aiLL 7/9/23 - 2 - Jump In! AI Is Here and It's Mind-Blowing

Description

Chapters

Transcript

0:00 foreign
0:22 name's Kyle Shannon
0:27 I'll lead the adventure this evening
0:32 this is my second live today
0:35 and uh just had some ribs and
0:39 did a little yard work hooked up the
0:41 hose
0:42 because it it's been raining so much
0:44 here in Denver we haven't had to water
0:45 anything so it's all watered now looks
0:48 like the next week's gonna be sun
0:50 so we'll see
0:54 welcome welcome welcome feel free to
0:56 share the live and hit the like screen
0:58 let people know we're here make them
1:00 believe like you're enjoying this and
1:02 then we can we can fool them we'll get
1:05 them in here and then you know and then
1:07 we'll extract cash from them like we do
1:10 every night right
1:11 God man I'm [ __ ] rolling in the dough
1:14 from these lives you you have no idea oh
1:18 my God it's fantastic
1:21 I think I think I'm up to
1:23 3.4 cents an hour
1:26 so at this rate if I had a nickel for
1:29 every hour I'd be getting a raise
1:34 anyway uh this is the AI learning lab
1:37 we'll wait for some more folks to come
1:39 in looks like the uh looks like the algo
1:41 Rhythm there he is hey Silver Fox what's
1:44 happening congratulations on uh building
1:48 a thing for a someone
1:50 with your restaurant knowledge that's
1:52 really exciting that's really flipping
1:54 exciting
1:56 um
1:57 and I I put this in in uh in the
2:00 comments but I think it's the right move
2:01 I think starting with service is is a
2:03 good way to go because all of this stuff
2:05 is going to change so fast that if you
2:08 spend too much time building something
2:09 you're going to build something that's
2:10 obsolete in three months and then you're
2:12 gonna be cranky and pissed off anyway
2:14 welcome everybody my name is Kyle
2:16 Shannon this is the AI learning lab all
2:18 these URLs around my head here I'll flip
2:21 it in a second
2:23 but here's why these are here if you
2:25 have not played with chat GPT I'll talk
2:27 about what the what the channel is about
2:29 and who I am if if that's of interest
2:31 but if you haven't played with chat GPT
2:33 while I while I chat here Sunday night
2:36 class
2:37 um go to chat.openai.com that's the
2:40 official chatgpt website and just start
2:43 playing with it if you don't have any
2:45 clue what to do there go to prompts.chat
2:47 that document down there
2:49 and that'll that'll give you
2:50 instructions on what prompting is and
2:52 how to talk with this thing it kind of
2:54 looks on the surface like a Google
2:56 search bar but it goes way deeper these
2:59 are some deep Waters deep deep Waters uh
3:02 that's chat EBD and then bing.com is
3:04 Microsoft search engine but if you click
3:06 on the chat button that's also chat GPT
3:09 but it's gpt4 connected to the internet
3:11 and it's free and you can make images
3:14 there you just say make me an image of a
3:16 penguin and a superhero costume and it
3:18 will and then poe.com gives you access
3:20 to six different chat GPT large language
3:23 models or not Chachi well chatgpt like
3:26 large language models including chat gbt
3:29 okay and then futurepedia.io that is uh
3:33 just a big old directory of AI tools so
3:36 let me let me flip these
3:37 so so while I ramble go there go there
3:40 now and go there until you have your
3:43 Kevin McAllister moment play with it
3:45 until this happens
3:49 and then you're ready to get started
3:55 okay so the way this works pop any
3:57 questions or comments you have uh down
4:00 in the uh comments below you can ask
4:02 questions about AI I'll do what I can to
4:04 answer them or point you in directions
4:07 um I do snarky commentary I cuss
4:10 occasionally so if you don't like that
4:12 get on out
4:15 and uh what else uh
4:18 uh I'm an entrepreneur
4:20 I've started 12 or 15 companies the uh
4:24 the re the this channel exists to talk
4:27 about generative Ai and
4:29 um the I I sort of see this this
4:32 inflection point of the stuff I talk
4:35 about started on November 30th 2022 so
4:37 I'm not a machine learning guy I'm not a
4:39 data scientist I'm not an engineer I'm
4:42 an entrepreneur but I love technology
4:44 and I love transformative technology the
4:47 last time I was even remotely this
4:49 excited about anything was in 1994 when
4:52 I stumbled upon this thing called the
4:53 World Wide Web
4:54 I started one of the first online
4:56 magazines I started one of the first
4:58 digital agencies
4:59 and I didn't think in my lifetime I
5:03 would be alive to see another
5:06 thing that transformative thank you
5:08 thank you for the hearty heart
5:11 uh something that transformative and
5:13 then Along Comes generative AI chat GPT
5:16 and mid-journey and stable infusion
5:18 those things and it's going to change
5:20 everything in fact my contention is it's
5:23 already changed everything it's just no
5:25 one knows it yet well 100 100 million
5:27 people a couple hundred million people
5:29 know it
5:30 um but
5:32 uh it's coming and it's coming fast and
5:34 I think faster than all of any of us are
5:36 prepared for so that's what this channel
5:38 is all about
5:40 um what do I like most about AI
5:41 technology
5:43 um
5:45 it is
5:50 somewhere between
5:54 having magical powers and having super
5:56 powers
5:57 Arthur C Clarke says any sufficiently
6:00 advanced technology is indistinguishable
6:03 from Magic it feels like magic
6:06 um
6:07 it felt like magic the first time I had
6:10 it write a python application for me and
6:12 I don't write python
6:15 um and it still feels like magic today
6:17 and then as the technology advances and
6:20 gets more and more sophisticated I get
6:22 more and more excited normally with
6:24 stuff like this like when I was playing
6:26 around with blockchain and nfts and
6:28 things like that I got super excited
6:29 about it initially and the more I dug in
6:32 the less real it felt the technology
6:35 felt real but the companies the projects
6:37 the crypto Bros just all that stuff was
6:39 just very surface level this is
6:42 completely the opposite
6:44 um
6:45 the harder you push it the better it
6:47 gets the deeper you dig the more
6:49 interesting it gets the more profound it
6:51 gets
6:53 um
6:54 and I've never felt more creatively
6:57 alive and free as I have in the past six
7:01 months because
7:03 this removes
7:05 creative blocks like like the idea of
7:08 writer's block or creative block or or
7:10 any of that like disappears with this
7:12 stuff so so that's it for me it's kind
7:15 of like it takes who you are and
7:17 amplifies it and it lets you to take
7:20 little insignificant ideas you have and
7:23 you kind of put them in these machines
7:24 and they come out like
7:27 bigger than you imagined much faster
7:29 than you imagine
7:31 and then you get to start playing with
7:33 them and refining them and
7:35 so
7:37 yeah I think that's the thing I like
7:39 most about it all right
7:43 Sunday night class bard
7:48 ee
7:51 hello Tobias what's happening
7:54 yeah feel free to if you want to get
7:57 other folks in here feel free to share
7:59 the live double tap tap the like but
8:01 like button looks like we hit a a bunch
8:03 of people and they all they all they saw
8:05 my face and they were like yeah yeah and
8:08 they took off so let's get some more
8:10 folks in here if we can
8:14 feel free to pop any comments down down
8:16 below I had my moment today using Vector
8:19 Journey awesome I don't know Vector
8:21 Journey your home alone moment
8:24 what's Vector Journey let me go look at
8:27 that
8:30 vector
8:35 Journey AI
8:40 oh
8:44 I don't know what Vector journey is you
8:46 mean mid-journey
8:49 or is there some other tool I'm missing
8:54 where do you think AI will be in a year
8:56 I'm glad you didn't say five years
8:57 because five years I don't have a
8:58 [ __ ] clue a year okay
9:03 in a year
9:06 so this time
9:08 July 2024
9:10 gpt5 will be out
9:17 it will be truly multimodal meaning
9:22 you can input audio video image words
9:27 and it will output all of those as well
9:33 we will likely have the first iterations
9:37 of personal agents that are trained on
9:40 our data
9:41 so rather than like having to go to chat
9:44 GPT and figure out what you want to
9:46 prompt it to do and do I need to add
9:49 plugins
9:51 like all the [ __ ] it is now
9:55 will will be more used to it and and
9:59 the the tools will have more autonomy so
10:02 we'll just say like go book me a trip or
10:04 go
10:06 get United Healthcare to pay for my
10:09 doctor's appointment or go reschedule my
10:11 this or go book a whatever haircut and
10:15 it'll just go do it so I think that's
10:17 gonna happen I think
10:21 generative
10:22 multimedia like generated video and
10:25 generative music and all that sort of
10:28 stuff
10:29 will be incredibly sophisticated
10:34 um
10:35 I think that
10:37 will be five months out from the 2024
10:41 election
10:42 and so I think the
10:46 the internets will be full of
10:50 misinformation deep fakes crazy [ __ ]
10:53 like that
10:56 oh
10:58 and I think that it so so
11:03 whatever chat gbt becomes will be
11:06 incorporated into all of Microsoft
11:08 products all of Google's products
11:12 anything you do on AWS
11:16 deeply integrated Within meta's products
11:19 and apple will have launched some AI
11:22 something probably personal agents is my
11:25 guess with Apple
11:29 I think that's oh and then the other
11:31 thing is
11:32 within businesses AI will not like right
11:35 now people are like
11:37 well I don't know if AI is real and you
11:40 know is it really ready for business and
11:42 I don't know what I'm doing back to bed
11:44 to bed
11:46 those are the CEOs I'm making fun of
11:51 um
11:52 I think most businesses will will be
11:56 will either jump in with both feet and
11:59 have ai strategies or we'll be forced to
12:02 have ai strategies I don't think I don't
12:04 think there will be many businesses a
12:06 year from now that can ignore this I
12:09 think it's just going to be too too far
12:11 into what we do
12:14 that was a lot good Lord does this man
12:17 ever shut up I don't know why are you
12:21 watching him I have no idea
12:25 hey Bob what's happening
12:27 hey could you grab me a cold one awesome
12:34 this guy again oh good lord I absolutely
12:38 love Bing but I haven't tried chat gbt
12:40 yet it so so you have
12:45 Bing chat is gpt4
12:48 it's just that it's it's uh it's
12:51 Microsoft's version of it basically so
12:53 you've got the open AI chat.openai.com
12:57 is chat GPT that's the original and then
13:01 Bing chat is built on top of gpt4 so
13:05 it's the same underlying uh large
13:08 language model that powers both of them
13:09 they just have slightly different
13:10 implementations and yeah I mean a couple
13:13 of things about Bing right now
13:15 um the the official chat GPT they pulled
13:18 their connection to the internet because
13:20 it was
13:21 um scraping content from behind paywalls
13:23 and apparently it did it to Fortune
13:25 Magazine and they time Inc didn't take a
13:28 shine to having their content stolen by
13:30 a large language model apparently
13:32 so they have to shut that pumping down
13:34 so bing.com right now is the only one of
13:37 these large language models directly
13:39 connected to the internet it also you
13:42 can generate images with bing.com you
13:45 can also upload images to bing.com and
13:48 it can read them
13:50 like you can upload a picture of a
13:52 unicorn and it will tell you it's a
13:53 unicorn with sunshine and rainbows and
13:55 Skittles
13:59 so yeah Bing's Bing's pretty solid you
14:02 are correct wait what was I correct
14:03 about God dang it tick tock it just
14:06 jumped me to the bottom again
14:09 this guy seems like he doesn't know how
14:11 to use tick tock shut up Marge he's
14:13 talking about AI
14:14 he's talking about AI
14:17 God lord
14:19 God lordy
14:22 was that like Minnesota Irish
14:27 all right where was I
14:31 I had my moment today where will I be in
14:33 a year I became number one wait someone
14:36 became number 41 member on the team
14:38 congratulations Susie Beach toes you're
14:41 number 41 on the AI learning lab team I
14:44 have no idea what that means but that's
14:46 awesome I think
14:47 [Laughter]
14:49 uh there's I love Bing you are correct
14:53 about the oh here we go the Creator
14:54 Creator block statement yeah uh telepo
14:58 the um
15:02 there's there's just something about
15:04 Chachi PT's ability to instantly
15:09 take vague ideas
15:12 and just go I got you you're like I kind
15:15 of have an idea generally
15:18 in crappy English with typos and you're
15:22 like you got any thoughts on that of
15:24 course I've got thoughts on that
15:27 and then somewhere in there you're like
15:29 oh yeah that's it that's what I was
15:30 thinking
15:31 but that's not quite right do it like
15:32 this do it like that and it it is
15:35 I
15:39 like it's a weird emotional thing
15:44 so okay so before if I had an idea for
15:47 something
15:48 and
15:51 and the next step was okay in order to
15:54 bring that idea to in any form of
15:56 existence I gotta go to Google I gotta
16:00 search for some sort of Wikipedia page
16:03 or some sort of something I gotta read
16:05 it I got it in it like if it had more
16:08 than five steps it had to be like a
16:11 radically compelling idea that like kept
16:13 me up at night
16:15 to motivate me to go do that horrible
16:18 [ __ ] work right like the the
16:20 blocking and tackling work of creative
16:22 ideation
16:23 and now with this thing you kind of have
16:26 that Vapor of an idea and you stick it
16:28 in there and it's just like
16:30 and you're like oh
16:32 that's good
16:36 a little more
16:42 that's just
16:44 if you haven't used it
16:46 why do I have this stuff on screen while
16:48 I talk
16:49 I'm not getting paid by them
16:52 if anybody knows anyone on open AI this
16:54 channel is like a [ __ ] ad for them
16:58 I'm not being serious about that I don't
17:00 care if they pay me
17:01 um
17:02 but that's why I have it on screen if
17:04 you haven't played with it go there go
17:06 play with it any one of those three
17:10 how do I feel about AI becoming a
17:13 catch-all for ML and other things I
17:15 think it is
17:17 I
17:19 I think machine learning is more
17:20 accurate
17:22 um but AI is what they call it in the
17:24 movies and tough [ __ ] ai's got a better
17:27 PR firm than ml does
17:31 I don't really care
17:33 I you know I think I think
17:36 I I think those distinctions are
17:38 important for the people building the
17:41 technology and for the researchers and
17:42 the scientists to understand the
17:44 difference between those
17:45 but for the 150 or 200 people million
17:49 people that are using chat GPT that
17:52 don't know the history of machine
17:53 learning and AI doesn't matter to them
17:57 what it is to them is it's the magic
17:59 thing that writes their project plans
18:01 now or it's the Met it's the thing that
18:03 helps them write their Great American
18:04 novelers the thing that helps them
18:06 organize their Garden they don't give a
18:07 [ __ ] what it's called
18:10 um
18:12 like for example
18:15 open ai's new
18:17 mind-blowing face melting tool
18:21 is called code interpreter
18:23 because from a technical perspective
18:25 what it's doing is called code
18:27 interpreting
18:29 human beings don't know that it's a
18:30 shitty name
18:32 it's a shitty name for a product that's
18:34 got a consumer product or a consumer
