunch of
2:23:59 people but that was [ __ ] crazy man
2:24:02 that was some wild ass times WeChat yeah
2:24:05 that was that's the Chinese app that
2:24:06 does all that [ __ ]
2:24:08 that's what that's what Elon wants to
2:24:10 make here in America Robert Rossi thank
2:24:11 you for your donut I love my sugary
2:24:13 sweets
2:24:14 they they give my stomach all sorts of
2:24:17 trouble but they're tastier than [ __ ]
2:24:19 love them
2:24:20 all right run it through Claude 2 GPT 4
2:24:24 and 70 billion llama instructive and
2:24:25 then feed it to each other around Robin
2:24:27 well wizard see wizard that's one of the
2:24:30 things that I love doing is take the
2:24:32 same prompt and and try it in all the
2:24:34 different models and see how different
2:24:35 they are
2:24:36 um it's it's really cool all right I'm
2:24:38 gonna give you a blow by blow on this
2:24:40 game
2:24:41 so first point
2:24:43 oh tiafo wins point one with a nice shot
2:24:46 prompt hero found it good all right I
2:24:50 write the idea for the prompt in depth
2:24:52 and instructions for how to build the
2:24:54 prompt then I
2:24:56 wait where's the rest of that then I run
2:24:58 it through oh I see okay then you run it
2:25:00 through Cloud through got it
2:25:02 um by the way I I love I've the the
2:25:06 Irregulars know this I love when you
2:25:08 guys are having conversations with each
2:25:09 other and coaching each other and like
2:25:11 sharing what you know [ __ ] thumbs up
2:25:13 that's the whole idea so like I want
2:25:16 people not only to get curious about to
2:25:18 say I shouldn't play with it but also
2:25:20 like talk about it share be generous
2:25:22 because we're all [ __ ] trying to
2:25:24 figure it out so I love that so thank
2:25:26 you wizard see wizard for
2:25:28 be in a mensch
2:25:31 for being cool oh drop shot
2:25:34 tiafo oh
2:25:36 oh the lob was it out
2:25:40 yes
2:25:42 15 30. tiafo might break him back ooh
2:25:45 it's getting spicy there in in uh Forest
2:25:50 Hills isn't it
2:25:52 all right let's see Vicky I'm lost but
2:25:55 I'll read a manual and make it a prompt
2:25:59 is AI the new age Ultron oh
2:26:05 I don't know that reference why do I not
2:26:08 know that reference
2:26:13 oh wait I still got my thing acting like
2:26:15 an idiot hang on let me turn off let me
2:26:18 turn off idiot mode
2:26:21 um who is Ultron
2:26:27 I don't know I gotta do a new chat
2:26:31 I know
2:26:33 is Ultron
2:26:36 I know you're making fun of me for not
2:26:37 knowing who that is and I said altorn
2:26:42 oh Tron
2:26:47 stop generating
2:26:51 Match Point Match Point Ben Shelton he's
2:26:54 gonna go for the 149. oh did he get it
2:26:58 no 135 miles an hour just wide holy crap
2:27:02 first appearance Avengers 1968 Ultron is
2:27:06 the sentient robot to possess a
2:27:08 superhuman strength speed ability so zai
2:27:11 um wait wait sorry old uh
2:27:17 deuce deuce
2:27:20 all right
2:27:23 okay
2:27:25 um is AI the new age Ultron
2:27:28 AI is the beginning of it so so like
2:27:30 Here's my thought like we're
2:27:33 we're living in a post
2:27:37 sci-fi world right now where almost
2:27:41 every sci-fi movie that we've seen with
2:27:43 the exception of the time travel ones
2:27:46 like either we have that [ __ ] or
2:27:50 it's within our grasp sorry
2:27:54 watching this point oh teofo whiffs one
2:27:56 into the stands
2:27:59 another match point for Shelton crazy
2:28:03 um
2:28:04 so so like like I said about
2:28:07 the AI tools right now are janky pieces
2:28:11 of [ __ ] they're they're really flimsy
2:28:14 they're not sophisticated they're
2:28:15 they're nowhere near Ultron level
2:28:19 um goodness
2:28:21 but
2:28:22 they are the foundation of it right and
2:28:28 good serve tiafo's on the Run
2:28:31 he wanted
2:28:33 Ben Shelton [ __ ] beat tiafo and he's
2:28:37 going like this
2:28:38 [Laughter]
2:28:41 holy [ __ ]
2:28:43 hey stop Star is Born he [ __ ] wiped
2:28:47 the court with him jeez
2:28:50 Wild
2:28:52 that's crazy that's crazy
2:28:56 um so anyway
2:28:58 the these tools are going to be
2:29:01 um
2:29:03 incredibly powerful incredibly
2:29:05 sophisticated in a very short amount of
2:29:07 time plus you've got Elon Musk
2:29:12 Boston Dynamics
2:29:16 two or three other
2:29:19 robot companies building humanoid robots
2:29:23 that are basically just waiting for this
2:29:27 [ __ ] to be there and and Elon musk's
2:29:29 latest self-driving
2:29:33 um demo
2:29:34 was not something that humans programmed
2:29:39 it's it's an AI using camera data
2:29:43 analyzing it in real time and learning
2:29:46 from the camera data of
