aiLL 7/18/23 - The Lightning Pace of AI is Electrifying

Description

Chapters

Transcript

0:06 ah
0:25 all right
0:28 it's it's the AI learning lab late night
0:32 Edition
0:33 and AI Salon tonight
0:35 we had some uh
0:38 some fans of The Tick Tock live showed
0:41 up at the AI salon
0:44 had a good good Salon tonight had a good
0:46 hangout
0:48 welcome welcome welcome
0:50 I don't know how much we're gonna hate
0:52 Tobias what's happening
0:54 I don't know how long I'm going to last
0:55 tonight we're getting a mighty late
0:57 start
0:58 but welcome to the AI learning lab if
1:00 you're new here check out these here
1:02 urls
1:03 there's lots of them why
1:06 because [ __ ] got way better real fast
1:08 this week there's a lot of stuff here to
1:10 check out and this is just sort of the
1:12 basics
1:14 so if you are new to General
1:17 if you're new to generative AI these are
1:20 things you should be checking out
1:23 that's the official chat GPT website if
1:26 you pay for the plus version you get
1:27 plug-ins you get gpt4 and you get code
1:30 interpreter code interpreter alone is
1:33 worth way more than 20 bucks a month in
1:35 my humble opinion bing.com is free
1:38 that's Microsoft's gpt4 connected to the
1:41 internet can make images can read images
1:44 cloud.ai is the competitive gpt4
1:49 lets you upload five files has a hundred
1:51 thousand token context window which
1:53 basically means you can put massively
1:55 long prompts in it like an entire novel
1:58 75 000 words and it'll just deal with it
2:02 poe.com gives you access to a bunch of
2:04 these things and then bard.com I found
2:06 something out today
2:08 bard.com suddenly mysteriously got
2:11 better
2:12 and it's because they completely replace
2:15 the core large language model with their
2:18 new I forget what it's called begins
2:20 with a G
2:21 not Gilead that would be funny
2:25 um not Galileo anyway it's something but
2:28 bard.google.com is is now better than it
2:31 was so worth checking out so all five of
2:35 these if you're new to chat GPT are
2:38 worth exploring
2:40 um go to prompts.chat that'll teach you
2:42 how to prompt and then
2:44 futurepedia.io will teach you how to
2:48 um
2:49 it's just a big long list of AI sites so
2:52 if you want to if you want to figure out
2:54 if there's some music thing you want or
2:55 a movie thing you want that's just a big
2:57 old pile of tools okay so that's the
3:00 basics my name is Kyle Shannon
3:04 [Music]
3:06 this is the AI learning lab
3:09 we have fun here I'm getting a very late
3:12 start tonight so uh if you've got
3:14 questions about AI pop them in the
3:16 comments below welcome everybody we had
3:19 a really good AI Salon tonight uh we had
3:22 a gentleman speak that has been uh
3:26 installing sort of corporate grade AI
3:28 systems for many many years that was
3:31 fascinating listening to him talk about
3:33 all the issues and what he's doing and
3:36 what he's seeing so that was awesome we
3:38 had
3:39 three that I counted members of The Tick
3:42 Tock live audience that were at the
3:43 salon tonight that were like hey Kyle
3:45 what's happening
3:47 one guy said do you ever sleep and I was
3:50 like no cheers
3:53 anyway
3:54 if you have uh questions comments uh
3:58 about the uh the old generative AI thing
4:01 popping below a little bit about my
4:03 background I'm an entrepreneur I'm the
4:06 founder and CEO of a company called
4:07 storyvine we've been around for 11 years
4:09 it's an automated video storytelling
4:11 platform that we figured out a way to
4:13 automate video production and
4:17 um we are now rolling some AI features
4:20 deep into our platform and we actually
4:23 just got a uh a hint from a large client
4:26 today that they might be paying us to go
4:28 build some of the AI stuff we've been
4:30 pitching
4:31 that we've got prototype down and I
4:33 think they want us to go build that [ __ ]
4:35 so