18:38 user base
18:40 it should be called like
18:42 my magic programmer or [ __ ] [ __ ]
18:46 anything other than code interpreter
18:47 which makes it sound like good Lord do I
18:50 take a nap now or when I start using it
18:54 but when you use it it's face melting
19:00 you upload a file to it
19:03 and say what is this and it starts
19:05 writing python code to figure out what
19:08 it is executing the python code and then
19:11 writing its own instructions to help you
19:14 understand what's in the [ __ ] file
19:19 that's insanity
19:22 well we've named it code interpreter
19:24 because from a technical perspective
19:25 that's the most accurate description of
19:27 what's Happening from from the sense of
19:29 uh you know it's also quite limited it's
19:31 just a python code at this point I mean
19:33 of course in the future it could could
19:35 of course be expanded to include other
19:37 languages and it could do things Beyond
19:40 interpreting code of course and right
19:43 right now the the the modalities are who
19:45 gives a [ __ ]
19:49 give it a good name
19:51 and tell people there's a there's a
19:53 little magic box inside that thing
19:58 that is going to change everything
20:04 maybe that's what we should do tonight
20:06 is is go to chat gbt and come up with a
20:08 new name for code interpreter because
20:10 it's the worst [ __ ] name
20:12 I bitched about that somewhere
20:15 and someone made a comment made a
20:17 comment to me well it's the accurate
20:19 description of what it is
20:24 so I don't care it's the answer to your
20:26 question sorry about that I know an
20:28 awful little bit there but you know
20:30 I just had ribs
20:34 I'm supposing I'm supposed my blood
20:37 pressure's a little spiked
20:42 um have I experimented with the pi AI I
20:45 have
20:47 um
20:47 installed it you install that one yeah I
20:51 think I think it's an app it's a yes
20:53 it's an iPhone app
20:55 and I talked to it and it responded to
20:58 me
20:59 but that's about that's about as far as
21:01 I got
21:02 um I'm definitely paying attention to it
21:04 because they just got funded for 1.3
21:06 billion dollars for their series B
21:08 that's a relatively significant series B
21:11 if even in the world of of
21:13 ridiculously overblown series B
21:17 valuations that's that's right up there
21:19 um so I'm paying so I'm now paying and I
21:21 think they're called inflection is the
21:22 name of the company behind pie
21:25 so I'm paying attention to them I'm
21:27 paying attention to anthropic which
21:29 anthropic I like a lot although
21:33 the sense I have about anthropic is I
21:36 bet they're building some amazing [ __ ]
21:37 we haven't seen yet
21:39 and they have no marketing department
21:41 because you just don't hear anything
21:42 about them but like the poe or the
21:45 Claude 100K model that you can play with
21:48 at poe.com 75
21:50 000 word context window holy [ __ ]
21:53 so I have a feeling anthropic's gonna be
21:55 big I'm paying attention to them
21:56 inflection open AI Microsoft
22:02 we'll see what Google does
22:04 I figure Apple will show up in a year
22:06 Apple should probably
22:09 I let's see
22:11 so they do their thing in June
22:14 and then when's the next thing September
22:17 June is the developers conference
22:19 September is the new products is that
22:20 how Apple does their thing or October
22:24 um
22:26 we might not hear anything from Apple on
22:28 AI this next time but but probably by
22:31 next June they'll be a pretty
22:33 significant announcement from them or or
22:35 they'll just be something built into the
22:37 operating systems what AI program Graham
22:40 can I use to copy a website
22:44 um
22:46 there are I would go to futurepedia.io
22:49 and go to there's there's probably a web
22:52 design section there I know there's some
22:53 that copy websites there's some where
22:56 you can you pop in a URL
22:58 it scrapes the website and then
23:00 separates it into a figma a figma design
23:04 Library based on that website so I know
23:07 there's some that do it I just I haven't
23:08 done it so
23:10 um I would go to futurepedia.io that
23:12 bottom one there that's why that's there
23:16 futurepedia.io its purpose on this list
23:18 is not that I love it it's just that it
23:21 keeps track of all these tools because I
23:23 can't
23:24 so I think it's what the guy the guy's
23:26 name is Matt wolf that started that he's
23:28 got a YouTube channel
23:31 um the ability to aify yourself is
23:33 coming very soon upload your digital
23:35 footprint getting it AI yeah I agree
23:37 Dave I think I think that's going to be
23:39 I think I think that's kind of the next
23:41 big thing right where it somewhere
23:44 between autonomous agents and personal
23:46 personal agents or you know personal
23:48 assistants or whatever we're going to
23:49 call them pis
23:51 um
23:52 yeah I think I think that's next because
23:54 because
23:57 like I only give so much of a [ __ ] about
23:59 outside data unless I'm a data scientist
24:02 or unless I have a client that's doing
24:03 it but everyone on the planet would like
24:07 to be able to leverage like if you think
24:10 about
24:11 depending on how long you've been on the
24:13 planet I've been on the planet for a
24:15 fair [ __ ] time and I've been on the
24:17 internet in a pretty significant way
24:20 since
24:22 late 80s early 90s
24:26 in some form or other and
24:29 wouldn't it be amazing to be able to
24:31 just
24:32 be rewarded for all of my digital
24:34 hoarding
24:35 right I've got all these hard drives
24:37 full of data I've got all these I've got
24:40 a MySpace page somewhere I've kind of
24:43 [ __ ] you know Facebook for a decade
24:46 and a half right we all have all this
24:48 [ __ ] everywhere wouldn't it be amazing
24:50 to just be able to
24:52 point this tool to all that and have it
24:54 kind of synthesize it and assimilate it
24:57 into some usable thing
24:59 that can that can amplify my
25:02 contribution
25:04 that would be amazing
25:06 so yeah I agree I think that's coming I
25:09 think that's going to be big
25:10 have you heard of people deleting their
25:12 negative comments about AI in case it
25:15 ever takes over the future I have heard
25:17 that I think it's I think it's listen
25:20 here's what I here's what I appreciate
25:22 about that it's at least critical
25:25 thinking like I the thing that
25:28 you know I evangelize the most on this
25:30 channel is get curious about this [ __ ]
25:32 like understand what it is understand
25:34 how it does what it does and what it
25:37 makes possible and all that but in the
25:39 end I just want people to be curious
25:41 about it and so if someone's curious
25:42 enough to go you know if this thing gets
25:45 powerful enough and I say shitty things
25:47 about Ai and it doesn't like that
25:51 um
25:52 yeah I'm gonna go get rid of that [ __ ]
25:55 [Laughter]
25:57 but then listen if if you're gonna go
26:00 down that paranoia rabbit hole then you
26:02 got to get rid of your Alexa because you
26:04 know you got to go delete all your
26:05 Amazon data and your Siri data and your
26:08 Google data like all the [ __ ]
26:10 machines that are listening to us
26:12 because all the stuff that you're saying
26:14 to your friends on the phone call that's
26:16 being recorded too it ain't just what
26:18 you typed on Tick Tock comments
26:21 foreign
26:23 so I think it's cute
26:30 I think people give
26:33 corporations and AI more credit like
26:35 like I I think people think they're more
26:38 important than they really are like if
26:39 you're actually a really important
26:40 person that would be targeted maybe so
26:44 I mean we're all just doing our thing
26:45 here I don't think AI or
26:47 corporations generally give too much of
26:49 a [ __ ] about what we do individually
26:51 they just want our money
26:54 uh
26:56 uh uh
26:57 real-time voice cloning is already here
27:00 imagine a personal AI powered voicemail
27:02 screening service Apple has already
27:04 announced that that's coming in fact in
27:07 the new version well not quite what
27:09 you're saying but the voice cloning
27:10 piece in the new operating system you're
27:13 going to be able to have a virtual
27:15 a synthesized version of you that can
27:18 like respond to voicemails when you're
27:20 driving and for people that can't talk
27:22 it's going to be able to do something
27:24 for them so
27:26 um yeah I agree I think that's coming
27:30 Dave's on top of it slash home slash
27:32 Dave I dig it
27:34 he's in it love your lives thank you
27:37 very much Nature's Blessing I appreciate
27:40 that
27:41 you have a Charming personality
27:45 there's there's a line from the Cherry
27:47 Orchard
27:50 Mama do you think I'm pretty
27:53 you have beautiful hair
27:55 [Laughter]
27:59 that nice personality thank you
28:02 thank you
28:02 see that these are the comments I like
28:06 in my lives it's a compliment right
28:08 right
28:09 laughs
28:12 oh it's good it's solid
28:15 all right let's see where are we so I'm
28:19 finding myself in the uh
28:21 comments here
28:31 [Music]
28:34 love your lives there's the Charming
28:37 personality you have a Charming
28:39 personality Mama am I pretty you have
28:41 beautiful hair
28:43 oh hey Joy
28:45 I feel like AI is just a buzzword at
28:48 this point still Eric mole Eric Eric
28:53 have you used it
28:59 AI is absolutely over hyped
29:03 this is absolutely over hyped this is as
29:06 overhyped if not more than blockchain
29:08 and nfts and we all know where they are
29:10 right now
29:12 they're in a turtlit
29:15 this is different if you have not played
29:17 with chat GPT Eric go there now
29:21 go there now and if you don't know what
29:23 to do there go to prompts.chat and it
29:25 will tell you what to do there it's just
29:27 like a Google search bar tell it to
29:29 write you a country song describe your
29:31 pet or your significant other and tell
29:33 it to write you a country song about
29:35 them
29:38 and then ask it a really hard question
29:40 about something you know something about
29:45 so yes it's over hyped yes it's a
29:47 buzzword
29:49 and it's the most profound technology
29:52 ever put in the hands of humans
29:55 both and the genius of and
30:00 I use it for song lyrics yep it's good
30:02 for that it's really good for that gpt4
30:04 in particular is quite good for that
30:06 can you type in images and prompt it to
30:09 make different variations kind of yeah
30:11 yeah you can it's
30:13 I wish it were that simple it's not
30:16 quite that simple so there there are a
30:18 number of different tools these are
30:21 things for you to go watch YouTube
30:23 videos on
30:24 um
30:25 mid Journey for image generation tools
30:29 is the one that's the the best in my
30:30 opinion right now it's the most
30:31 sophisticated it's got a thing called
30:35 um blend mode and you can also upload an
30:38 image and then prompt against it and it
30:39 will sort of do a variation of it
30:42 so there's mid-journey
30:45 stable diffusion has a number of
30:47 different things including a thing
30:49 called control net
30:51 which allows you to
30:53 upload images to use as kind of the the
30:55 base
30:57 um
30:58 layout of the image and then you can do
31:00 new stuff with it
31:02 um
31:02 stable diffusion is good but it's
31:05 complicated and you need to be pretty
31:07 geeky
31:08 then there's photoshop's generative fill
31:11 and generative feels really powerful you
31:14 could upload a picture of a person and
31:16 like select everything but the person
31:18 like the whole background and say
31:20 you know put them in a snowy Forest
31:22 rather than the beach that they're on
31:24 and it will and then you can select like
31:26 their coat and say give them a puffy
31:29 parka and you can select their hair and
31:30 put a hat on them so so there's there's
31:33 lots of different ways to do it but
31:37 um
31:42 it's not as
31:49 the the generative image stuff right now
31:51 the the the the thing the the thing that
31:54 I'll say about it that's the most
31:56 frustrating
31:57 is that it's
31:59 it's pretty hard to get predictable
32:01 results or if you get a result you like
32:03 and then you want to recreate kind of
32:05 exactly that character in a different
32:07 scene
32:08 that's not super easy right now
32:11 generating newly original stuff super
32:14 easy super awesome just go to Mid
32:16 Journey generate [ __ ] but if you want to
32:18 take that person put them in a different
32:19 scene and and and have it remember that
32:23 and all that sort of stuff it's it's
32:24 just it's a little bit of spitting duct
32:26 tape right now for that kind of work my
32:28 new BFF is code interpreting that's
32:31 awesome
32:35 interpreter amazing absolutely amazing
32:41 um that's a good idea new name for code
32:43 interpreter I might do that that might
32:45 be something fun to do tonight I helped
32:47 raise 39 million dollars for series a
32:49 and was thrilled oh that's awesome
32:51 for your series a or for someone else's
32:54 like either way that's that's that's a
32:56 solid series eh
32:59 um do you watch the all-in Pod they were
33:02 talking about pie in today's podcast
33:04 yeah I watched those guys I actually I
33:07 know Jason falcanus pretty well he he
33:09 was the editor and publisher of this
33:11 back in the day that's me
33:14 I know Jason pretty well
33:20 you know how you can get rich with
33:22 yeah exactly
33:25 you give me money
33:30 you know we should do some night is
33:32 we'll get everybody ready to start
33:34 tapping the like button
33:36 and then I'll just I'll just say I'll
33:38 just say that over and over like 30
33:40 times in a row so that people flicking
33:42 through Tick Tock just see that and
33:44 they're like I can make money with
33:45 chatgpt
33:46 foreign
33:48 and they'll be disappointed with me like
33:51 figure out a good business and then use
33:54 AI That's how you make money with chat
33:55 gbt business first AI second
33:59 it's such a disappointment when you you
34:02 know when you're just like I want to get
34:03 rich quick
34:04 sorry
34:07 go make a course on how to get rich with
34:09 chat EVT and make an annoying video and
34:11 maybe you'll get rich
34:17 chat GPT can write the course
34:21 um is that what Zucker Buck wait is that
34:23 what Zuckerberg done for Twitter oh copy
34:26 and paste yeah exactly
34:28 3G classic rock that's what's awesome
34:30 isn't it you can get rich but touching
34:32 me hey
34:34 um if you put it on the Internet it's
34:36 there for eternity I know like but we
34:39 all thought that I'll put it on the
34:41 internet no one will ever really find it
34:42 no one's gonna go looking for it well
34:44 now ai can just like
34:46 bring it all back to life
34:48 as if we intended it like that oh well
34:52 actually you know what oh that's okay so
34:56 so these lives for the past few nights
34:58 have become business sites I'll give you
35:00 a business idea you want to go start you
35:01 want to get you want to make money
35:04 here's one
35:07 um create a tool that goes and scrapes
35:10 someone's Social graph
35:14 and identifies potentially embarrassing
35:17 [ __ ] for them and then they can choose
35:20 whether or not to delete it
35:22 that would be perfect for AI right so so
35:25 I just log in I give it an account for
35:28 14 for 14.99
35:32 it will go out I give it permission for
35:34 14.99 it can scrape up to three social
35:37 media services and for 29.99 it'll do
35:41 you know 10.