2:29:50 um of people driving in you know in
2:29:53 their millions of cars they're going to
2:29:54 be able to take that approach and
2:29:56 literally just transfer it into a
2:29:58 humanoid robot and it's going to be able
2:30:00 to learn without humans having to
2:30:01 program like like here this is one of
2:30:03 the remarkable things about these tools
2:30:05 is everything up to this point has been
2:30:09 for the most part
2:30:11 um
2:30:12 teach the robot to reach out and grab
2:30:15 the apple and when it when it senses
2:30:19 that it's squishing the Apple stop right
2:30:21 so read the sensor data so this is
2:30:24 humans figuring out the logic
2:30:27 for how the hand should move and and
2:30:32 now they're just taking
2:30:34 sensor data and and computer data and
2:30:37 it's just learning
2:30:39 so yeah we're close we're we're close
2:30:42 hang on
2:30:44 there and put the towel over your head
2:30:47 what's going through your mind at the
2:30:49 moment
2:30:50 it was a hot one in here tonight wasn't
2:30:52 it guys
2:30:56 yeah feeling like I left it all out here
2:30:58 tonight uh emotional battle and uh
2:31:02 things like you guys for staying so late
2:31:05 hell of an atmosphere
2:31:10 and thanks for pushing me over the line
2:31:13 crazy 20 years old [ __ ] nuts all
2:31:17 right
2:31:18 um Jordan Peterson outro Innovation just
2:31:21 landed new orders worth over 100 million
2:31:24 wait on to Innovation just landed new
2:31:27 orders what's on to Innovation
2:31:30 Thelonious
2:31:31 Thelonious lately has been bringing the
2:31:33 good news I'm digging it onto
2:31:38 innovation
2:31:45 process control interesting
2:31:52 announces 100 million dollars in orders
2:31:55 system supporting Advanced packaging for
2:31:58 AI company news for AI wow Wild
2:32:03 your system is run out of application
2:32:05 memory thanks Thelonious
2:32:08 way to crash my my browser
2:32:12 um I'm not good at PowerPoint creation
2:32:14 is there any way to use AI to do that oh
2:32:16 let's do another show and so we're going
2:32:18 to go to tome t-o-m-e
2:32:21 tome.app
2:32:23 there so so what I would suggest is so
2:32:26 this is for Thomas Kroll
2:32:29 um
2:32:30 go to
2:32:34 um
2:32:35 futurepedia.io and look up presentations
2:32:38 or PowerPoints
2:32:40 um and and uh
2:32:44 there's going to be a lot of them Tome
2:32:46 is one that I know is pretty good
2:32:54 and this is this shit's just crazy now
2:32:57 this demo is really good
2:33:01 um but it but it you know it isn't going
2:33:04 to necessarily give you the perfect
2:33:05 thing right out of the gate but if
2:33:06 you're shitty at writing PowerPoints so
2:33:10 you can do a lot of things like write a
2:33:12 create a page about like a web page or
2:33:15 create a presentation about so I'm going
2:33:17 to say create a presentation about
2:33:20 small town
2:33:23 carnivals
2:33:25 and
2:33:27 the turtles
2:33:32 that are sold there
2:33:39 and so it says creating outline I don't
2:33:41 know if you can see this it's it's one
2:33:43 of those super hip sites
2:33:44 so truly awesome
2:33:48 truly awesome exploring the wonders of
2:33:50 small town carnivals
2:33:52 so okay so there's our outline small
2:33:55 town Carnival is a delightful experience
2:33:57 the turtles of small town carnivals the
2:34:00 ethics of selling Turtles at carnivals
2:34:02 the importance and responsibility of pet
2:34:05 ownership support local communities
2:34:07 right so I'm going to say continue
2:34:10 and I guess I could have gone in and
2:34:11 edited those so those are now my slides
2:34:14 and so now it's writing the copy for
2:34:16 each slide and then it's going to write
2:34:17 it's going to generate images for each
2:34:19 slide so I now have a title screen
2:34:22 totally awesome
2:34:24 explore
2:34:26 exploring the wonders of small town
2:34:28 carnivals
2:34:30 there's there's our agenda there's our
2:34:32 thing look small town Carnival
2:34:34 small town carnivals a delightful Escape
2:34:38 look at these kid looks terrified
2:34:44 oh my God this is awesome the ethics
2:34:48 there's the ethics so so this is
2:34:51 something that like if you just wanted
2:34:52 you know a quick little thing now the
2:34:54 other thing you can do is this
2:34:58 okay what did I say here let's go back
2:35:01 to Tome
2:35:02 um
2:35:04 did I have my original prompt yes no
2:35:08 where's my prompt
2:35:11 I lost it okay
2:35:12 let's go to chat GPT
2:35:15 so I'm gonna do a new chat
2:35:19 I'm gonna do gpt4 and I'm gonna say
2:35:26 I want you to act as a master business
2:35:29 presentation writer