4:37 you can make money
4:38 [Laughter]
4:42 oh okay let's get to the questions
4:47 hello Apocrypha what's happening Mama
4:49 magic okay what are the essential skills
4:52 needed to land a prompt engineering job
4:58 I would say at this point and probably
5:01 for the next year
5:02 [Music]
5:05 um
5:06 deep curiosity a sense of adventure and
5:09 exploration
5:10 um and and for example here's here's why
5:12 so
5:14 at the beginning of this week or end of
5:16 last week this list only had three sites
5:19 on it I've added two I could also add pi
5:23 and today meta released their llama 2
5:26 model
5:28 open sourced with a commercial license
5:31 for free for anyone to use unless you
5:34 have more than 700 million customers so
5:36 basically you know
5:38 uh Instagram scale
5:40 companies can't use it for free everyone
5:42 else
5:43 free to go build on it
5:46 so
5:50 understanding how these models work
5:53 understanding which ones are better for
5:55 what understanding which ones have
5:57 security issues or not which ones have
5:58 apis or not
6:01 um if you're brand new to prompting
6:03 um that document prompts.chat will teach
6:06 you about prompting
6:07 there's some primers in there it also
6:10 teaches you this concept that you can
6:12 set these large language models up as a
6:14 particular Persona so
6:16 act like an air traffic controller act
6:18 like a magician act like a I don't know
6:22 consultant act like a psychologist and
6:25 when you do that these things get better
6:27 there's a reason for that I'm happy to
6:28 describe it but understanding the basic
6:31 concepts being able to get these things
6:34 to do what you want them to do is is
6:36 probably the first skill and then the
6:38 second skill after that is understanding
6:41 business these are just tools so
6:44 the trick with tools is don't get
6:46 seduced by the tool
6:48 just learning prompt engineering is not
6:50 enough but understanding how to ask
6:54 questions for who your customer might be
6:56 so if if it's someone that's hiring you
6:58 understanding what are they trying to
7:00 accomplish and
7:02 how quickly can you get this tool to
7:06 give them what they want to accomplish
7:07 so focus on the problem then figure out
7:11 which of these tools is the best to do
7:12 it with or if it's mid-journey or stable
7:15 diffusion or adobe Firefly whatever it
7:18 might be
7:19 um I I think those are the those are the
7:21 the biggest skills and then and then the
7:23 kind of marketing piece of it right now
7:25 quite frankly is talk about it so as you
7:28 start teaching yourself these tools put
7:31 it out in the world put it on LinkedIn
7:33 hey I just built this thing on chatgpt
7:35 you're at a dinner party oh let me tell
7:37 you about this thing I did with chat gbt
7:38 most of the people right now will look
7:40 at you like they're [ __ ] like you're
7:42 crazy
7:43 let them who gives a [ __ ] [ __ ] them
7:46 they're the ones that'll call you three
7:48 months from now go and I should have
7:49 listened to you at that dinner party
7:50 sorry I was such a dick well now I've
7:53 got a job and you don't [ __ ]
7:56 because you never know who you're
7:58 talking to right now
8:00 is going to be in that moment where
8:02 they're like holy crap I need to figure
8:03 this stuff out does anybody know that oh
8:05 mama magic does she was talking about
8:06 chat gbt the other day right so be that
8:09 person
8:10 right now I think that's it that's
8:12 that's the primary piece of it
8:15 and then you can watch YouTube videos
8:17 and you can download PDFs don't pay for
8:20 any of that [ __ ] because the the
8:21 prompting game is changing way too
8:23 quickly right like I thought my head I
8:26 had my head around prompting and then
8:28 code interpreter came out and now now
8:30 you don't really need much of a prompt
8:32 for code interpreter but you got to
8:33 figure out what the [ __ ] you're gonna do