35:43 and then I give it my credentials for
35:45 all my social media Services it goes out
35:47 and it crawls my entire [ __ ] history
35:50 and then it runs
35:51 GPT analysis and find stuff you sure you
35:56 wanted to say this
35:58 would you like me to delete that for you
36:01 for another five dollars I can delete it
36:03 in the 37 other places it exists on the
36:05 internet would you like that too
36:09 go start that business you'll be [ __ ]
36:12 Rich give me
36:13 send me a send me a whatever you do on
36:17 Tick Tock here if you do that if you
36:18 make some money
36:22 AI learning lab free business ideas so
36:25 you too can make money bye
36:30 [Music]
36:31 can chat GPT manipulate images it well
36:35 hmm
36:37 code interpreter sure can
36:42 I wonder if code interpreter
36:47 I wonder if it could recolor an image
36:50 want to go play with code interpreter
36:57 all right so this thing on screen so
36:59 this was something code interpreter did
37:01 so I
37:03 uploaded
37:10 [Music]
37:12 this image
37:14 so I uploaded this image which is this
37:16 cool sort of dark Moody fantasy image
37:19 I didn't make it someone from the AI
37:22 Salon did and then I asked it I asked
37:25 code interpreter
37:27 to find the 15
37:30 uh average colors you know top 15
37:33 average colors in the image and and then
37:35 put their relative percentages in a pie
37:38 chart and then make each segment of the
37:40 chart that the color of the color and
37:43 name them a creative name and so it did
37:46 all this it wrote code to go find all
37:48 that [ __ ] which is just mind-blowing
37:51 that it does that it gives itself
37:52 instructions and then goes and writes
37:54 and executes python code to go do it
37:57 options and then goes and writes and
38:00 executes python code to go do it so then
38:02 it named all the colors gave me their
38:04 percentages and then
38:07 it ended up creating
38:11 um
38:11 that chart so
38:14 what I don't know now let's see
38:18 um
38:20 please
38:23 list the colors and their
38:29 um
38:31 hex values
38:41 all right so it's going back I assume
38:44 it's
38:46 there's an unexpected error okay
38:49 oh
38:51 let's see we'll convert the RGB values
38:53 into hexadecimal which is a common way
38:55 of representing
39:00 so here's another thing that code
39:02 interpreter does so the first time I
39:04 asked it for hex values so it went in
39:06 and it saw RGB values and it said well
39:08 that's not good because there are GB
39:10 values so let's change them to hex
39:12 values
39:13 then it said
39:16 we lost
39:17 we lost the library that we installed so
39:20 we're going to do that
39:24 all right so this has all been lost so
39:26 we're going to start over so screw this
39:27 we're going to start over okay
39:30 I want to see if we can do something
39:31 here
39:33 so I'm going to go code interpreter
39:36 and what you do here is there's a little
39:38 upload button next to your prompt now so
39:42 I'm going to go find an image
39:56 oh that's good okay
40:01 so it's a big old
40:03 tractor pull tractor pull tractor
40:07 so I'm going to say please
40:09 find
40:12 the top
40:15 15 average colors in this image
40:22 and name that uh wait let's see uh and
40:26 and yeah name them creatively
40:32 and make a table
40:36 with the names
40:39 percentages
40:41 and hex values
40:47 okay so if you if you're not don't have
40:49 any clue what I'm doing so so I'm in
40:51 chat GPT there's this new thing called
40:54 code interpreter where you can upload
40:55 files
40:58 I asked it to do some crazy [ __ ] I
41:01 uploaded a picture and I asked it to
41:04 give me the top 15 average colors name
41:06 them give me their percentages and their
41:08 hex values
41:12 so it couldn't get it because they were
41:14 RGB
41:16 so
41:18 it's not liking that uh
41:22 but what's cool is every every time it
41:24 goes to do something it writes
41:27 um python code so so it figures out what
41:30 libraries it needs it writes the code it
41:33 tries its thing so it's doing a k-means
41:35 cluster to get the colors
41:38 and then if it messes up it says Oops I
41:41 messed up I apologize for messing up so
41:44 let me write some more python code
41:46 and then it does that and says oops I
41:49 screwed up and now it says great we've
41:50 successfully identified the top 15
41:52 dominant colors in the image
41:55 we've calculated their proportions now
41:57 let's give these colors some creative
41:59 names
42:01 okay Golden Harvest deep Moss Sky
42:04 reflection ocean deep dark Forest
42:11 I'm gonna say can you make
42:16 a horizontal
42:21 bar chart
42:24 with the length
42:29 based on the percentage the color
42:35 of the bar
42:38 should
42:41 be the color
42:43 I don't know if that's right
42:45 wording that's some English words in a
42:49 sentence the color of the bar should be
42:51 the color and the name of the color
42:56 should be inside the bar
43:00 if the if
43:03 the color is
43:08 light the text should be dark
43:14 if the color
43:17 is dark
43:20 the text
43:25 should be light
43:26 let's see if that does anything
43:28 interesting
43:34 so it's writing code what's it doing it
43:37 brought in a
43:38 matplot so it brought in a a plotting
43:42 Library
43:44 it's writing python code to determine
43:46 the length of the bar charts
43:49 and there you go
43:51 and it actually did it look at this Dark
43:53 Forest is in light
43:55 like it made the dark colors have light
43:57 fonts the light colors have dark fonts
43:59 it named them all down here
44:02 here's the relative percentages Okay so
44:06 so now we have
44:08 a color palette from that image so now
44:11 we're going to say
44:14 can you characterize
44:21 this color
44:23 palette
44:28 um
44:30 as if
44:34 you were
44:36 a creative
44:38 writer let's see what it does
44:42 absolutely imagine I don't want that
44:47 sorry I just want a label for
44:54 the kinds of colors here
45:02 I would call it rustic nature Symphony
45:05 great
45:07 so I'm going to say perfect
45:09 great job
45:11 by the way saying thank you and please
45:13 to your AI is is not a bad
45:17 great job it's not a bad thing to do it
45:20 seems to enjoy it and get get better if
45:22 if it knows you're happy with it perfect
45:23 great job
45:26 now I'd like you to name a color palette
45:31 that is very different than this and
45:37 where
45:38 Anything Green is now red
45:43 what
45:45 would be the name of that palette
45:55 Autumn's fiery Serenade
45:58 perfect
46:01 is it
46:02 possible
46:06 for you to show me
46:09 that palette
46:13 keeping
46:14 the same color per
46:19 sentences but
46:22 change out the colors for your new
46:26 palette
46:30 yes apparently so
46:32 now it's going off and writing some more
46:34 python code
46:36 I mean
46:38 I don't know about you
46:40 but like the fact that this is anywhere
46:42 in the neighborhood
46:44 of remotely possible is staggering to me
46:55 so it's creating new RGB colors
47:02 they gave us some new names sun-dried
47:04 tomato lemon zest
47:07 there's the hex values so it did it
47:10 peppermint snow
47:14 Driftwood gray charcoal brat cotton
47:16 candy Sky velvet night
47:20 sun-dried tomato there's the colors
47:25 where's sun-dried tomato in there
47:31 why is that not in there oh there it is
47:34 oh it's kind of brown
47:36 um
47:37 can you make
47:42 a sun
47:44 dried tomato
47:47 more red
48:03 [Music]
48:05 so what did this one do
48:08 I guess that just made that just fixed
48:10 the colors
48:11 and now I guess this next one is writing
48:13 the
48:15 yeah yeah it's making that
48:19 all right well it made sun-dried tomato
48:20 orange now let's try one final thing
48:22 okay
48:23 is it possible to
48:28 recolor the original
48:32 image
48:34 by mapping
48:36 this new palette
48:40 to
48:43 the colors in the image
48:47 it may not because it because it just
48:49 took the top 15.
48:52 but let's see what it does and so so
48:54 apparently I haven't actually done this
48:57 apparently what what code interpreter
48:59 can also do is it can process
49:03 oh it did it holy [ __ ] oh my God
49:09 okay I I'm I'm shooketh
49:13 is anyone paying attention to what's
49:14 going on here let's see open image in a
49:17 new tab
49:19 where did it do it
49:21 did it do it where's my where's my image
49:25 what'd you do with my image chat jpt
49:28 what Dang God dang it
49:32 let's see
49:34 copy image link let's go over to here
49:38 and paste image like let's go look at
49:40 this thing
49:46 come on what what is happening
49:55 is it going
49:59 well anyway
50:01 can I make it bigger there we go so that
50:04 original image
50:06 the tractor was Green
50:08 so now it's done this sort of poster
50:09 posterized thing it made
50:12 holy crap
50:14 so I uploaded an image I found out a
50:17 bunch of stuff about the image and then
50:19 I told it to recolor the image and it
50:21 just recolored the image
50:23 uh you know there you go people
50:27 I don't know
50:28 I don't know what to say
50:31 if you don't find that mind-blowing you
50:33 ain't been paying attention
50:34 have you been paying attention
50:37 whoever said this [ __ ] was over hyped
50:40 it's for good reason
50:42 sometimes
50:44 the hype is there because the shit's
50:47 real
50:48 just because it's over hyped doesn't
50:51 mean it's over hyped it just might be
50:53 appropriately hyped
50:55 what if it's under hyped
51:00 equally plausible
51:04 so yes Chachi BD manipulate images
51:10 um
51:10 so a couple of things
51:13 let me let me I'm gonna
51:15 if you go to bing.com
51:21 so you come to Bing it's just like a
51:23 Google search box but there's this chat
51:25 button so you go chat
51:30 and
51:31 you can
51:33 just ask it to generate an image so you
51:35 could say
51:36 create
51:38 a purple
51:44 tractor
51:46 um racing
51:50 on a dirt road
51:56 so it's gonna go
51:59 do that
52:02 why they use this image as the
52:04 placeholder image for when it's
52:06 generating images I guess it's pixels
52:08 but whatever stupid
52:11 it's a very Microsoft thing to do look
52:13 there's our purple tractor racing down
52:14 the road so it can generate images but
52:17 the other thing Ben can do that chat GPT
52:19 can't do
52:22 which I find kind of cool
52:25 let's go find an image
52:28 let's see
52:30 okay there's a good one
52:33 this is an image from the AI salon
52:40 so I'm going to drag this in here
52:43 so you can drag an image in here and you
52:45 can say what
52:47 is
52:49 the main subject
52:53 of
52:55 this image and what does the text say
53:02 and so it's a zebra and it should say AI
53:05 salon so let's see if it
53:08 if it figures it out I mean
53:20 the image you requested is not available
53:23 yet it got confused let me I'm going to
53:26 create a new topic I'm gonna have it
53:27 forget the other thing it got confused
53:30 with the uh
53:31 with the purple tractor
53:34 what is the main
53:38 subject in this image and what does the
53:44 text say
53:49 this is one of the one of the common
53:52 experiences that you'll have with these
53:54 GPT tools is that sometimes they're just
53:57 weird and do [ __ ] wrong
54:02 the main subject of this image is a
54:04 zebra in a running motion with a main
54:07 and it's tail flying in the Wind
54:12 the background is solid yellow with AI
54:15 Salon written in Black that's not quite
54:18 right
54:19 the artist name alizel Wizzle is written
54:22 in the top left corner no it's in the
54:24 bottom left corner but it got that right
54:26 the illustration is in a comic book
54:28 style unfortunately I couldn't find any
54:30 more information on what the
54:33 topic says so it got that mostly right
54:35 so
54:36 so bing bing can both make images and
54:40 understand images so you could probably
54:42 there's probably a fair amount that you
54:44 could do with that looks like we have
54:46 some new people here welcome everybody
54:47 this is the AI learning lab
54:50 um let me go show you this little
54:53 document or this little
54:55 graphic page whatever the hell it is
54:59 um if you're new to generative AI if
55:01 you've not played with chat gbt go to
55:04 that go there
55:06 go go right now go play with that
55:10 listen to the people that have been here
55:12 that hang out in my lives they went
55:15 there and then they went like this
55:21 and
55:24 so what that is is that's chat gpt's
55:26 official website Bing is what I was just
55:29 showing that's Microsoft's version of
55:31 chat GPT
55:33 if you don't know anything about this at
55:35 all go to prompts.chat that document
55:37 right there
55:39 will teach you about prompting and we'll
55:41 give you a bunch of ideas for what you
55:42 can do with this thing
55:44 um if you haven't played with it just go
55:45 start playing and then if you want to
55:48 feel free to ask any questions here and
55:51 um I'm sorry I just got distracted
55:53 someone said I have a great laugh which
55:54 I appreciate and then they said I have
55:56 great hair which I think is comical
55:57 because actually today it's looking
55:59 pretty good the glue's holding but when
56:02 the when the glue goes it can get it can
56:03 get it can get pretty pretty uh
56:06 disturbing all right let me let me flip
56:08 this back around but thank you I
56:10 appreciate the compliment
56:13 um The Prompt is everything not too much
56:16 not too much too little yeah the prompt
56:19 is everything I mean the the thing about
56:21 chat GPT it's all about
56:24 um
56:26 giving it enough context to understand
56:29 kind of what's inbounds and what's out
56:31 of bounds of what you want
56:35 and if you get better at asking
56:37 questions it turns into a [ __ ]
56:39 superpower hey when you have a minute
56:41 please explain where to upload the image
56:43 for mid-journey to include in a prompt
56:45 okay I'll go show you that right
56:48 here we are in mid Journey
56:50 so if you've never used mid-journey
56:53 then I have to explain two tools to you
56:55 because you can only make mid-journey
56:57 images right now in Discord why is that
56:59 because
57:04 whatever doesn't matter that's where
57:06 they started they were probably nft Bros
57:08 when they started and they're like hey
57:09 we can make images
57:11 let's do it in Discord every time
57:12 everyone's on Discord so anyway if you
57:14 go to midjourney.com It'll point you to
57:16 Discord you get it give yourself a
57:18 Discord account if you don't have one so
57:20 when you're in Discord all these circles
57:22 down the left left are the different
57:24 servers so those are basically different
57:25 communities or different you know
57:27 different servers and then when you're
57:29 in a server when you're in a community
57:30 you've got these are channels and so I'm
57:33 in the AI Salon channel mid-journey has
57:35 one of these where everything's
57:37 mid-journey related but I'm just in the
57:39 mid Journey channel on the AI Salon okay
57:42 so here's some images I I produced on my
57:44 last live and there's some buttons here
57:46 I'll show you all those but the the
57:48 person here Tobias
57:51 um wanted to know how how do you up how
57:53 do you basically use an image as a base
57:56 to create another image so what we're
57:57 going to do is we're going to go find an
57:59 image
58:00 so I'm on a Mac
58:03 um let's see
58:09 yeah that's a good one this is this is
58:10 when I was experimenting with something
58:15 trying to create a persona so
58:20 so what I did was I just dragged that
58:22 image from my finder into the prompt
58:24 window and I'm going to hit return and
58:26 so what that's going to do is it's going
58:27 to upload that image so now that image
58:29 is in there so now what I do is I don't
58:32 know why it says it's processing it but
58:34 whatever I'm going to right click it or
58:36 if you're on a Mac control click it and
58:38 then I'm going to copy the link
58:40 to that image
58:42 and so now I'm going to do I'm gonna
58:44 make an image just like I normally would
58:45 slash imagine
58:50 and then what you do is you paste in the
58:52 URL of that image so basically you have
58:54 to upload it to Discord right click it
58:56 get the URL then type slash imagine as
58:59 you normally would paste in the image