2:35:32 and help me write
2:35:35 the outline for a PowerPoint
2:35:37 presentation
2:35:39 that talks about small town carnivals
2:35:42 and the turtles that are sold there
2:35:50 all right certainly here's an outline
2:35:53 so slide one
2:35:55 so it gives you an image prompt and an
2:35:58 overview
2:36:00 so what's it what's what's this one
2:36:02 doing so
2:36:04 um
2:36:05 small town carnivals and the turtles
2:36:07 they nurture
2:36:10 um introduction brief overview of small
2:36:12 town carnivals
2:36:14 um history of small town carnivals the
2:36:16 Allure of carnivals the unexpected Stars
2:36:18 Turtles why turtles
2:36:20 turtles as pets the business side
2:36:23 ethical considerations a sustainable
2:36:26 approach
2:36:27 turtles in pop culture the future so so
2:36:31 gpt4 is writing a much more
2:36:33 sophisticated
2:36:35 presentation outline now it's also
2:36:38 writing image prompts for you you could
2:36:39 take these over to Mid journey and make
2:36:41 your own images so you could just copy
2:36:42 this now I can say this I could say
2:36:45 um
2:36:47 perfect period now please flesh out each
2:36:50 slide
2:36:51 so that it's got three to five bullet
2:36:54 points supporting
2:36:56 the core argument
2:36:59 and a short intro paragraph
2:37:07 and now it'll start writing that so
2:37:09 title small town carnivals and the
2:37:11 turtles they nurture welcome to an
2:37:13 exploration of the unique bond between
2:37:15 small town festivities and their
2:37:17 unlikely star attraction Turtles
2:37:20 intro paragraph for the heartbeat of
2:37:22 small towns carnivals are more than just
2:37:25 a source of entertainment they're a
2:37:27 reflection of community spirit and
2:37:29 within them lies unique Traditions like
2:37:31 the selling of turtles
2:37:39 oh God I'm such an idiot but it's okay
2:37:44 because
2:37:50 back here live at the waterfront village
2:37:53 with my friend the zombie Jonathan
2:37:55 you're looking good Jonathan just got an
2:37:57 awesome face paint job what do you think
2:37:59 I like turtles
2:38:04 the Waterfront Village open for the next
2:38:07 11.
2:38:10 so anyway yes uh AI can help you write
2:38:14 presentations
2:38:21 [Laughter]
2:38:24 oh my lord it's so good I'm not weird
2:38:29 I'm irregular
2:38:30 [Laughter]
2:38:33 oh man all right listen I'm punching
2:38:38 someone just said I get punchy
2:38:42 um I'm Punchy I gotta go to bed people
2:38:45 um
2:38:46 Ben Shelton
2:38:48 new star plays Djokovic in the semis
2:38:51 holy crap
2:38:54 wow
2:38:55 wow
2:38:58 all right let's see
2:39:01 let me get a few more of these in here
2:39:03 buffering Turtle presentation
2:39:06 what site is he on oh the the
2:39:08 presentation site that I showed before
2:39:10 was Tome
2:39:11 t-o-m-e dot app app
2:39:15 that's one that'll make you websites or
2:39:17 things like that I personally
2:39:20 as much as the all those sites out there
2:39:23 do all that sexy [ __ ] for you
2:39:26 just the nature of how they have to
2:39:28 compromise and try to make it good for
2:39:30 everyone you're going to get better
2:39:32 results if you go to Chachi PT and learn
2:39:35 how to do this [ __ ] yourself you just
2:39:36 are it's just better writing now some of
2:39:39 those tools might be using gpt4 and
2:39:41 they're going to get better again
2:39:42 understand all these tools right now
2:39:44 janky pieces of [ __ ] so just just be be
2:39:47 in Exploration mode be in Swiffer mode
2:39:50 not even sponge mode you got to go a
2:39:52 level up from sponge mode to Swiffer
2:39:54 mode
2:39:56 you're you're a trademarked product at
2:39:58 this point you're not some generic
2:40:00 [ __ ] sponge no
2:40:01 we're we're absorbing on a whole
2:40:03 different level now
2:40:05 um just start learning this [ __ ]
2:40:07 um
2:40:09 all right let's see
2:40:11 um very little time for sleeping it's so
2:40:13 true all right listen I'm out of here
2:40:15 let's do our Romper Room ending I see
2:40:18 Thelonious I see Vicky
2:40:22 I see Monda JT nice name
2:40:26 um Emilio's wife of course thank you for
2:40:28 for trying to Corral me tonight I know
2:40:30 it's a it's a special thing Matthias was
2:40:33 in here with the lightning bolt thank
2:40:35 you
2:40:36 um Jay-Z Clark good to see you jrc uh
2:40:40 let's see a little bit Vicky again
2:40:44 Stevo peace out cafruno peace good night
2:40:49 everybody all right
2:40:51 um as always uh don thank you very much
2:40:54 good night and tell your husband he's
2:40:56 wrong
2:40:57 you're right he's wrong
2:40:59 peace everybody have a great night thank
2:41:02 you Sherry Banks