8:34 with it because it's remarkable
8:37 but like what do you actually do with it
8:39 I don't know I don't know yet I haven't
8:41 figured that out so uh be flexible be
8:45 hungry be curious be adventurous and
8:48 talk about it that's that's it
8:50 that's all it's easy
8:53 oh no it's simple it's not easy actually
8:55 it is easy because it's fun
8:59 if you like learning new [ __ ] this this
9:01 is amazing amazing stuff what site will
9:03 create
9:04 a singer performing someone else's music
9:08 um I don't know which one they're using
9:10 to like to to sort of map the voices
9:13 um to someone else's song I would go to
9:16 futurepedia.io or just go to Twitter I
9:19 know when the Drake and the weekend song
9:21 came out a bunch of people posted here
9:23 was the tools that they used I haven't
9:25 really paid attention to that all that
9:27 much I've been paying more attention to
9:28 the talking ones not the music ones
9:31 so I don't really know so I'd go to
9:33 futurepedia and then I'd go to Just
9:34 Twitter and search search that up what
9:37 are you selling
9:38 um nothing really
9:41 I'm you know I'm trying to
9:44 I'm trying to learn this [ __ ] I'm trying
9:46 to teach as many people as I can
9:48 um people here have asked me to do a
9:49 workshop I don't have a workshop but I
9:52 have a Graphic for a workshop that I
9:54 made
9:55 and I have a survey to learn what people
9:58 might want so at some point I might sell
10:00 that but I'm not selling it now
10:02 why you buy it
10:06 I love the I love how people have a
10:09 cynical default he's up to something man
10:12 he's got to be selling something oh I
10:14 know he's hired by one of these
10:16 companies he's just a shield for chat
10:18 GPT
10:22 nope I'm not getting paid for this
10:25 I do these two to three hours a night
10:28 um
10:29 I'm a CEO of a company
10:31 uh it's an 11 year old company
10:34 um I'm doing this instead of sim racing
10:36 because I enjoy it and because I'm
10:37 trying to keep my head in the game
10:40 this [ __ ] is moving so fast it is making
10:43 I I can't I can't comprehend how fast
10:47 shit's moving so what this channel is
10:50 about is telling as many people as I can
10:53 as enthusiastically as I can get [ __ ]
10:57 curious lose the cynicism
11:00 gain the Curiosity gain the empathy get
11:03 adventurous and learn this [ __ ]
11:06 because if you don't you're gonna find
11:09 yourself on the other end of a [ __ ] slap
11:12 Fest and it's not gonna be fun
11:14 so I'm not really selling anything
11:17 I am I am selling hope
11:21 some
11:23 no all right
11:30 what's a good one to write a movie
11:32 script gpt4 the top one
11:36 claude's not bad Bing is gpt42 but it's
11:40 been neutered put it in creative mode
11:41 it's not bad but the the gpt4 at open AI
11:46 is still the best by far and now
11:50 it looks like it's getting worse
11:54 uh I don't know what's going on at open
11:56 AI something's going on at open AI
11:58 they're they're having all sorts of
11:59 issues they they pulled browse with Bing
12:01 functionality there's something going on
12:03 over there with lawyers and cutting
12:05 deals and getting sued and something's
12:07 going on but
12:09 to write a screenplay I've written seven
12:11 screenplays so
12:13 I know the world well
12:15 um that one
12:17 and don't so so one of the tricks if you
12:19 haven't played with these tools a lot
12:21 don't just go in and go write me a
12:23 screenplay
12:24 it'll write you a shitty screenplay it
12:25 knows what they are
12:27 um but go in and start with your story
12:28 say hey I got a story and here's what it
12:31 is can you help me flesh out the outline
12:33 and let's use the save the cat story
12:36 framework or let's use you know Sid
12:39 fields or let's use Joseph Campbell's
12:41 Heroes Journey whatever the [ __ ] let's
12:43 take the basic story framework of Star
12:46 Wars but remove all the major plot