59:01 and then let's see let's say
59:06 um
59:08 woman
59:10 um standing
59:12 on
59:15 a
59:17 rooftop
59:19 overlooking Manhattan
59:24 um
59:26 during sunset let's just see what that
59:30 does
59:30 so that should put her on a roof
59:33 overlooking Manhattan
59:35 during sunset
59:37 or some version of her or or maybe like
59:40 it'll keep the composition and change
59:43 other things we'll see what it does
59:45 so there's at least a woman and if you
59:47 haven't seen mid-journing or any of
59:49 these tools how they make images they
59:51 all start with this blurry sort of noise
59:53 noisy
59:56 um completely nondescript image and then
59:59 it denoises over a series of like 70
1:00:03 steps they don't show you all the steps
1:00:05 but it's like you know somewhere between
1:00:07 30 and 70 steps they're denoising
1:00:09 denoising denoising so it gets more and
1:00:11 more detailed so these images are
1:00:13 actually being generated from
1:00:17 just random pixels and then they get
1:00:19 increasingly more
1:00:22 detailed as it goes it looks like it did
1:00:25 this pretty well
1:00:27 so she's got on kind of the same kind of
1:00:30 top
1:00:32 her hair is Sun lit
1:00:36 yeah this is pretty good actually
1:00:43 like I mean that one is definitely like
1:00:46 her on a rooftop with you know Manhattan
1:00:48 in the background at sunset
1:00:54 and this one you can't quite see it with
1:00:56 the thing but it's got like sun rays
1:00:58 coming out
1:00:59 so anyway that's how you do it
1:01:04 crazy right
1:01:05 cuckoo
1:01:10 [Music]
1:01:18 using topaz AI mask AI can do the
1:01:21 selection background removal tool yeah
1:01:23 there's gonna be lots of tools that do a
1:01:25 lot of the tools now are doing really
1:01:27 sophisticated background removal quite
1:01:29 easily which is good
1:01:31 I use it every day to write social media
1:01:33 posts it gives me more time to create
1:01:35 content yes yes
1:01:37 these tools what they're really good at
1:01:40 is the shitty repetitive Soul crushing
1:01:46 like blocking and tackling work of
1:01:49 knowledge work they do all the shitty
1:01:51 work frees us up to do the higher level
1:01:53 thinky thinky work
1:01:56 if you can do more think you think he
1:01:58 work we can do better [ __ ] for clients
1:02:00 and for ourselves and work may not suck
1:02:03 as much
1:02:06 been playing with chat EBT since January
1:02:08 just got four where I can learn
1:02:11 about plugins and how to use plugins are
1:02:14 weird and
1:02:17 they're somewhere plugins are somewhere
1:02:19 between remarkable and why are these
1:02:22 here what do these do
1:02:24 um
1:02:26 the way I'm thinking about plugins right
1:02:28 now
1:02:29 is like the early days of the iPhone so
1:02:32 when the iPhone first came out
1:02:34 the App Store didn't exist it said
1:02:36 here's this phone and it's got these
1:02:38 things and they didn't call them apps
1:02:40 they called them widgets which was a Mac
1:02:43 holdover
1:02:44 and widgets were actually just it was
1:02:47 just a web browser so when you clicked
1:02:49 on a widget it just took you to a web
1:02:51 page that was sort of an application in
1:02:53 a website
1:02:55 um and it was like that for a year or so
1:02:56 and everybody made fun of them they're
1:02:58 like what's the point of the widgets
1:02:59 which are just websites this is stupid
1:03:01 why don't you just go to a website
1:03:02 Apple's an [ __ ]
1:03:06 they completely redefined you know how
1:03:08 we interact with all information and
1:03:11 people are like but but it's not perfect
1:03:13 so so that's where I think plugins are
1:03:16 right now and even I'm whining about
1:03:18 them and Sam Altman himself said there's
1:03:20 not a product Market fit for widgets and
1:03:22 it's probably because they suck and the
1:03:24 interface isn't there and they're you
1:03:25 know they're they're not there that said
1:03:27 some of them are really good like
1:03:28 there's some that read PDF files and
1:03:30 read
1:03:32 um websites and things like that code
1:03:34 interpreter can replace a lot of the
1:03:36 widgets actually
1:03:38 um
1:03:40 so there but it's worth playing with
1:03:41 because what what widgets represent
1:03:45 is is kind of what I showed earlier with
1:03:47 with code interpreter where
1:03:50 you kind of give it a goal you you give
1:03:52 chat CPT or whatever your agent you give
1:03:56 your virtual agent thank you very much
1:03:57 for the rose
1:03:59 you give your virtual agent a goal
1:04:03 and then it will just figure out okay
1:04:05 which plugins do I need to use which
1:04:07 code do I need to write do I are there
1:04:09 any apis available if there's not I'll
1:04:11 create one I'll go get the data I'll
1:04:13 bring it back
1:04:16 all right
1:04:19 just like a trained monkey you put a
1:04:22 cowboy hat or a heart on my head and I
1:04:23 Just Dance I Just Dance
1:04:27 um
1:04:29 but if you've got access to plugins you
1:04:31 also have access to code interpreter so
1:04:33 not only should you be learning about
1:04:34 plugins but start figuring out what you
1:04:37 can do with code interpreter I don't
1:04:39 have a flipping clue right now because
1:04:41 I've only had it for two days and I know
1:04:44 like I'm still at at parlor tricks stage
1:04:47 with code interpreter like what I just
1:04:49 did with that tractor image like pull
1:04:51 the colors out of it put it in a chart
1:04:53 give them fancy names and now I learned
1:04:55 change the colors and redo that image
1:04:58 and it did it
1:05:00 um
1:05:02 I haven't quite figured out
1:05:06 what it can really do and what it
1:05:10 what I'll get excited about with it but
1:05:12 I know it's big
1:05:16 I know there ain't been nothing ever in
1:05:19 my life that is even remotely close to
1:05:21 what code interpreter is doing right now
1:05:25 I think before I go tonight we're going
1:05:27 to rename code interpreter
1:05:29 and then oh you know what we could do ah
1:05:31 okay
1:05:32 there's 99 of us here if we if we can
1:05:34 keep a decent amount of people here
1:05:36 um tap tap the heart button maybe we can
1:05:38 get some more people here so what we're
1:05:39 gonna do is we'll we'll go to chat GPT
1:05:44 we'll rename code interpreter we'll come
1:05:46 up with some some clever name for it and
1:05:50 then we can all go go to Twitter or
1:05:52 something and tag open Ai and tell them
1:05:56 tell them we we all approve of the new
1:05:58 name that we think they should rename
1:05:59 code interpreter because it's a shitty
1:06:02 name and we don't care if it's accurate
1:06:04 all right maybe we'll do that later
1:06:06 that'll be fun
1:06:12 I don't know we did some kind of Anarchy
1:06:17 there's this old dude on Tick Tock and
1:06:20 he was a little feisty
1:06:21 we renamed something and we tag Sam
1:06:24 Altman
1:06:26 I feel like a rebel
1:06:29 what are you talking about Marge I'm
1:06:32 talking about AI it's going to change
1:06:33 the world Sally
1:06:40 Marge you've been drinking again
1:06:43 you know I don't do that
1:06:46 I've used it to create a lot of good
1:06:48 synthwear style software scenery oh
1:06:50 that's cool yeah I bet it's good at that
1:06:51 that's that's sweet
1:06:54 what's this code interpreter of gpt4 so
1:06:57 I so I assume I'm behind on my comments
1:07:00 here
1:07:01 um
1:07:02 you can I'll go show it again
1:07:05 and and that's what we're talking about
1:07:07 we're gonna have to rename it because
1:07:08 it's a horrible name
1:07:10 um but code interpreter so so chat GPT
1:07:13 you can go in and you can type [ __ ] and
1:07:14 then it
1:07:15 does amazing things right make me a poem
1:07:17 write me a song write me a project plan
1:07:21 compare this philosophy with that that
1:07:23 philosophy rightly python code it does
1:07:25 all the [ __ ] right it does all this deep
1:07:27 deep amazing stuff
1:07:29 code interpreter
1:07:32 allows you to upload a data file up to
1:07:35 100 megabytes
1:07:36 so the data file can be it it seems like
1:07:39 almost anything you can upload a zip
1:07:41 file that like a compressed single zip
1:07:44 file that has a hundred files in it and
1:07:46 it will unzip the file and then read all
1:07:49 the the 100 files within it which I
1:07:52 didn't think it could do I thought you
1:07:53 could just uh upload like one
1:07:55 spreadsheet and it could do that no no
1:07:57 you can upload a zip file full of a
1:07:59 bunch of other files and it will read
1:08:01 all of them it will understand all of
1:08:03 them
1:08:04 and then you can tell it
1:08:07 to do [ __ ] with that data or with that
1:08:10 file
1:08:11 so you can say hey what's what are the
1:08:14 interesting Trends in this data and
1:08:16 visualize them for me and then it gives
1:08:20 itself a set of instructions to meet
1:08:23 your goals and it starts writing python
1:08:25 code and executing it and like if you
1:08:28 upload a zip file the first thing it
1:08:30 does is it writes python code to unzip
1:08:33 the zip file and look at what's in it
1:08:36 and then it writes more code to look at
1:08:37 the [ __ ] that's in it and categorize it
1:08:39 and coalesced it into a single file that
1:08:42 it can deal with and then it makes
1:08:44 charts and graphs and like it's just
1:08:46 Insanity it's it's just [ __ ] bonkers
1:08:50 and this is like version like this is
1:08:52 like it's beta right like this is just
1:08:55 out of alpha this is this is like it it
1:08:57 only does python
1:09:00 so like you can imagine a future version
1:09:02 of this will happen instantly and it
1:09:05 will write all the languages and you'll
1:09:07 just give it a goal and it's going to go
1:09:09 do a bunch of [ __ ]
1:09:12 that's what it is dance Rosie I know
1:09:17 I'll dance you put a cowboy hat on me
1:09:20 I talk like a cliche cowboy
1:09:24 just happens
1:09:29 you can take the actor off the stage but
1:09:31 you can't take the stage out of the
1:09:33 actor
1:09:34 you know what I'm saying
1:09:37 you know what I'm saying cow poke
1:09:40 sorry I lost my place in the comments
1:09:42 again
1:09:43 this happens
1:09:45 old guy trying to use tick tock he's an
1:09:48 old guy trying to use tick tock his hair
1:09:51 is receding he's
1:09:53 nose is hairy
1:09:55 he's got something wrong with his neck
1:10:03 okay here we go Vector journey I
1:10:05 absolutely love Bing oh I went back too
1:10:07 far hold
1:10:11 oh I went back way too far
1:10:14 dang it
1:10:19 um
1:10:20 what'd you do tonight I don't know
1:10:22 that's some old guy trying to find his
1:10:24 way on tick tock
1:10:27 all right there was the please explain
1:10:29 how to do something in mid-journey I did
1:10:31 that they should have asked chat GPT to
1:10:33 name code interpreter we're gonna do
1:10:35 that tonight and we're we're gonna we're
1:10:38 gonna Twitter Bomb Sam Altman and open
1:10:41 AI we're gonna rename it tonight it's
1:10:43 it's it's the worst [ __ ] name
1:10:46 um okay here we are no date yet on the
1:10:49 musk Zuck cage match not yet but I think
1:10:52 they've they've picked they picked a
1:10:55 venue in Vegas
1:10:57 and and Dana White from from UFC said if
1:11:00 they do it it's gonna be the biggest
1:11:01 fight in history like the biggest purse
1:11:04 in history so they could they could end
1:11:06 up raising a [ __ ] ton of money for
1:11:08 charity which is cool like great go
1:11:11 punch the [ __ ] out of each other I don't
1:11:12 care [ __ ] idiots
1:11:17 um what's the best place to learn
1:11:18 plugins and zapier uh plugins
1:11:23 I don't know just start playing I'll
1:11:25 show you a couple of them here I've been
1:11:27 on chat GPT since January just subbed
1:11:30 into four
1:11:32 um
1:11:33 zapier is a different thing so if if you
1:11:36 if you just got into chat gbt
1:11:40 um the gpt4 thing you should probably
1:11:42 also get into if you want to do zapier
1:11:45 you should get into
1:11:47 um the playground and pay for API access
1:11:51 um which it sounds like it's double
1:11:53 dipping but it's just it's it's a
1:11:55 different thing and and the the API
1:11:57 access is not that expensive if you're
1:11:59 just developing small scale stuff
1:12:02 um but let me let me go show you I'll go
1:12:04 show you a couple of things here
1:12:12 okay so for plugins so I assume you I
1:12:15 assume you turn them on in settings but
1:12:17 for those of you
1:12:19 um who may be new to chat GPT plus if
1:12:21 you want gpt4 plugins and code
1:12:24 interpreter you have to pay for GPT plus
1:12:27 if you do that and you don't see them
1:12:29 you have to enable them in your settings
1:12:32 this is a non-obvious thing I don't know
1:12:34 why they did this I don't like it but
1:12:36 whatever
1:12:37 so you you come into your settings you
1:12:39 go to Beta features and then you have to
1:12:41 turn on access to plugins and code
1:12:43 interpreter and then when they bring
1:12:45 back browse with Bing you'll have to
1:12:46 turn that on as well
1:12:48 okay
1:12:50 so I assume you've used plugins but if
1:12:53 you haven't
1:12:55 um the the way they work is in any given
1:12:58 chat session you can enable up to three
1:13:00 of them at the same time
1:13:02 so here's
1:13:04 um
1:13:05 instacart let's see these are ones I've
1:13:08 got available to me that I installed
1:13:12 and then in this plug-in session I can
1:13:16 you know include up to three so let me
1:13:18 do I'll do get your guide
1:13:22 um
1:13:24 and trip.com we'll do those too so I'll
1:13:28 say
1:13:29 um
1:13:30 make me
1:13:32 a plan for my trip
1:13:36 to Toledo
1:13:39 Ohio
1:13:41 and September
1:13:46 um I want
1:13:48 to have
1:13:50 fun during the day and I like turtles
1:13:55 [Laughter]
1:13:58 so so I've been able to plug-ins that
1:14:03 are related what exact date in September
1:14:06 how many days do you plan to stay oh
1:14:08 let's say
1:14:10 September 14th
1:14:15 for three days
1:14:19 from Denver
1:14:22 so that's probably going to go book a
1:14:24 trip on trip.com is my guess
1:14:29 and then we'll see
1:14:33 great
1:14:34 here's a three-day itinerary
1:14:38 visit the collegial Cathedral of Toledo
1:14:42 explore the Alcazar historic Stone
1:14:44 formation
1:14:46 stay at the
1:14:48 Palacio Buena Vista
1:14:55 it's sending me to a bunch of religious
1:14:57 places
1:14:58 that's weird
1:15:05 a museum dedicated to El Greco
1:15:08 the synagogue wait
1:15:11 I'm sorry I'm going to stop it and stop
1:15:12 it uh this is weird
1:15:16 um they are all
1:15:19 religious
1:15:21 activities and I like turtles
1:15:26 what about turtles
1:15:33 Turtles
1:15:34 so now it's using get your guide so
1:15:36 maybe it realized uh it needed some I
1:15:39 found some activities that involve
1:15:41 Turtles
1:15:45 oh my God wait Cruz the Strait of
1:15:48 Gibraltar
1:15:55 I'm gonna stop it again
1:15:58 no it has to be in Ohio
1:16:03 isn't there a zoo question mark
1:16:07 all right so anyway so that's how you
1:16:08 use so so plug-ins and and actually what
1:16:12 I'm experiencing right now a lot of the
1:16:14 plugins
1:16:15 like are are this kind of frustrating
1:16:18 where where they seem like they should
1:16:20 be better than they are some of them are
1:16:21 really good some are bad
1:16:23 so here's a river cruise where I could
1:16:25 see a turtle here's a sightseeing Cruise
1:16:27 as for the Toledo Zoo it's one of the
1:16:29 top rated zoos that's what we're talking
1:16:31 about
1:16:33 and there's my little river boat all
1:16:35 right so I can go see some turtles there
1:16:36 so anyway that's how you use those
1:16:38 things
1:16:39 um what was the question
1:16:41 where's the best place to learn plugins
1:16:44 oh and zapier okay
1:16:45 so so for the plugins probably YouTube
1:16:49 um or go to
1:16:50 um
1:16:51 go to Reddit and and just just hang out
1:16:54 in like the chat GPT forums and Reddit
1:16:57 and search for plugins and
1:17:00 there's a there's a lot of plugins the
1:17:02 the other thing about plugins
1:17:04 um if if you haven't played with them is
1:17:07 how you add new ones is you go to this
1:17:09 list that you have to enable them and
1:17:10 there's a plug-in store
1:17:13 and then
1:17:16 we'll see how many there are now last
1:17:18 time I checked there were like 620.
1:17:20 so I'll go to all
1:17:22 there's 163 pages of four plug-ins each
1:17:25 so whatever that math is that's how many
1:17:27 plugins there are so there's a lot of
1:17:29 them like I said a lot of them are crap
1:17:33 um a lot of them are just experiments
1:17:35 where people are experimenting how to do
1:17:36 plug-ins and then there's some
1:17:38 occasional good ones
1:17:41 um let's see I'll tell you the ones that
1:17:42 I have installed that I like
1:17:46 uh uh ask your PDF is good although with
1:17:49 code interpreter that's completely
1:17:51 irrelevant now
1:17:53 um
1:17:54 diagrams not bad chat with videos not
1:17:57 bad but again I think I think code
1:18:00 interpreter is going to make a lot of
1:18:02 these obsolete
1:18:03 instacart's kind of fun you can have it
1:18:06 you know give you a meal plan and then
1:18:08 give you a shopping list for the meal
1:18:10 plan and have it delivered to your door
1:18:12 kayak is not bad
1:18:15 um
1:18:16 ml paper reader is machine learning
1:18:19 papers like scientific papers OpenTable
1:18:21 you can book restaurants why you would
1:18:24 do that here and not through
1:18:26 open tables Beyond me plug-in pedia is a
1:18:29 a thing that'll tell you about plugins
1:18:31 prompt perfect is actually kind of fun
1:18:33 you type the word perfect before prompt
1:18:35 and it'll rewrite it
1:18:37 show me diagrams is good stories is good
1:18:42 um web pilot I use a lot but again I
1:18:44 have a feeling I'm gonna I'm gonna be
1:18:47 using Code interpreter more than
1:18:49 stuff connected to the internet and yeah
1:18:52 oh and then there's a zapier one where
1:18:54 you can sort of describe an Automation
1:18:56 and it'll go make it for you I don't
1:18:58 really use that one though because so
1:19:00 anyway so that's that okay
1:19:01 so now if you want to do zapier so one
1:19:04 thing that you're going to want to do is
1:19:05 you're going to want to get on the
1:19:07 playground so it's the open AI
1:19:09 playground so you go to
1:19:13 platform.openai.com playground
1:19:16 and that'll bring you here and it's it's
1:19:20 sort of like uh chat GPT except you have
1:19:23 to give it a system message like you are
1:19:25 an expert
1:19:27 smart ass
1:19:30 and no matter what someone says
1:19:35 to you
1:19:37 you can answer them
1:19:41 but
1:19:43 do it in a snarky
1:19:47 and
1:19:49 cynical sin
1:19:51 send me a call but Hume or us way you
1:19:57 don't want to hurt
1:20:01 anyone but
1:20:04 just make them laugh
1:20:09 it may not enjoy that so I'm going to
1:20:12 say
1:20:12 um
1:20:13 hello there
1:20:16 um what
1:20:19 um is the best way to get
1:20:23 to Carnegie
1:20:27 Hall
1:20:31 the answers practice
1:20:35 okay
1:20:39 submit
1:20:42 ah the famous Carnegie Hall the best way
1:20:44 to get there well I suppose you could
1:20:45 try hitchhiking on the unicorn's back
1:20:47 okay so it's bad but anyway so so this
1:20:50 is where you can prop practice your
1:20:52 prompt writing and then you've also got
1:20:54 all the different settings and different
1:20:57 models right so this is 3.5 turbo
1:20:59 there's gpt4 in here there's the chat
1:21:03 um mode there's the complete mode which
1:21:06 they're deprecating they're getting rid
1:21:07 of that most of the stuff I've written
1:21:09 is incomplete so hopefully they don't
1:21:12 kill it too soon
1:21:14 or my shits couldn't break
1:21:16 um so anyway but this this is where you
1:21:18 go to sort of
1:21:19 um get your prompts dialed in and then
1:21:21 when you go to zapier
1:21:24 oh the other thing you do at the
1:21:25 playground is and this is important
1:21:30 um under your account there's view API
1:21:34 keys this is where you get your API key
1:21:36 to put into zapier
1:21:38 so here's my my API Keys you create a
1:21:41 new secret API key it gives you the key
1:21:43 and that's what you're going to put into
1:21:44 zapier
1:21:45 so then in zapier
1:21:49 I just did a uh I just did a workshop on
1:21:52 this for for a couple of guys I work
1:21:53 with
1:21:56 so I've got a good example here that's
1:21:58 nice and simple
1:22:00 I didn't rename it it's still called
1:22:02 Untitled zap because I don't have any
1:22:04 creativity in me let's call it
1:22:07 um
1:22:08 simple example
1:22:11 zap that's good okay
1:22:14 so this is kind of the the the most
1:22:17 simple thing you can do in in zapier as
1:22:20 an automation is there's kind of three
1:22:22 three there's three modes there's
1:22:25 get some data in there's
1:22:28 process that data and then there's do
1:22:31 something with what you processed and so
1:22:33 in this case I'm pulling in data from a
1:22:36 Google form
1:22:37 I've got a prompt here that's going to
1:22:40 go to open AI that uses the data from
1:22:42 that form
1:22:44 and then I'm going to take the output of
1:22:47 what chat GPT gives me and send that to
1:22:50 some to that person via email so so
1:22:54 there's three components so so what
1:22:56 zapier is is it's kind of these Lego
1:22:58 blocks of different
1:23:00 services
1:23:01 so so basically when you set up you know
1:23:04 like your Google form you say okay when
1:23:06 someone gives you a new form response
1:23:09 from this account
1:23:13 from this form
1:23:15 you know here's the data that they're
1:23:17 going to input
1:23:18 and then you're going to go to the next
1:23:20 thing
1:23:21 so I'm gonna send a prompt to open AI
1:23:24 from this account
1:23:27 using this prompt right that
1:23:30 incorporates you know the the answer
1:23:32 from that Google form or you know
1:23:34 multiple fields from the Google form
1:23:36 and then I'm gonna get I think in this
1:23:39 case I got back a poem yeah so I'm gonna
1:23:41 get back a poem
1:23:43 and then I'm going to take that poem and
1:23:45 I'm going to stick it in I'm going to
1:23:47 send an email from Gmail
1:23:50 from this account
1:23:52 with you know from this person with this
1:23:55 subject line where I Incorporated you
1:23:57 know the thing from the form as well and
1:24:00 here's the body and then you know it's
1:24:02 going to send an email so so that's
1:24:04 that's what that's a really simple
1:24:06 automation
1:24:08 um but that's basically what it is so so
1:24:10 you'll you'll need your accounts in your
1:24:13 different Services you'll need your API
1:24:15 key if you're doing anything with open
1:24:18 Ai and then you're off to the races so
1:24:22 I know there's a lot holy [ __ ] there's
1:24:23 still a hundred people here I figured
1:24:25 there'd be like three people left it's
1:24:26 like is he just gonna point to his
1:24:28 screen for the next hour this is awful
1:24:32 what is this channel
1:24:35 it is what it is man you know man you
1:24:38 know you know sometimes man it's just
1:24:40 like I get to you know you know man
1:24:43 sometimes I point at the screen man
1:24:45 that's how it goes man you know man
1:24:54 um
1:24:56 I just thought chat GPT doesn't have
1:24:58 real-time browsing capabilities
1:25:00 it it doesn't but an application layer
1:25:04 on top of it can right so if you go to
1:25:06 bing.com
1:25:08 and you go to chat
1:25:09 that's
1:25:11 gpt4 connected to the internet
1:25:14 so the core large language model itself
1:25:17 so when you're searching when you're
1:25:19 chat gbting and you say write me a poem
1:25:21 or this or that the core large language
1:25:24 model is not connected to the internet
1:25:26 it's doing all that language processing
1:25:27 down there
1:25:28 but they've written it in such a way
1:25:32 that it can understand what you're
1:25:33 asking for and it says if I if they're
1:25:37 asking for something that's more recent
1:25:39 than what I have in my core model go do
1:25:41 a search on the internet and then
1:25:42 incorporate that response into whatever
1:25:45 they asked me to do or where so it sort
1:25:47 of combines the large language model
1:25:49 with the internet search and provides
1:25:51 you a coherent answer
1:25:53 so it makes it seem like it is connected
1:25:55 to the internet but it's actually two
1:25:56 operations that are kind of being munged
1:25:58 into one thing
1:26:03 um extortion via chat gbt I think that's
1:26:05 going to be a major business I think I
1:26:09 think the mob is probably all excited
1:26:11 about chat gbt right now I'd pay 40
1:26:14 bucks for that service for which service
1:26:19 uh oh the scraping the one where I said
1:26:23 yeah yeah four yeah 40 bucks for um
1:26:28 here's here's all my social media
1:26:29 accounts go scrape them all and
1:26:33 bring back to me anything that you think
1:26:36 might be compromising or embarrassing
1:26:39 and then I can just go keep it kill it
1:26:43 keep kill keep kill keep kill keep kill
1:26:46 oh God I can't believe I said that kill
1:26:50 off it goes
1:26:53 and for another 1995 I'll go find all
1:26:56 the places on the internet that
1:26:59 that exists Beyond
1:27:01 where you posted it yourself you [ __ ]
1:27:04 [ __ ]
1:27:05 [Laughter]
1:27:08 I like code interpreter because it keeps
1:27:11 trying methods that don't work right
1:27:13 away yeah respect exactly yeah it does
1:27:16 what's cool about code interpreter is
1:27:18 it's got no shame right it's just like
1:27:20 yeah I gave it I gave it my shot
1:27:22 apologies for that not working you know
1:27:24 what we could try
1:27:27 yeah that failed too hey but you know we
1:27:30 could try
1:27:31 it's amazing is code interpreter going
1:27:34 to be available for ILS mobile version
1:27:37 it might be now is it not I don't know
1:27:39 you might have to turn it on so just
1:27:42 like you have to turn it on in your beta
1:27:44 features here
1:27:47 I would assume it is and then even if
1:27:49 it's not like on mobile I just use a
1:27:53 browser I just use the chat
1:27:56 gpchat.openai.com in a browser
1:27:59 because I think it's I think in a lot of
1:28:01 ways it's easier to use than their app
1:28:03 because their app doesn't have parity
1:28:05 with the web app which I find stupid but
1:28:08 they also named this remarkable thing
1:28:10 code interpreter
1:28:12 let's let's see Team
1:28:15 we want to come up with a name for this
1:28:17 new feature that's going to
1:28:18 revolutionize all of humanity
1:28:21 well it's technically interpreting code
1:28:24 what if we called it code interpreter
1:28:26 I like it it's simple it's direct no one
1:28:29 will know what the [ __ ] it means I like
1:28:31 let's go with that code interpreter it
1:28:34 is
1:28:36 I mean seriously did they have a [ __ ]
1:28:38 meeting
1:28:41 they know that a hundred million people
1:28:43 started using chat EBT that were not
1:28:45 Engineers right they know that right
1:28:47 like someone there knows that right
1:28:50 code interpreter
1:29:02 please don't be recording this and send
1:29:05 it to Sam Altman I don't want to be on
1:29:06 some [ __ ] podcast two weeks from now
1:29:08 sorry Sam I just got carried away I was
1:29:12 kind of excited I was on live people
1:29:13 were hitting the heart button
1:29:16 [Laughter]
1:29:18 say say
1:29:20 it's your fault it's not my fault
1:29:23 I take no responsibility here none
1:29:31 oh good code interpreter names Dream
1:29:34 Craft dream code dreamify code spark
1:29:38 idea engine
1:29:40 code cave recovery
1:29:47 maybe something around agent
1:29:51 magic agent no
1:29:55 I don't know
1:29:56 I don't know keep those cards and
1:29:58 letters coming keep the ideas coming
1:30:00 we'll get there
1:30:01 we're gonna rename this [ __ ] that's
1:30:03 good to give you a downloadable file of
1:30:05 the analysis including text and charts
1:30:07 oh that's cool idea okay I'll do that
1:30:09 just watched a good 60 Minute show on AI
1:30:11 nice
1:30:14 um can you just ask it to scramble the
1:30:16 colors and recreate the photo that's
1:30:18 what I did
1:30:19 oh yeah I mean like oh oh just oh and
1:30:21 recreate the photo probably let me let
1:30:24 me try that
1:30:28 all right we're gonna put a photo in
1:30:30 here
1:30:34 200
1:30:35 completely randomize them or do we want
1:30:38 to cycle them
1:30:40 let's cycle them
1:30:45 some degree around the color wheel
1:30:50 we'll do this one because it's a good
1:30:52 one
1:30:56 um
1:30:58 recolor this image
1:31:01 by
1:31:03 increasing
1:31:06 the luminance
1:31:11 loom in ants and cycling
1:31:15 the colors
1:31:18 um
1:31:21 87.5 degrees
1:31:26 on the color wheel I don't know if
1:31:29 that's possible but I assume it is
1:31:32 like one of those mat this is one of
1:31:34 those things where like
1:31:35 having some confidence in math probably
1:31:38 would be a good thing but now you don't
1:31:40 have to
1:31:42 but look at this it does so look
1:31:45 it said okay load the image convert it
1:31:47 to hsl Hue saturation lightness so it
1:31:51 can increase the luminance at 87.5
1:31:53 degrees to the Hue convert the image
1:31:56 back to RGB save it and display the new
1:31:58 image now let's go ahead and carry out
1:32:00 these steps
1:32:02 so it wrote all our python code
1:32:06 brought in numpy color CIS
1:32:10 image
1:32:15 and now it's thinking
1:32:19 it's a big image so what it might end up
1:32:21 doing is saying I'm sorry we can't do
1:32:24 that but we could do it if we down
1:32:26 sample the image that's happened with
1:32:28 this image before it's a big image it
1:32:30 doesn't like big images
1:32:34 all right
1:32:37 um the youngins will think nothing of it
1:32:39 as it was always here I know that's the
1:32:41 way it is though that's that's just
1:32:42 tough on us you know we we get to say in
1:32:46 my day you actually had to do your own
1:32:48 python coding what's python Grandpa shut
1:32:52 up
1:32:54 get off my porch
1:32:58 um the image has been successfully
1:32:59 recolored here's the output
1:33:02 um I
1:33:03 see no link
1:33:06 or no image can you please
1:33:12 give it to me oh I guess I should
1:33:15 flip these things so you're not
1:33:29 wow it did it
1:33:32 that's [ __ ] insane
1:33:35 so there's that same image
1:33:39 I don't know that it's
1:33:42 um
1:33:43 luminance is is that much brighter but
1:33:47 maybe let's go look
1:33:53 yeah I guess it is
1:33:55 yeah
1:33:57 no not really I don't think the
1:34:00 luminance is brighter
1:34:02 maybe a little bit but it definitely
1:34:05 cycled
1:34:06 cycled the colors
1:34:08 that's just insanity
1:34:11 it's
1:34:12 just insanity
1:34:16 wow
1:34:17 wow
1:34:20 I I got I got nothing
1:34:25 I got nothing I can't I
1:34:30 I don't know how to process what this
1:34:33 does but it does
1:34:38 the configuration of your screen shows
1:34:40 different colors than the configuration
1:34:42 of the robot you have to re-calibrate
1:34:44 your screen for the AI robot I don't
1:34:46 know what that means but okay I think it
1:34:48 needs to be trained on what colors are
1:34:50 names are usually often my attempts
1:34:54 um it I did it I did it one time and the
1:34:57 colors were nowhere close
1:34:59 and then what did I do that made it
1:35:01 better
1:35:02 oh
1:35:07 if you haven't described the colors
1:35:10 first and then give it creative names it
1:35:12 did much better so like if it knows a
1:35:14 color is green
1:35:16 then when it does creative names it
1:35:17 tends to get them right
1:35:20 um but yeah I had one attempt for the
1:35:22 colors were all screwed up
1:35:25 try printing that stuff I bet they come
1:35:27 out looking different everything needs
1:35:28 to be calibrated oh yeah I know that
1:35:30 well RGB images and CMYK images are
1:35:32 nowhere in the same neighborhood so all
1:35:34 that [ __ ] was super RGB heavy so those
1:35:36 are going to print like crap totally
1:35:38 agree yeah but whatever this this is
1:35:40 just
1:35:41 this is just me trying to get my head
1:35:43 around what code interpreter is and why
1:35:45 it has such a shitty horrible name
1:35:48 um you can ask it to give you a
1:35:49 downloadable file of your image we just
1:35:50 did that that was super cool
1:35:52 code interpreter name holy [ __ ] okay
1:35:55 that's good we're getting there we're
1:35:56 getting there
1:35:58 oh well listen there I Can't Believe
1:36:01 It's Not Butter remember that all you
1:36:02 gen xers remember I can't believe it's
1:36:04 not better butter does that product
1:36:05 still [ __ ] exist
1:36:07 the commercials
1:36:10 I Can't Believe It's Not buttered try I
1:36:12 Can't Believe It's Not Butter today
1:36:18 yeah what if they called it
1:36:20 holy [ __ ] it does that
1:36:24 wait it does what holy [ __ ] yeah there's
1:36:27 something there's something there
1:36:33 UV bended a spoon Mr Geller Yuri Geller
1:36:36 nice it's a good Gen X reference
1:36:39 reminding me of the fly where the
1:36:41 computer needed to understand the flesh
1:36:43 pretty much that's that you know
1:36:49 it's not this similar than that I I mean
1:36:52 and that and code interpreters kind of
1:36:54 like uploading the Flesh and quite
1:36:56 frankly like 100 megabyte file I bet if
1:36:59 you if you if you did some sort of Flesh
1:37:02 sampling where where you know the uh the
1:37:07 machine and the sensors you know read
1:37:09 the components and the DNA strands and