12:47 elements but keep the structure whatever
12:50 then have it fill in your outline then
12:52 have it give you describe some of the
12:54 characters and have it give you a
12:55 character sheet
12:58 then I have it write you the first page
13:02 of the movie
13:03 and just keep [ __ ] with the first
13:05 page of the movie until it gets
13:06 somewhere close to what you want and if
13:08 you're a good writer already
13:10 paste in three or four pages of a senior
13:13 particularly happy with
13:15 and tell it to mimic that style and it
13:18 will
13:19 and then you should have what I call
13:21 your Kevin McAllister moment
13:26 I had no idea did that oh yeah
13:32 um so yeah that's why I have these on
13:33 screen these are
13:36 we're lucky to be alive at a time when
13:38 we get to play with this [ __ ] they're
13:40 [ __ ] remarkable
13:42 now do they make mistakes do they
13:44 hallucinate are they wrong a lot of the
13:46 times do they write crap a lot of the
13:48 times yep
13:50 but have you ever had writers blocked
13:52 for say a week or three months or three
13:54 years not anymore
13:58 take that shitty Vapor of an idea that
14:00 you have that you're like I don't really
14:02 feel like writing it I'm not a very good
14:03 writer put it in there watch what
14:05 happens to it
14:07 within 10 minutes you'll your brain will
14:09 be on fire with creative ideas
14:11 so yeah they can write movie scripts
14:14 well no you can write a movie script
14:17 using that as your assistant
14:19 it's like having an idiot savant as your
14:21 assistant you have to give it lots of
14:23 Direction because it's a little dum-dum
14:25 but
14:27 it knows everything
14:34 you want to take the surprise ending of
14:37 The Sixth Sense
14:39 and combine that with the
14:43 Charming Narrative of Forrest Gump
14:47 it'll do it
14:51 hey all hey Joy
14:54 um new here can you explain please what
14:56 would you like me to explain who I am
14:59 why I act like this what the [ __ ] is
15:01 going on with my hair
15:03 I'll explain it all
15:06 so a marketing background will be an
15:08 asset yes uh Mama magic here's my
15:13 here's my thing
15:15 I think that
15:17 um
15:19 being older is an advantage so I think I
15:22 think we are entering a revenge of the
15:24 Gen X
15:26 phase
15:27 where
15:29 the more World experience you have the
15:31 kind of the better you are with these
15:33 things because they'll produce stuff and
15:34 you're like Ah that's crap well that's
15:36 good
15:37 and it's also revenge of the liberal
15:39 arts major so yeah having a marketing
15:42 perspective understanding storytelling
15:44 understanding audience understanding
15:47 um intent understanding what you're
15:49 trying to accomplish is brilliant with
15:51 this [ __ ] because these are just [ __ ]
15:53 tools
15:55 the art of these is understanding what
15:58 am I trying to accomplish and if you can
16:00 articulate that
16:02 then you can figure out how to talk to
16:04 this thing and it's more like prompt
16:06 psychology than prompt engineering I
16:09 think we should strike the term prompt
16:11 engineering from the vocabulary
16:13 I'm good with prompt crafting
16:16 um I saw on Twitter I don't know a week
16:18 or so ago someone said it's more like
16:20 prompt psychology this is a these are
16:23 called large language models large what
16:25 language models who uses language humans
16:29 these things basically were trained on
16:32 all of the text of the public internet
16:34 or as much as they could get but like a
16:36 [ __ ] ton of it
16:38 so so imagine for the last 30 35 40
16:42 years people have been dumping stuff on
16:44 this thing called the internet
16:47 in decades
16:50 of all the stuff it's all in there
16:55 I mean have you ever tried to like surf
16:57 Wikipedia
16:58 it's massive and that's like one little
17:01 ants ass of what's in here it's like all
17:04 the blogs all the websites
17:06 all the tweets which Elon Musk is pissed
17:10 off about
17:11 whatever they scraped it they can see it
17:14 all the universities all the research
17:16 papers all the programming languages all
17:18 the all the code
17:20 from those programming languages all of
17:22 it
17:24 so yeah so marketing background is good
17:27 because you go in there and you treat
17:29 that thing like a co-worker
17:32 you understand what you want you ask it
17:35 for it and then it gives you some
17:36 version of it when you get good with
17:38 these things you can get to like 80
17:41 percent of what you want like that
17:43 and then the last 20 takes you about the
17:45 same amount of time as you would have
17:47 written the first draft in the first
17:48 place if you want it to be good
17:53 yeah
17:54 you're too funny Kyle oh but looks
17:57 aren't everything
18:04 good morning from Finland thanks for
18:07 doing what you do hello young thank you
18:10 for the compliment I do appreciate that
18:12 it's late I got started very very late
18:14 tonight because uh I had an AI Salon
18:17 which by the way if you're interested in
18:19 the AI Salon here's the website so
18:23 Kyle what's the AI Salon you've never
18:25 talked about that before what is that
18:29 um it's a group of AI adventurers we
18:32 call us
18:35 that are on this [ __ ] adventure of
18:37 like what is this stuff how do we use it
18:40 what do we do what if I'm a copywriter
18:42 what if I'm I'm an artist what if I'm an
18:45 actor a writer
18:48 a screenwriter what do I do
18:50 come hang out with us we're trying to
18:52 figure it out
18:54 you know who has it figured out
18:58 no one
19:00 no one
19:02 I just sounded like Tom Papa no one
19:08 thanks for the heads up you're welcome I
19:09 don't know what that was in reference to
19:10 but you are welcome Zuckerberg announced
19:13 today about the rollout of llama version
19:14 2 thoughts big [ __ ] deal not big
19:17 [ __ ] deal like in big massive [ __ ]
19:21 deal
19:22 like when Obama signed the healthcare
19:24 thing and Joe Biden leaned into him and
19:26 said this is a big [ __ ] deal the The
19:28 Meta thing is a big [ __ ] deal
19:31 um so if you don't know
19:33 meta Facebook released llama2
19:38 um and they released it open source with
19:41 a commercial license so they released
19:43 their 7 billion 13 billion and 70
19:46 billion parameter models
19:48 and said unless you have more than 700
19:50 million active customers you're free to
19:53 use it with no royalties no compensation
19:55 no restrictions go
19:58 it's huge it's huge
20:02 because the llama's been out there for a
20:04 while
20:06 um but it was only for research purposes
20:08 so you couldn't use it for commercial
20:10 and then some Engineers figured out some
20:12 ways to Wrangle around that and come up
20:14 with semi-usable things but this is now
20:17 out there for everyone to use and then
20:19 and then bard
20:22 like I I can't believe it wasn't a
20:25 bigger deal they completely replaced the
20:27 the core engine sitting underneath bard
20:30 Gemini it's called with their new
20:33 technology called Gemini so I guess they
20:35 flushed Palm 2 down the toilet or that's
20:37 going to be used for something else but
20:39 Gemini is this multimodal engine sitting
20:42 under that Claude that's
20:45 um anthropics gpt4 competitor
20:48 it's just like this stuff is it's
20:50 hopping it's happened I heard tonight I
20:53 got to confirm this
20:54 someone said this but I don't I I don't
20:57 trust it someone said it at the salon
20:59 tonight that gpt4 is apparently HIPAA
21:02 compliant which if you're in healthcare
21:04 and Pharma we do a lot of work in
21:05 healthcare and Pharma that's a big
21:07 [ __ ] deal so I've got to confirm or
21:09 deny that um but that's what I heard at
21:11 the salon tonight so that's just a rumor
21:14 don't follow rumors you know confirm