1:37:12 this and the molecular structure I bet
1:37:14 that could all go into a data file and
1:37:16 then that goes into code interpreter and
1:37:19 you can sort of cycle six DNA strands to
1:37:22 the right
1:37:23 and out comes a frog
1:37:27 I mean
1:37:30 we're not
1:37:31 we're conceptually in in that
1:37:34 neighborhood it's amazing
1:37:36 awesome I'm in love with AI I know I
1:37:39 know me too me too
1:37:41 it's like
1:37:45 someone said my new BFF is
1:37:47 is a code interpreter
1:37:50 [Applause]
1:37:54 definitely under hyped open AI hire this
1:37:56 man yeah yeah just hire me I'm gonna be
1:38:00 your product neighbor I'm actually not
1:38:01 going to be your product neighbor I'm
1:38:03 just gonna be like the this the smarmy
1:38:05 Gen X are sitting in the corner someone
1:38:07 goes well uh what if what if we called
1:38:10 it a code interpreter nope no
1:38:16 no get them out of here
1:38:18 I'll be that guy
1:38:24 what if we call it holy [ __ ] okay now
1:38:26 we're getting there that's close
1:38:28 [Laughter]
1:38:32 chat no longer talking to people in a
1:38:34 room
1:38:36 um
1:38:37 yeah you know what I the fact that these
1:38:40 are called chat Bots I I think that's
1:38:42 another [ __ ] blunder another naming
1:38:45 blender because chat Bots for the past
1:38:47 decade have sucked ass because they've
1:38:51 just been logic trees right and so you
1:38:54 know and like there's whole companies
1:38:56 built around this I feel horrible for
1:38:57 them like the companies built around
1:38:59 building all these chat Bots where
1:39:01 they're they're writing all the logic
1:39:02 trees
1:39:03 that the problem with chat Bots
1:39:05 historically is you had to write all the
1:39:06 logic and you could never anticipate
1:39:08 what humans would actually do so the
1:39:10 Bots just tended to suck
1:39:12 and this new thing while it's
1:39:14 technically a chat bot it's it's not it
1:39:16 is the experience is nothing like a
1:39:19 [ __ ] chatbot so I agree with you I
1:39:21 think chatting is like people talking
1:39:23 and this is this is something
1:39:25 fundamentally different but again I
1:39:27 don't have uh
1:39:29 anything better right now so tractor
1:39:31 racing on a dirt road is cool thank you
1:39:34 that's good uh
1:39:36 uh it sees a tail it it must have meant
1:39:39 Maine it just made up the tail thing I
1:39:41 know there's no tail I know for the for
1:39:44 the zebra thing I was looking at
1:39:46 but it did it get the other stuff okay
1:39:48 like like it pulled some of the stuff
1:39:50 from the name of the file which I
1:39:52 thought was interesting
1:39:54 because it when it didn't search it
1:39:56 searched on the file the the names the
1:39:58 person's name was in the file name
1:40:00 best place to learn how to use different
1:40:02 AI programs in depth
1:40:05 is it all just learning prompts
1:40:07 um
1:40:11 it is and it's not
1:40:13 um
1:40:14 I think to start
1:40:17 so so I think to start it's important to
1:40:19 understand
1:40:21 that
1:40:24 the real breakthrough here
1:40:26 is that you're using the English
1:40:29 language or whatever your native
1:40:31 language is you're using the English
1:40:32 language
1:40:34 as a programming language
1:40:37 and so you have to think critically and
1:40:40 you have to think
1:40:42 um
1:40:43 creatively
1:40:46 and in some weird way you have to kind
1:40:48 of think empathetically like if you were
1:40:50 telling
1:40:52 a real simpleton
1:40:54 like let's say you got an intern that
1:40:56 knew nothing about your business
1:40:58 but was super smart
1:41:00 like Savant smart like
1:41:05 like Ace is the bar exam and the biology
1:41:07 test which chat CPT does right like aces
1:41:11 all the tests but knows nothing like
1:41:13 really just a adult a really simpleton
1:41:16 how would you talk to that person well
1:41:18 you'd say okay what we're gonna do today
1:41:21 is I want you to think about all the
1:41:24 stuff you know about law I want you to
1:41:26 be a lawyer and he's like okay
1:41:28 and then I've got this problem and and
1:41:32 my problem is that the this person is
1:41:35 suing me and here is the document that
1:41:38 they sent me saying they're suing me
1:41:40 okay got it now I want you to read that
1:41:43 document
1:41:44 and then I want you to take all your
1:41:45 knowledge of all the law and I want you
1:41:47 to tell me what you think we should do
1:41:49 okay here you go all right here it is
1:41:53 um
1:41:55 so so everything whether it's
1:41:57 mid-journey for images or audio
1:42:00 generation or you know some of the
1:42:04 animation tools or chat GPT
1:42:07 are all using some flavor of this
1:42:09 language model like stacked up with
1:42:11 other Technologies so understanding how
1:42:15 to interact with chat GPT is just good
1:42:18 foundational knowledge now I think
1:42:20 moving forward
1:42:22 the requirement of doing complicated
1:42:24 prompting is going to get smaller and
1:42:26 smaller and smaller if you're in Tech
1:42:28 development you're probably going to
1:42:30 need to have these skills but just as
1:42:31 human beings
1:42:33 you may not need to so for example in
1:42:35 mid-journey I showed earlier you can
1:42:37 upload an image to mid-journey and just
1:42:40 give it a really simple prompt and or or
1:42:42 no prompt at all
1:42:45 and let it sort of interpret that image
1:42:47 and make a new one so so there's ways
1:42:49 that you can use this generative AI
1:42:51 without doing prompting or minimal
1:42:54 prompting
1:42:56 um but it's a good thing to learn so the
1:42:57 the
1:42:59 that URL right there prompts.chat is a
1:43:02 document that teaches you about
1:43:04 prompting it teaches you what it is why
1:43:07 it works how these large language models
1:43:09 work and then it gives you a bunch of
1:43:11 examples that you can put in chat GPT to
1:43:15 tell it to be the lawyer to tell it to
1:43:17 be a bunch of different things so
1:43:24 uh what's the difference between these
1:43:27 you sharing I don't know what I was
1:43:29 sharing so
1:43:30 have I tried pie I tried it a little bit
1:43:33 but it's it's on my list of [ __ ] to try
1:43:37 I mean my list right now is subsumed
1:43:40 with figure out what the [ __ ] to do with
1:43:42 code interpreter and give it a new name
1:43:45 the glue is holding I know right
1:43:48 Bam Bam Bam
1:43:52 pretty funny
1:43:54 why you not talk about Bard because Bard
1:43:57 sucks I Bard will go on this list when
1:44:01 it doesn't
1:44:02 have you used it now listen
1:44:05 I hear it's much better at coding
1:44:08 um than it is at language but I mostly
1:44:11 use this [ __ ] for language so if it's
1:44:14 great at coding for you and you're
1:44:15 getting value out of it great do I think
1:44:17 Bard is going to be good at some point I
1:44:19 do but I think Google has has kind of
1:44:22 ignored
1:44:24 making it good for a long long time
1:44:26 because it threatens their search engine
1:44:29 model and because of that chat EBT is
1:44:32 way ahead of them like no one's close to
1:44:34 gpt4 right now that I've experienced
1:44:37 so it's just not there like all of these
1:44:41 all three of those options all have
1:44:44 access to gpt4 pose got access to some
1:44:48 of the
1:44:49 um anthropic models the cloud models
1:44:51 which I think are actually pretty strong
1:44:53 but those are open AI people
1:44:55 so
1:44:56 so I just don't have it here because it
1:44:58 sucks but there's there's like there's a
1:45:00 Google technology called style drop
1:45:04 which is mind-blowing
1:45:18 if you haven't seen this this is just
1:45:20 insane so this is something from Google
1:45:23 so so
1:45:24 I'm not hating on Google Google I think
1:45:27 Google's really really far behind but
1:45:29 Google invented this [ __ ] they invented
1:45:32 the Transformer so so the core
1:45:33 technology that all of the stuff that
1:45:35 we're playing with is built on
1:45:37 Google invented so this is something I
1:45:40 don't think this is I don't think you
1:45:41 can play with this yet but this is a
1:45:43 paper that that Google put out
1:45:46 I don't know a month or so ago about
1:45:48 this new technology called style drop so
1:45:51 what you're looking at here is here's
1:45:53 your Source image so it's like a little
1:45:54 an image like a little kid drew a little
1:45:57 watercolor that a kid Drew and then all
1:46:00 of these images are just prompts so this
1:46:03 one is a hat in that style this is a
1:46:06 mousse in that style an F1 race car in
1:46:09 that style a butterfly a robot a coffee
1:46:10 maker so check this out I'm just gonna
1:46:13 flip through the different styles
1:46:17 I I like this I find mind-blowing
1:46:20 so there's starry starry night look it
1:46:23 completely changes the style of
1:46:24 illustration
1:46:26 there it is without blue in it there's a
1:46:28 couple of Van Gogh things in here here's
1:46:30 like a 90s icon style and look at the
1:46:33 Moose it's like a moose hanging on a
1:46:35 wall and the Hat's like this really
1:46:37 simplified hat and this is totally the
1:46:39 style that the robot would be
1:46:42 there's like one so it's like one clip
1:46:46 art image and you can just prompt
1:46:49 anything else it'll be perfect perfectly
1:46:52 in that style
1:46:53 look at this like stickers
1:46:56 I mean
1:46:58 this is insanity
1:47:01 I don't think it's out yet but like this
1:47:03 is the kind of stuff Google's doing
1:47:06 so Google's gonna make my list
1:47:12 but
1:47:13 they ain't there now Bard sucks
1:47:19 have you argued with chat GPT
1:47:23 I tell it I tell it when it gets things
1:47:25 wrong I haven't I haven't gone into like
1:47:28 a debate with it when I've asked it when
1:47:31 I
1:47:32 one of the things that I've been
1:47:34 impressed with with chat GPT is that
1:47:36 when I give it
1:47:39 when I ask at things that would be
1:47:41 polarizing in a conversation with a
1:47:44 human
1:47:46 it gives me nuanced answers
1:47:49 and that impresses me because I'm so not
1:47:51 used to that anymore like there's just
1:47:53 whole categories of topics that I refuse
1:47:56 to talk about in my life anymore because
1:47:59 I don't know who I'm in the conversation
1:48:00 with and there's like 872 trigger words
1:48:03 that's just like oh so you're a commie
1:48:06 liberal oh so you're a [ __ ] Trump
1:48:08 support like all that [ __ ] drives me
1:48:10 [ __ ] crazy
1:48:11 and so I can have conversations with it
1:48:14 without all that [ __ ] overhang
1:48:17 so but no I haven't really argued with
1:48:20 it Bart is fast it it doesn't surprise
1:48:23 me that Bard is fast and efficient and
1:48:25 technically sound what it's not is good
1:48:28 at words and humanities
1:48:31 because what what what Sam Altman and
1:48:34 open AI recognized is that the quality
1:48:37 of the data that they train it on is
1:48:38 everything
1:48:40 and they spent years finding the best
1:48:43 data sources and tagging the best data
1:48:45 sources so that their base large
1:48:47 language model was really really good
1:48:50 what I suspect Google did was did
1:48:54 something that was technically efficient
1:48:56 but they didn't Focus as much on the
1:48:58 humanity side
1:49:00 right like why is Apple
1:49:03 different because they merge art and
1:49:06 technology and I have a feeling that
1:49:08 open AI is the apple of these large
1:49:11 language model companies
1:49:14 until Apple comes along and does it and
1:49:16 then they're gonna [ __ ] rock But but
1:49:19 so I think I think Google has a steep
1:49:22 hill to push up because of that because
1:49:24 they're so engineer heavy and their
1:49:26 Humanity's light
1:49:29 I think this is going to be a rough push
1:49:31 for them on the language side
1:49:32 can you make money out of it you can
1:49:34 make money out of anything
1:49:36 out of AI yeah
1:49:39 um start doing consulting start learning
1:49:41 about it become a prompt engineer
1:49:44 um
1:49:45 um
1:49:46 openai just announced they're putting 20
1:49:49 percent of their resources into this
1:49:51 thing called super alignment and they're
1:49:53 hiring and the starting salary for
1:49:56 anyone doing super alignment for open AI
1:49:58 is four hundred thousand dollars a year
1:50:01 yeah you can make money at it
1:50:03 when arguing with Chachi PT it thinks
1:50:05 it's always right yeah it's so I like to
1:50:07 describe chat GPT as mansplaining as a
1:50:09 service
1:50:12 it will confidently answer you even if
1:50:14 it's wrong
1:50:15 so yeah it doesn't surprise me that's
1:50:17 how it argues
1:50:19 um somehow like I can't get fingers and
1:50:21 hands right I just welcome yeah that's
1:50:24 we're just new have people cross their
1:50:27 arms I don't know is mid-journey free to
1:50:30 use no it's not and you can have the
1:50:33 image you can have the image but I think
1:50:36 it's like 10 bucks a month is there is
1:50:37 their cheapest version that's what I
1:50:39 have because I don't use it there's a a
1:50:42 site worth signing up for is uh
1:50:46 leonardo.ai it's built on top of stable
1:50:48 diffusion not mid-journey but it's
1:50:50 really good it's really good interface
1:50:52 and I think they give you 150 credits a
1:50:54 day which is I think it's eight
1:50:57 eight credits per four images
1:51:02 um so you can make enough images there
1:51:04 for free so leonardo.ai I think you have
1:51:07 to get on a wait list though but it's
1:51:09 it's really good version 5.1 has fingers
1:51:12 and eye Improvement yeah 5.2 is better
1:51:14 at it as well thank you for explaining
1:51:15 you're welcome
1:51:17 oh I've got 90 okay so so if you're new
1:51:21 here my name is Kyle Shannon this is the
1:51:22 AI learning lab I obviously go through
1:51:24 and answer questions I've I am more than
1:51:26 99 messages behind so if you're like why
1:51:29 won't he answer my questions I've put it
1:51:31 in here 37 times and you just won't
1:51:32 answer it it's because I haven't gotten
1:51:35 to it yet
1:51:36 and I I I'm codependent and OCD and I
1:51:39 like to read all the questions and I
1:51:41 like to answer them all so if I haven't
1:51:42 gotten to yours yet it's just because I
1:51:43 haven't and you're a poopy head
1:51:49 and I get a little Punchy about two
1:51:52 hours in on this and let's see
1:51:54 I'm about two hours in on this
1:51:57 I used to create a lot of call synthwear
1:52:00 style Star Wars scenery cool
1:52:03 what's this code interpreter why am I
1:52:05 seeing the same things I've seen before
1:52:09 you are on The Cutting Edge
1:52:13 wait what is that
1:52:15 um
1:52:17 I'm on a roll this there's an old Harry
1:52:20 Anderson joke
1:52:21 I'm on a roll that's why I got but wait
1:52:24 wait I'm on a roll that's why I got
1:52:26 butter on my pants
1:52:28 uh Harry Anderson he was hilarious
1:52:32 all right so I can upload my Django
1:52:35 project as a report and ask it to design
1:52:37 an API
1:52:40 you mean code interpreter probably I
1:52:42 think there's a hundred megabyte
1:52:45 file size limit but you can upload a
1:52:48 compressed zip file so if you've got a
1:52:50 zip file that's less than 100 megabytes
1:52:52 you can upload it
1:52:56 and yes you could probably do that
1:53:01 that would be a good experiment
1:53:04 zip your Django project
1:53:08 uploaded into code interpreter
1:53:11 tell it to document
1:53:13 what's there if it's not documented
1:53:17 and then say I want you to help me
1:53:20 create an API with this it'll probably
1:53:21 ask you some questions Blown Away by pie
1:53:26 personal assistant who is High On
1:53:27 Emotion has anybody tried that
1:53:30 um I haven't if you're Blown Away by it
1:53:32 I'll definitely try it because there
1:53:33 it's definitely on my list of one like I
1:53:36 want to dig deep in it because they just
1:53:38 raised 1.3 billion dollars for their
1:53:40 series B
1:53:42 it's like what the [ __ ] are they gonna
1:53:44 do with 1.3 billion dollars I assume
1:53:46 they're going to buy lots of gpus and
1:53:47 and build foundational models like all
1:53:49 all the other people are but holy
1:53:52 [ __ ] Jesus that's a lot of money it's
1:53:54 a lot of money 1.3 billion dollars
1:53:58 I mean the 10 billion
1:54:00 for open AIS a lot but most of that's in
1:54:03 compute time and hosting and like all
1:54:05 the all the Microsoft [ __ ]
1:54:08 I don't know how much of that's actually
1:54:10 cash but
1:54:11 Magnum CI code interpreter Magnum CI
1:54:16 Thelonious
1:54:18 that's solid
1:54:22 Magnum CI all right I like it all right
1:54:27 until we have something better code
1:54:29 interpreter is Magnum CI it plays into
1:54:32 the Gen X thing which I'm down with
1:54:36 it retains the legacy of code
1:54:39 interpreter so the Geeks can be like
1:54:41 well they've abbreviated code
1:54:43 interpreter but technically what it's
1:54:45 doing is code interpretation so yeah
1:54:47 that is technically correct it's a
1:54:49 magnum is probably a definition that you
1:54:52 know I don't know if it's a Magnum or a
1:54:53 minimum but you know it's really easy
1:54:59 it's good it's solid it's solid
1:55:02 Thelonious bounce you should and I