21:16 them yourself
21:19 go ask chat gbt when it tells you yes
21:21 it's HIPAA compliant don't believe it
21:23 because it lies
21:24 yeah all right started with pi Ai and
21:29 finished with Claude and Adobe for a
21:31 presentation and step Z
21:34 that's very cool Silver Fox started with
21:36 pie finished with Claude
21:39 and Adobe for presentation oh Adobe like
21:41 Firefly and generative generative fill
21:44 that's cool super cool silver fox
21:50 I'm thirsty I'm a little punched it was
21:52 almost 100 degrees here in Denver today
21:55 legal Tech is a huge use case yes yes
22:00 ever like right now I just I mean these
22:04 things are so
22:06 staggeringly capable at almost anything
22:09 you throw at them that I feel like
22:11 everything right now is a huge use case
22:14 made a 13-day social plan for getting
22:17 tourists back to flooded Vermont with
22:19 pie pie Ai and Claude this morning wow
22:22 that's awesome
22:23 that's awesome tell us more about that
22:24 and tell us how it how it um how it
22:27 works
22:28 and what it did and like what it worked
22:30 what what didn't that's that's really
22:32 awesome
22:33 Tobias can't figure out how to install
22:35 llama just code files on GitHub more
22:37 research needed okay so here's a trick
22:42 whether you're on a Mac or a PC go to
22:44 GitHub go to the repository where
22:46 llama's there well a couple of things go
22:49 to Reddit
22:51 and and I would assume at this point
22:53 there's a llama subreddit and if there's
22:55 not I just can't imagine that there's
22:57 not but just go to Reddit and search for
23:00 open source large language models see if
23:02 you can find a llama subreddit there's
23:04 probably even a llama 2 subreddit at
23:06 this point
23:09 and then just find people talking about
23:11 oh here's the thing to install llama on
23:13 your Mac or on this or on that or here's
23:15 the GitHub okay so that's a way to find
23:17 some things
23:19 um
23:21 go to the readme file of your GitHub
23:26 copy all of the readme file
23:30 and go to claude.ai that has a hundred
23:33 thousand token context window and paste
23:36 that [ __ ] in there paste the
23:38 readme file from that GitHub well and so
23:41 do this say
23:43 based based on on this documentation
23:47 shift return for a line break paste
23:50 shift return
23:52 um
23:55 write me a step-by-step tutorial on how
23:59 to get this running on my in my case
24:02 MacBook Pro whatever in your case PC or
24:05 Mac
24:06 and don't move on to the next step until
24:09 I've told you until I've said okay to to
24:12 the the present step
24:14 and then it will it will give you the
24:16 first step okay install this
24:18 and then you're like I don't know how to
24:20 install that and it'll go okay here's
24:22 the step-by-step step instructions for
24:24 how to install that and then you go try
24:26 to install it and if you can get it
24:27 there then you go to the next step then
24:30 you go to the next one then you yeah
24:31 exactly Tobias I know that was what
24:33 because I was struggling it took me to
24:35 be you know to be fair
24:37 um or not fair to be to be transparent
24:40 when I tried to install stable diffusion
24:42 I failed nine times for exactly that
24:45 reason I couldn't I couldn't get um
24:48 Google collab working like I could get
24:50 it working and I could put hit the play
24:53 buttons but then when it would error out
24:55 I didn't know what the [ __ ] to do
24:57 and then I'm looking around on the on
24:59 the Twitter and someone's like oh did
25:02 you know that you can have
25:04 GPT read your readme file on GitHub and
25:08 just turn that [ __ ] on it's like [ __ ]
25:12 it's amazing
25:13 hello Daisy sunshine how are you welcome
25:16 welcome
25:19 um oh looks we had a bunch of people in
25:21 here and they all went away apparently
25:23 apparently uh it's so late for people
25:26 because I started at like 10 30.
25:31 ah if you've got uh questions about uh
25:35 generative AI pop them in the comments
25:37 below
25:38 let's go look at something what's worth
25:41 looking at there's so much
25:48 I'll do a Bard thing
25:50 so
25:56 so Silver Fox pie gave me bones and ran
25:59 out of space so moved process to Claude
26:01 then did summary and action items that's
26:04 super cool that's super cool
26:07 man
26:08 I hope to I hope the flooding's done
26:12 that's screwed up
26:16 um
26:17 so on bard
26:20 um
26:20 it does all sorts of stuff like Chachi
26:23 BT does like you can say oh here I'll do
26:25 wait this thing
26:30 write me a python script
26:33 of
26:35 a tic-tac-toe game you can never win
26:43 I don't know if it's possible to write a
26:45 tic-tac-toe game you can never win
26:47 well I guess you could if you get the
26:49 third thing in a row then you could
26:51 start the game over
26:53 here's a python script of a tic-tac-toe
26:55 game you can never win
26:57 let's see
26:59 it's designed to ensure the computer
27:00 always wins the check when functions so
27:05 anyway so it writes code does all that
27:06 but but then check this out so
27:09 um what are the top
27:12 um
27:14 minimalist shoe
27:16 companies
27:20 oh sleep at that
27:28 and so what what Bard is doing now that
27:31 no one else is doing is if it's if it's
27:34 relevant to your
27:36 results it's giving you previews of the
27:39 websites
27:40 and descriptions of the websites
27:44 in context of what you're asking for
27:47 this seems like a very big deal to me I
27:49 mean
27:51 this is
27:54 this is counter to their business model
27:56 now
27:58 you know I assume they've got something
28:00 in here that if I click on this maybe
28:01 you know
28:03 the pay-per-click goes to someone but
28:06 maybe not I don't know maybe they're
28:07 just experimenting with with this crap
28:09 right now
28:11 um but that's that's just wild isn't it
28:16 um buffering Max Headroom that's because
28:17 I switched screens probably
28:23 oh yeah
28:28 yeah Daisy I think you ran everybody off
28:31 they were all in here everybody was in
28:32 here there were hundreds thousands of
28:35 people were in here
28:38 and then Daisy said hello to me and they
28:40 all left
28:42 how was your 8 A.M meeting my 8 A.M
28:44 meeting was awesome oh okay here's a you
28:49 can make money with Chacha bite so for
28:52 the past
28:53 six months
28:57 my co-founder and I have been talking to
29:00 a very large pharmaceutical company
29:03 and pitching them Story vine and
29:06 pitching them some semi-new things with
29:10 storyvine and then pitching them a lot
29:12 of AI where we're rolling AI into the
29:15 storyline product and talking about
29:17 doing Omni cam omnichannel content
29:20 generation and I've been doing a lot of
29:23 hand waving
29:24 for the past six months and the call
29:27 this morning
29:29 um looks very promising that they want
29:31 to do some sort of pilot and what's
29:33 fascinating
29:34 is the pilots that they want to work on
29:36 with us are the AI stuff
29:39 so it's very possible that that this
29:42 large client
29:43 um is going to give us money
29:46 to go build some of the things I've been
29:49 designing which is super exciting
29:51 because
29:53 if you build it for a big Pharma company
29:55 it's got to be good
29:57 and so if we build it for them we will
30:00 retain the intellectual property and
30:03 we'll be able to go sell that to other
30:04 people
30:05 and then we're gonna have more money
30:07 and you can make money
30:11 [Laughter]
30:14 and all of the all of the prototypes
30:17 that I've been uh showing to people
30:19 we're all done with with GPT playground
30:22 and chat GPT
30:24 so if indeed I make money
30:27 going to build some of this stuff then I
30:30 will have made money with chat gbt but
30:32 it was a six month pitch
30:35 and probably you know 15 or 20
30:37 conversations
30:40 buffering Max Headroom
30:42 uh uh a Marley 95 I think we need a
30:45 better text I think we need better text
30:47 to video models yes we do but here's the
30:50 deal
30:51 go look at I assume you can Google this
30:54 somewhere probably find it on Twitter go