1:55:05 didn't use chat GPT to name it all on my
1:55:08 own see people are still needed
1:55:11 what do you mean the robots are gonna
1:55:12 take our job Magnum CI
1:55:15 GPT wouldn't have come up with that
1:55:18 Maybe
1:55:20 that's solid though it's good
1:55:22 can you organize oh just jump me to the
1:55:25 bottom oh tick tock
1:55:26 could you fix that stupid bug God
1:55:32 for those that you know I can't even
1:55:35 talk anymore for those of you that don't
1:55:37 know why I'm whining
1:55:39 if I'm scrolling with my finger on the
1:55:41 screen at the same time someone posts a
1:55:44 comment it jumps me to the bottom
1:55:46 and I lose my place in a [ __ ] line
1:55:49 and it makes me cranky holy toledo
1:55:54 that's how far oh my God that's how far
1:55:57 oh that was Toledo Spain
1:55:59 duh thank you sorry cat Tamer I
1:56:05 it's a little slow sometimes but
1:56:07 occasionally it'll occasionally catch up
1:56:09 with something
1:56:12 oh good God
1:56:17 well he seems to know a little bit about
1:56:20 some stuff but not a lot about a lot of
1:56:23 stuff
1:56:27 uh Bart is fast okay now here's where we
1:56:30 were before okay
1:56:33 thank you for explaining you're welcome
1:56:38 Anarchy Rebels AI Magnum CI it's so good
1:56:42 it's so good Thelonious
1:56:45 all right like I have a slap sloppy
1:56:47 folder of music files could it organize
1:56:50 it for me yes
1:56:53 in fact
1:56:57 I don't depending on what metadata you
1:57:00 have in your music files if you've got
1:57:01 metadata in there like descriptions or
1:57:04 anything like that it could potentially
1:57:05 even categorize them if you if you have
1:57:07 decent meta or it could
1:57:10 it could sort them by genre and things
1:57:14 like that or by by all sorts of
1:57:16 different criteria but yeah you can take
1:57:18 a bunch of files and have it create a
1:57:19 CSV file that that is an organized view
1:57:23 of all that it that would be a good use
1:57:25 case for it actually open table has an
1:57:28 AI plug-in yes why why so can I
1:57:32 Turbo Charge my restaurant reservation
1:57:35 yeah you can say I want to have steak
1:57:38 dinner Wednesday night you know find me
1:57:40 some restaurants and it'll go find them
1:57:42 and then you say book it down it'll go
1:57:44 book it for you
1:57:47 again I don't know why you wouldn't just
1:57:48 do that at an OpenTable but
1:57:50 like I
1:57:51 get chained together multiple
1:57:54 plug-ins there's a reason you'd do it
1:57:57 what the plugins actually point to for
1:57:59 me is is this idea that we're all going
1:58:00 to have personal agents and you're not
1:58:02 gonna have to select plugins I think
1:58:04 this is a temporary state
1:58:06 I think what's gonna happen is we're
1:58:08 just gonna have these agents and you're
1:58:10 gonna have like your nightlife agent
1:58:12 that's gonna go like book dinner
1:58:14 reservations for you and you know
1:58:16 schedule [ __ ] with friends and and then
1:58:18 you're gonna have your professional
1:58:20 agent that helps you write emails and
1:58:23 you know does project plans for you and
1:58:25 then you're gonna have your I don't know
1:58:26 whatever you know dating agent is gonna
1:58:29 go do that [ __ ]
1:58:33 um
1:58:35 [Music]
1:58:40 I'm a travel agent Kyle I'll get you a
1:58:43 turtle ride
1:58:46 I like turtles
1:58:48 [Applause]
1:58:54 [Laughter]
1:58:57 it's really fun it was Toledo Spain
1:58:59 Toledo Spain all the religious [ __ ]
1:59:02 makes sense now I'm like why is it
1:59:04 giving me monasteries and [ __ ] like that
1:59:06 for Toledo Ohio like isn't there a zoo
1:59:10 there's to lead us Spain
1:59:14 every now and again
1:59:18 do you have code interpreted yes it's
1:59:20 been the topic of much conversation here
1:59:23 tonight we like to call it Magnum CI as
1:59:26 of tonight
1:59:27 um can we start a group list
1:59:29 of plugins we have used and would
1:59:32 recommend it's not a bad idea
1:59:34 could you make a list of your plugins I
1:59:36 don't
1:59:40 it's it's gonna be useless
1:59:42 because
1:59:45 once I sort of figured out what plugins
1:59:47 were
1:59:49 and then I also figured out that most of
1:59:52 them are crap and there's 600 of them
1:59:56 I could make it my full-time job to just
1:59:59 catalog which plugins are worth playing
2:00:01 with and I have a feeling
2:00:05 there
2:00:06 they're not going to survive in their
2:00:08 current form they're going to be
2:00:09 something different
2:00:11 so I'm kind of sort of sitting back just
2:00:14 waiting right now so I don't actually
2:00:16 use a lot of them
2:00:18 um code interpreter Magnum CI I'm going
2:00:20 to use a lot
2:00:22 I could use zapier for my work analysis
2:00:25 yes
2:00:26 I mean zapier's been around for like a
2:00:28 decade I mean it's it's a very robust
2:00:31 automation tool you can do all sorts of
2:00:33 [ __ ] with it it's just that now that you
2:00:35 can send stuff over to open Ai and get
2:00:38 back images or get back you know chat
2:00:41 GPT like things it it makes it
2:00:45 mind-blowingly powerful
2:00:51 are you rich no wish I were
2:00:56 I didn't I did take a company public but
2:00:58 I didn't sell any public shares took a
2:01:01 little bit of money off the table when
2:01:02 Omnicom invested and when we sold it to
2:01:04 Omnicom but I have since reinvested that
2:01:08 money and my current company
2:01:12 is where all of my net worth is
2:01:16 thanks for that walk through
2:01:18 thanks group for answering yeah I know
2:01:21 it's cool that you are answering one
2:01:22 another because I don't get to all these
2:01:24 questions it was a nice demo man thank
2:01:26 you you're the best your energy and
2:01:28 positive vibe gives me strength
2:01:30 carry on boom boom boom wow
2:01:34 you could make money with chat
2:01:39 why does he do that
2:01:42 he's a little touched
2:01:44 what did I miss everything and nothing
2:01:49 going to bed
2:01:51 Mike
2:01:53 but my renames are Dreamcoat or dreamify
2:01:56 Coda C's not bad
2:01:59 rap Sonic reverie code craft code craft
2:02:03 is interesting
2:02:07 we gotta capture something about the
2:02:09 fact that it
2:02:10 writes executes like like it plans
2:02:14 writes execute so they call those things
2:02:16 autonomous agents
2:02:22 I don't know yet we'll get there we're
2:02:24 gonna get there exactly the web version
2:02:26 is better than the app yep agent agent
2:02:29 oh agent agent that's funny
2:02:32 code code pelation
2:02:35 I like it conceptually but it's too hard
2:02:38 for human beings to pronounce because
2:02:41 you actually have to think about how
2:02:43 it's spelled because it's a it's a hard
2:02:45 word
2:02:46 window is too tiny man which window oh
2:02:49 because I had that thing up I know
2:02:52 sorry flip the screen please I did that
2:02:54 this was old I knew that
2:02:58 flip the screen so we can see the code
2:03:00 instead of up your nose I know I know I
2:03:03 figured it out
2:03:05 it's chiffon it does it definitely still
2:03:08 exists where's the beef another solid
2:03:10 Gen X reference
2:03:12 call it Moments by Home Alone code
2:03:15 interpreter
2:03:18 the
2:03:20 let's see the the McAllister
2:03:23 I don't know something there it's 10 p.m
2:03:25 Do You Know Where Your Children Are
2:03:26 another solid Gen X reference I Can't
2:03:29 Believe It's Not Butter has have
2:03:32 half-life of plutonium I know listen all
2:03:35 that [ __ ] we grew up on it wasn't good
2:03:37 but we're still here
2:03:39 hey crazy love you [ __ ] thank you thank
2:03:42 you Brian I appreciate the uh the
2:03:45 compliment
2:03:46 rename code interpreter the Kevin
2:03:48 McAllister experience
2:03:51 that's not bad
2:03:53 Master experience these are all solid
2:03:55 Kyle you have code interpreter right now
2:03:57 I it just went live
2:03:59 for all plus users I know that's what
2:04:02 we've been talking about it's [ __ ]
2:04:04 crazy
2:04:05 how do you adapt documents to chat GPT
2:04:08 you go to code interpreter so you have
2:04:12 to so you have to be a plus member
2:04:15 then you have to
2:04:18 go to your settings and go to Beta
2:04:20 features and turn on code interpreter
2:04:23 then you go to code interpreter and
2:04:25 there's a little plus button in the chat
2:04:28 icon that's how you upload files there
2:04:32 you give the path to the file
2:04:34 now well you might be able to give the
2:04:36 path to the file but the little plus
2:04:38 button lets you upload things
2:04:40 what's the difference between chat GPT
2:04:42 and Bing okay
2:04:44 so down here is this thing called the
2:04:47 large language model this is sort of the
2:04:48 core engine gpt4
2:04:52 chat GPT is an application that sits on
2:04:54 top of it and it's the thing that makes
2:04:58 it talk to you politely and it's it's
2:05:00 the thing that says I'm sorry Dave I'm
2:05:02 afraid I can't do that right it's got
2:05:03 all the rules right it's the application
2:05:06 and so that's chat GPT and then Bing
2:05:10 chat is also sitting on top of that same
2:05:13 gpt4 large language model and it's got
2:05:15 very very similar
2:05:18 um
2:05:19 functions as chat GPT but it's got
2:05:22 different ones and it's it's got
2:05:23 different rules so it's connected to the
2:05:26 internet it it incorporates
2:05:29 um Dali image generation
2:05:32 um it's it's a bit more restrictive and
2:05:34 a bit less
2:05:35 good
2:05:37 than chat GPT so so they're basically
2:05:40 just two flavors of the same thing
2:05:43 sitting on the same enabling technology
2:05:47 can it generate 3D model from a 2D image
2:05:50 not yet that's coming I would say that
2:05:52 like there's there's a ton of white
2:05:54 papers right now
2:05:56 on text to 3D object and then there's a
2:05:59 ton of white papers on extracting 3D
2:06:03 um information from 2D images and doing
2:06:06 that so
2:06:08 I don't know give it six months and
2:06:10 that'll be here and it's there's
2:06:11 probably tools you can play with now if
2:06:14 you go to
2:06:15 futurepedia.io there's probably stuff I
2:06:17 don't know about that's doing
2:06:19 text to 3D object that doesn't suck but
2:06:23 it's all it's all super new right now so
2:06:24 I'd say give give it six months what
2:06:27 kind of computer do I need mine is old
2:06:29 and I need a new one I want to play with
2:06:30 AI
2:06:31 um it depends what you want to play with
2:06:33 AI if you want to like host your own
2:06:35 models and and actually run large
2:06:37 language models locally
2:06:39 then if you're a Mac person get it get
2:06:43 the highest spec M2 you can get if
2:06:46 you're a PC person you're going to need
2:06:48 to get a a cranked out
2:06:51 multi-gpu system like what's your budget
2:06:55 like George hotz is is designing a
2:06:58 fifteen thousand dollar machine learning
2:07:02 you know Beast
2:07:05 so it depends on what you want to do
2:07:07 with it if you just want to play with
2:07:09 chat GPT
2:07:10 any [ __ ] computer your phone works
2:07:13 right because all the compute is
2:07:15 happening on Microsoft servers somewhere
2:07:17 and gpus
2:07:20 so it depends what you want to do if you
2:07:22 want to do local
2:07:24 highfalutin
2:07:26 processing then you need a fancy
2:07:28 computer if you just want to use these
2:07:30 tools you don't need [ __ ]
2:07:35 um other than plugins what's the
2:07:37 distinct difference between 3.5 and 4.
2:07:39 so
2:07:40 the quality difference when it comes to
2:07:43 language and program just just the
2:07:46 quality of the response you get back
2:07:48 from gpt4 versus GPT 3.5 is
2:07:53 it's subtle but it's like staggeringly
2:07:56 different
2:07:56 like if I have GPT 3.5 write me
2:07:59 something it gets me to like 65 or 70
2:08:01 percent gpt4 gets me to like 80 to 90
2:08:05 percent and like you know like there
2:08:08 there like I don't have to edit nearly
2:08:11 as much if I'm using gpt4 and it's
2:08:13 apparently better at code as well
2:08:16 yeah
2:08:17 and then the other difference is code
2:08:19 interpreter
2:08:20 so if you pay for
2:08:23 so so a couple of things if you want
2:08:25 gpt4 for free use Bing chat Bing chat is
2:08:30 gpt4
2:08:31 if you want access to all of open ai's
2:08:34 newest Technologies as soon as you can
2:08:36 get them
2:08:38 pay for their API and pay for a chat GPT
2:08:41 here GPT plus and you'll get early [ __ ]
2:08:44 here and then if you're paying for apis
2:08:46 you'll get access to the API stuff over
2:08:49 there so it depends what you want to do
2:08:51 if you don't want to spend any money
2:08:52 just use Bing chat that's gpt4
2:08:56 GPT is faster for smarter that's right
2:08:59 gpt's like don't get me wrong GPT 3.5 is
2:09:03 still really good and it is very fast so
2:09:05 if you want stuff fast do that Google is
2:09:08 doing a Xerox Park in the 70s where they
2:09:11 ignored the future and they developed
2:09:12 CRX part they're doing a Kodak right
2:09:16 Kodak [ __ ] invented the digital
2:09:18 camera
2:09:20 the dude walks into the
2:09:23 boardroom and says here's the digital
2:09:25 camera it's the future of Kodak and
2:09:27 they're like boy we're in the silver
2:09:30 halide business that stupid Contraption
2:09:32 of yours ain't gonna sell more film now
2:09:34 is it now get out of here you cow poke
2:09:40 yeah they are I think
2:09:43 it's [ __ ] up but it's you know
2:09:46 I've argued with it before I knew it it
2:09:49 had a character limit it was creating
2:09:50 something very long and that's
2:09:52 interesting
2:09:53 how do you rank kyber for video
2:09:55 generation what's better I've only
2:09:57 really played with
2:09:59 um
2:10:00 with Runway with the video generation
2:10:02 I've heard kyber
2:10:03 um from from a lot of people I trust
2:10:05 kyber is really really good I just
2:10:07 haven't played with it
2:10:08 um there's there's uh descript
2:10:14 um there's the script Runway kyber are
2:10:17 the three biggies that I know of
2:10:20 Kyle I can tell you how to access code
2:10:22 interpreter right now I have it
2:10:26 uh
2:10:27 thoughts on open llms
2:10:30 um
2:10:32 cool I can't keep up with them a new one
2:10:35 comes out every three days so I'm just
2:10:37 waiting until the the day that one comes
2:10:40 out where they're like
2:10:41 here's here's the seven billion
2:10:43 parameter llm that's as good or better
2:10:45 than gpt4 that you can install on your
2:10:48 phone or your
2:10:49 your Mac and you don't need any fancy
2:10:51 computer when that comes out then I'll
2:10:54 then I'll pounce on open llms
2:10:58 I just don't have time to keep up with
2:11:00 it right now and I'm not geeky enough to
2:11:02 be in any of the projects I I
2:11:05 pay attention to to the announcements
2:11:09 and all the announcements are like we're
2:11:11 almost as good as 3.5 we're just better
2:11:13 than 3.5
2:11:15 I'm waiting for it's tiny and it's as
2:11:19 good as for go get them
2:11:21 that I'm excited about the alignment
2:11:23 layer on chat gbt straight up refuses to
2:11:26 discuss my innocent topics yeah I know
2:11:29 so you can either jailbreak it go look
2:11:32 up uh Dan do anything now or go start
2:11:35 playing with the open source large
2:11:37 language models that don't have any any
2:11:39 of the constraints on them
2:11:41 if you want to if you want to sort of
2:11:44 push the boundaries and it's getting
2:11:45 frustrating it's yeah those the the the
2:11:48 safety guideline stuff can be really
2:11:50 frustrating depending on your topic
2:11:53 I thought he said he was still waiting
2:11:55 for it nope I got it I was I said that I
2:11:58 don't know two lives ago
2:12:01 I'm so addicted to them I'm not sure to
2:12:04 what but
2:12:06 let's see that that mid-journey is super
2:12:09 awesome that and chat gpt4 is worth
2:12:12 paying for I agree yo Ronda what's
2:12:14 happening
2:12:15 let's chat this week okay I'm around
2:12:17 I don't have a busy week this week so
2:12:19 let's talk Rhonda
2:12:21 um
2:12:22 we're 46 with 130 saved for retirement
2:12:26 wait
2:12:27 is this good
2:12:29 or not real honest opinion please
2:12:34 I don't know what you mean Joel is what
2:12:36 good or not real is is generative AI
2:12:39 good and it's very real
2:12:41 it's very good it's profoundly good I
2:12:44 don't know why you're asking are you
2:12:45 asking about investment advice don't ask
2:12:48 me I have a degree in acting
2:12:50 I'm [ __ ] horrible with money
2:12:53 but but yes AI is real it's gonna
2:12:56 [ __ ] change everything it's gonna
2:12:58 profoundly transform the world
2:13:02 um
2:13:03 but where would I place a bet I am not
2:13:06 the right guy to ask can code
2:13:08 interpreter make an Excel spreadsheet
2:13:10 from a mySQL database my guess is yes I
2:13:13 would I would yes I think I don't know I
2:13:17 would
2:13:18 try it I think it can do a lot I think
2:13:22 it can do way more than we
2:13:25 no
2:13:26 because what I'm seeing right now is if
2:13:29 you go to Twitter
2:13:31 and
2:13:33 um
2:13:33 look up code enter like code interpreter
2:13:35 was trending today which is good right
2:13:37 people are using it but if you go look
2:13:40 at what people are posting most of what
2:13:41 they're posting is them trying [ __ ]
2:13:45 other people already talked about two
2:13:46 months ago
2:13:48 it's almost all that so I'm kind of
2:13:50 waiting for a week or two for some
2:13:51 actual real smart people to do
2:13:53 interesting [ __ ] with it Ethan Malik put
2:13:55 out
2:13:57 um a post today
2:13:59 about a prompt that he wrote to to
2:14:01 create some really interesting
2:14:03 um charts and graphs based on Edward
2:14:05 tufty which I think is really smart
2:14:08 um so go look up Ethan Malik his his
2:14:10 post today on on the prompt that he
2:14:12 created it's really really cool and he's
2:14:14 had code interpreter for like three
2:14:15 months
2:14:16 if you if you joined Leonardo's Facebook
2:14:18 group they'll Grant you instant access
2:14:20 oh but Lauren lay thank you so much
2:14:24 that's great news to hear okay so
2:14:29 that's huge that I mean that's a huge
2:14:31 tip so join Leonardo's Facebook if if
2:14:35 you want to do image generation
2:14:38 leonardo.ai is amazing and if you join
2:14:41 their Facebook group and they give you
2:14:42 instant access go do that right now it's
2:14:45 it's a it's an amazing tool it really is
2:14:48 and they give you a fair amount of
2:14:50 tokens to play with for free every day
2:14:52 so if they're not stingy
2:14:55 all right I gotta go it's 11 o'clock
2:14:57 it's time I don't know where my children
2:14:59 are well they're upstairs but you know
2:15:01 other than that
2:15:05 Tick Tock just jumped me to the bottom
2:15:06 of comments so I'm just gonna go I'm
2:15:08 gonna grab some of these bottom comments
2:15:09 here and then I am gonna get out of here
2:15:13 how do I get a job in AI
2:15:17 go there start getting AI literate
2:15:20 here's my recipe
2:15:26 go start playing
2:15:29 get good at first parlor tricks like
2:15:32 just understand it can write songs it
2:15:33 can write poems it can write code it can
2:15:35 make tables it can it did it like start
2:15:37 learning it
2:15:39 then take what you do
2:15:42 and learn how that can help you do what
2:15:45 you do better
2:15:47 so if you're a project manager figure
2:15:49 out how you can use chat GPT to do
2:15:51 better project management if you're a
2:15:53 writer if you're a whatever whatever you
2:15:55 do whatever you're interested in
2:15:58 dig deep in your areas of knowledge as
2:16:01 you're doing all of this share it
2:16:06 join groups put it on LinkedIn put it on
2:16:10 Facebook
2:16:11 tell people at dinner parties tell
2:16:14 people at work ah I I'm playing with
2:16:17 chat gbt and I had it write me a country
2:16:19 song
2:16:20 start with that
2:16:22 and let them look at you weird like I
2:16:25 heard the robots are gonna kill us why
2:16:27 are you playing with that stupid AI
2:16:28 thing I think it's the devil isn't it
2:16:30 the devil I heard it's the devil isn't
2:16:32 it aren't the robots gonna kill us I
2:16:33 heard it's a waste of time you know what
2:16:35 I heard I heard it lies
2:16:36 I heard it's stealing all this
2:16:38 information you're not going to use that
2:16:39 right it's stealing everyone's
2:16:41 information ignore all that
2:16:43 and just keep talking about it because
2:16:46 at some point one of those craziest
2:16:48 people in your life is going to go you
2:16:50 know I need to figure out to say I
2:16:51 should
2:16:52 who should I talk to oh unbridled son
2:16:55 has been talking about that for the last
2:16:57 three weeks he won't shut up about it I
2:16:59 got to talk to him can you introduce me
2:17:01 that's how you get a job
2:17:03 put it on your LinkedIn join groups we
2:17:06 had the AI salon so this URL go here
2:17:11 the salon.ai
2:17:13 that's a community I started for people
2:17:16 curious about AI we've got a Discord
2:17:19 server we meet every other week it's a
2:17:21 group of people that are just kind of on
2:17:23 this adventure together
2:17:24 we have a lot of fun and
2:17:28 people in this group have put the fact
2:17:30 that they're a part of this group on
2:17:31 their LinkedIn profiles and gotten jobs
2:17:33 because of it I'm not guaranteeing that
2:17:35 at all like
2:17:36 but just start talking about it because
2:17:38 people are getting increasingly curious
2:17:40 about AI so get literate about it
2:17:43 and and start talking about it and start
2:17:45 helping people the other thing to do
2:17:46 start helping people
2:17:48 because one of the things that
2:17:52 um that I've noticed is people like well
2:17:54 I don't know where to start I just look
2:17:55 at that that you know the open AI box
2:17:58 and I don't know what to do
2:18:00 rather than thinking what can I do with
2:18:02 AI flip it upside down
2:18:04 and say what problem needs solving
2:18:08 and then ask okay how can I use AI to
2:18:11 solve that problem
2:18:13 so so one of the things that you can do
2:18:15 is you can go to people and say hey I'm
2:18:18 learning this AI stuff and I know you've
2:18:20 been struggling you know finding a
2:18:22 marketing agency I don't know anything
2:18:24 about marketing but
2:18:26 would you mind if I used chat GPT to
2:18:30 come up with a marketing plan so I could
2:18:31 learn it and maybe you could benefit
2:18:33 from that too sure you can do that for
2:18:35 free for me like give some [ __ ] away for
2:18:39 free and go learn it
2:18:40 thank you for the balloons
2:18:44 um just start playing get curious get
2:18:46 curious get curious play play play play
2:18:48 play play
2:18:49 make it an adventure
2:18:52 like any good Adventure it should be fun
2:18:55 it should be scary
2:18:59 there should be a lot at stake
2:19:01 you should do it with others
2:19:03 ideally if you're on a good Adventure
2:19:05 you're doing it with others there's
2:19:06 there's like the meat head that can
2:19:08 swing an ax there's like the little
2:19:10 Brainiac mousy one that can't talk but
2:19:13 can solve problems
2:19:15 there's the charismatic one there's the
2:19:17 wizard right you go you get your group
2:19:19 of four or five people that all have
2:19:20 different skills go on this [ __ ]
2:19:22 adventure together with them
2:19:25 because when you start talking about it
2:19:27 you're gonna find out oh there's someone
2:19:29 else in your group that's been secretly
2:19:30 doing this too
2:19:32 oh you're doing this too
2:19:34 you want to go have coffee and talk
2:19:35 about it yeah do you know what you're
2:19:38 doing no
2:19:42 that's how you get a job
2:19:44 you can't monetize AI chat tools unless
2:19:47 you do a subscription
2:19:48 you can monetize Consulting
2:19:52 that's just not true
2:19:57 you could go use bing.com for free and
2:20:00 go tell people I'll I'll use
2:20:03 I'll use AI to give you what your agency
2:20:06 is charging you ten thousand dollars for
2:20:09 I'll do it for five and I'll deliver it
2:20:11 twice as fast
2:20:12 want to take a shot
2:20:15 you don't need a subscription to do that
2:20:16 awesome work thank you you're welcome
2:20:19 headlines Sarah Silverman suing
2:20:22 open AI meta for copyright infringement
2:20:25 like I said the next 20 years is going
2:20:28 to be nothing but a string of one
2:20:30 lawsuit after another at some point
2:20:33 there's going to be some precedent set
2:20:34 where it's like well Sarah Silverman's
2:20:37 you know DVDs are also in libraries and
2:20:41 libraries are there for you know fair
2:20:43 use and knowledge consumption hey hey
2:20:47 um
2:20:49 just the lawsuits are going to keep
2:20:52 coming doesn't surprise me
2:20:56 it
2:20:57 whatever it's you know it's the this is
2:21:01 there's a there's going to be a lot of
2:21:03 what I what what I'm calling business
2:21:05 model Panic coming
2:21:07 which is my business model relied on
2:21:10 this particular interpretation of
2:21:12 copyright law which these tools
2:21:14 fundamentally [ __ ] blow up
2:21:17 so there's going to be a lot of
2:21:19 companies a lot of big organizations
2:21:21 going we can't lose our business model
2:21:23 we can't lose our sue them sue them sue
2:21:25 them sue them it's what they've always
2:21:27 done it's what they'll always do I think
2:21:29 this this is so much different than
2:21:33 um
2:21:34 like music sampling in the 80s and 90s
2:21:37 with hip-hop remember all the remember
2:21:38 all the copyright suits you're stealing
2:21:40 our music and this and that and and it
2:21:42 took him a decade to sort of figure out
2:21:44 fair use and and how all that stuff
2:21:46 worked this is just a fundamentally
2:21:48 different thing
2:21:50 sampling music for hip-hop you're
2:21:52 actually taking part of someone else's
2:21:54 song copying it and pasting it into your
2:21:57 song this stuff is generating new
2:22:00 original content based on some
2:22:02 amalgamation of training data that may
2:22:05 include something Sarah Silverman said
2:22:07 but it may not
2:22:09 and how you determine what percentage of
2:22:12 all of the [ __ ] that was on the Internet
2:22:14 is that Sarah Silverman thing that ends
2:22:16 up in that one line of that one poem
2:22:19 is essentially impossible so it's a
2:22:21 fundamentally different thing it doesn't
2:22:24 mean
2:22:26 everyone's not gonna sue the [ __ ] out of
2:22:28 everyone they are
2:22:30 so this is just the beginning like the
2:22:32 lawsuits the lawyers are like I said
2:22:35 I've said to this year before the
2:22:37 lawyers right now are pre-ordering their
2:22:39 new yachts
2:22:40 they're gonna make a [ __ ] fortune
2:22:43 on both sides
2:22:45 there's gonna be lawyers going ah Sarah
2:22:47 we can help you sue open AI they just
2:22:49 got 10 billion dollars from Microsoft
2:22:51 let's go get some of that cold cash
2:22:53 and then on the other side they're gonna
2:22:55 be like yeah open Ai and Microsoft we
2:22:57 can help defend against all these
2:22:58 frivolous lawsuits
2:23:00 let's go
2:23:02 the lawyers are thrilled
2:23:05 imagine if every third sentence in AI
2:23:07 response was a sponsored reply that's
2:23:10 someone's gonna do that Google's
2:23:12 probably gonna so so Microsoft doesn't
2:23:14 have to do that because they can
2:23:16 monetize this [ __ ] with all of their
2:23:18 applications Google can't
2:23:23 but then
2:23:25 if Google's
2:23:27 chat
2:23:29 tool
2:23:32 is too saturated with sponsored content
2:23:35 and you compare it to open AIS or open
2:23:38 source models that have no sponsorship
2:23:40 in them
2:23:41 is there a point at which it's just not
2:23:44 useful and it drifts to the background
2:23:47 probably
2:23:50 any free AI tools that have Enterprise
2:23:53 or commercial rights that you know well
2:23:55 not free Adobe if you're doing images
2:23:58 the Adobe stuff is commercially safe
2:24:02 um
2:24:05 all of this generative large language
2:24:07 model stuff depending on how
2:24:09 conservative the organization is
2:24:12 might steer clear of this for a while
2:24:14 but but you should go read the article
2:24:17 from Ethan mullet called
2:24:19 um
2:24:20 is it discovering detecting detecting
2:24:22 your secrets the secret cyborgs
2:24:25 so what what Malik talks about in that
2:24:27 piece detecting the secret cyborgs is
2:24:30 that
2:24:32 individuals in in big companies are
2:24:35 using large language they're using chat
2:24:37 GPT as personal productivity enhancers
2:24:40 but they're not telling the company
2:24:41 about it and the company's not using
2:24:43 these things because they're not ready
2:24:44 for prime time
2:24:45 and so what's happening is you've got a
2:24:48 class of AI literate employees that are
2:24:50 being more efficient and more innovative
2:24:53 in secret and corporations that aren't
2:24:56 benefiting from that
2:24:58 and they're afraid to say anything
2:24:59 because maybe there's a policy that we
2:25:02 don't use AI or maybe there's a policy
2:25:04 that uh that you know you can only use
2:25:07 it in a certain way or maybe one of
2:25:09 these employees is now doing in four
2:25:11 hours what used to take them 30 hours
2:25:13 and they don't want to tell anyone about
2:25:14 it
2:25:15 so so companies are going to need to
2:25:18 find a way to close that Gap it's like
2:25:19 we're in a fascinating time right now
2:25:22 um
2:25:24 so I don't know like like these these
2:25:27 tools are all new right now
2:25:30 um the the legal questions around
2:25:32 anything generated with these tools is
2:25:34 is all over the place
2:25:36 um
2:25:38 when Chachi PT is built into Microsoft
2:25:41 Word
2:25:43 and I'm writing an email
2:25:45 and there's a button there that says
2:25:47 help me write
2:25:48 at what percentage of that email does it
2:25:52 have to be AI generated for the whole
2:25:54 email to be considered AI generated is
2:25:57 it just a word is it just a sentence is
2:26:00 it the whole body
2:26:02 I think the reality of these tools as
2:26:04 they get more and more integrated into
2:26:06 our business tools is that everything
2:26:08 produced everything produced by people
2:26:11 will be some hybrid of AI and people
2:26:15 and so I don't think it's going to be as
2:26:17 clear as this is AI and this isn't
2:26:20 so I I like I have no answers for you
2:26:23 unfortunately other than start learning
2:26:26 it if you own a company start learning
2:26:29 it if you don't own a company start
2:26:31 learning it
2:26:33 if you're afraid of it start learning it
2:26:38 just start getting your head around what
2:26:39 it is and then as the legal stuff gets
2:26:43 sorted
2:26:44 over the next 20 years
2:26:48 do the appropriate thing
2:26:54 what type of accessibility features are
2:26:56 found in chat GPT for blind low vision
2:26:59 people so so there's a lot of things you
2:27:02 can do you can
2:27:04 there are what are they called are they
2:27:07 called automations on the iPhone
2:27:12 I forget what they're called on the
2:27:14 iPhone anyway they're automations but
2:27:16 there are things you can do on the
2:27:17 iPhone where you can basically invoke
2:27:19 Siri so you can talk to Siri and say
2:27:21 like Siri GPT or something like that and
2:27:24 you'll actually be talking to GPT I also
2:27:27 think the new
2:27:28 pie tool from uh inflection
2:27:33 is is voice is voice activated
2:27:37 um
2:27:40 so I think I think whatever
2:27:41 accessibility you use with voice to text
2:27:44 and text to voice
2:27:46 um is is all still in play with chat gbt
2:27:50 it's it's really just it's just like a
2:27:51 Google search so to the extent that you
2:27:54 can do Google searches you can do things
2:27:56 with chat gbt um but I don't know that
2:27:59 they're explicit
2:28:00 accessibility tools
2:28:02 uh built into it but but there's there's
2:28:05 plenty certainly within the Apple
2:28:07 ecosystem to deal with all that
2:28:10 um thanks for the info midnight here
2:28:11 hate to go but have to I know I gotta go
2:28:13 too it's 11 10 here
2:28:15 all right uh plus someone needs to
2:28:16 Chicago I gotta say all right listen I
2:28:18 gotta go I'm gonna do one more
2:28:19 one more
2:28:22 why is host not reading the comments
2:28:24 because I I am
2:28:26 can you join us for 10 seconds sorry
2:28:29 folks unless you have a server Farm
2:28:30 dozens of full-time computer scientists
2:28:33 okay that's funny I have a quick
2:28:35 question ask it
2:28:37 uh
2:28:39 um you get nothing from AI tools okay so
2:28:42 I think this is not live because he's
2:28:43 not responding to ghost questions I am
2:28:46 live
2:28:47 faith hope and love
2:28:50 see if I weren't live could I say your
2:28:52 name
2:28:53 awkwardly and as if I can't read it
2:28:56 because you didn't use capital letters
2:28:58 and it just seems like alphabet soup
2:29:00 if you capitalize the first letter of
2:29:03 each word it'll it'll improve usability
2:29:05 of your name
2:29:06 Faith open love
2:29:10 what's Blue Willow Blue Willow
2:29:13 looks like
2:29:14 mid-journey because you make images in a
2:29:17 Discord server but it's built on top of
2:29:19 stable diffusion which is not quite as
2:29:21 good as mid-journey but it's free so far
2:29:25 out
2:29:29 all right did you ever hear NPR's
2:29:31 broadcast regarding ai's ability to take
2:29:33 the reporters jobs yet
2:29:35 no but that doesn't surprise me but it
2:29:38 could also
2:29:39 I I'm I'm excited about the return of
2:29:43 local journalism where it's not local
2:29:46 journalists but it's AI journalists
2:29:48 because I think you could
2:29:49 I think we could help our country by by
2:29:53 re-inventing local journalism with AI
2:29:56 so drink water thank you
2:30:01 I know my voice has gone all crispy all
2:30:03 right listen peace out everyone I gotta
2:30:05 go I'm I'm moist
2:30:08 I can't talk my ass is numb I've been
2:30:11 doing this for three hours I just can't
2:30:13 people I can't do it anymore I've been
2:30:16 on the back of this horse for days now
2:30:18 it's days
2:30:21 thank you very much
2:30:23 I appreciate the Roses I appreciate the
2:30:26 hearts I appreciate the questions I
2:30:28 appreciate you all being willing to be
2:30:31 open to giving this [ __ ] a try
2:30:34 I think it will help you
2:30:36 and uh I'll see you tomorrow peace