30:56 look at
30:57 where image generation was in like April
31:02 of 2022
31:04 it was laughably bad
31:07 stable diffusion and mid-journey had
31:09 just come out
31:10 and you put in text and it sort of
31:12 looked like a dog
31:13 and that woman's face it almost looks
31:17 human and sort of is it is it a blob of
31:20 clay or is that someone's face
31:23 and a year later they were brilliant and
31:25 so video is going to take the same
31:27 trajectory and it's going to get better
31:29 much better much faster in fact there's
31:33 the recent interview with the CEO of
31:36 stability AI who he's a bit of a uh if
31:40 you think I'm a tech Optimist go listen
31:42 to the the CEO of stability please
31:47 he's like everything
31:52 um but but he said that the internal
31:55 model that they're working with at
31:56 stability
31:58 like right now if you go make a image in
32:00 stable diffusion it takes whatever it
32:01 takes 10 seconds to render one
32:04 the internal model that they're working
32:06 with at stability is rendering 60 images
32:09 a second
32:12 30 frames a second is video 60 frames a
32:14 second is video games
32:17 so think about the implications of that
32:20 dynamically generating 60 frames a
32:23 second
32:28 as good as stable diffusion is right now
32:31 which with the Laura's and the dream
32:33 booth and the fine tuning and the
32:35 control net it's gonna get
32:39 Mighty powerful Mighty fast so just have
32:42 a little have a little uh a little
32:44 patience and I mean I assume I assume
32:46 you've played with Runway
32:49 um gen 1 and gen 2.
32:51 um I agree they're kind of you know just
32:54 janky
32:56 well that's okay they'll get better go
32:59 look at Wonder Studio you want to have
33:00 your mind blown off your head go look at
33:02 Wonder Studio
33:04 upload a little blob of video Let It
33:06 process it for 30 minutes and it
33:08 replaces the human being in the scene
33:10 with this like a CGI robot
33:13 and you can export the alpha channel the
33:16 depth map
33:18 and and the you know the green screen
33:21 animation and the full video
33:25 I showed it to a filmmaker tonight A
33:27 filmmaker came to the uh to the AI Salon
33:29 tonight
33:30 I mean I was like oh have you seen
33:31 Wonder studio and I showed it to him he
33:33 was like
33:36 because that's very impressive
33:41 like I know
33:43 like this is a guy that puts actors in
33:45 those suits with the ping pong balls on
33:46 the hands
33:47 he's like you can just shoot video on a
33:49 cell phone and get that
33:51 me
33:55 hey everybody hey Mr cgo thank you very
33:58 much that's Chief generative officer to
34:00 you Ma'am
34:01 yeah I am the chief generative officer
34:03 of an organization called content
34:05 evolution
34:07 and if you're impressed by that title
34:10 it's just made up
34:14 the first one with that title
34:16 uh
34:19 uh all right let's see it's going down
34:23 ask some more questions folks
34:26 being shy just because it's [ __ ]
34:28 midnight almost in in Denver
34:32 if anyone's on the east coast watching
34:33 me right now you really need to take a
34:36 nap
34:37 I feel like search engines and SEO in
34:40 general won't be as important I I so I
34:44 firmly that's I was in that camp for a
34:47 while but then
34:51 this thing where where
34:53 Bard is when I asked about minimalist
34:57 running shoes and my buddy Steve he owns
35:00 this company zero shoes
35:02 we actually met filming an episode of
35:05 America's Most Wanted
35:07 back in 1988 or 1889
35:11 and we filmed it in Connecticut he
35:12 played the killer I played his best
35:14 friend
35:15 and it turned out we lived like a block
35:17 from each other in Manhattan and so we
35:19 became good friends and we're still good
35:21 friends to this day so anyway
35:23 he owns this company and it was founded
35:26 in 2009 and
35:29 like he's been doing like you know SEO
35:31 optimized content and advertising for
35:35 like 13 years
35:37 and so
35:39 the site you know I'm sure it ranks at
35:42 the top of everything you know every
35:44 possible thing so when I just asked
35:47 um Bard for the top companies
35:50 um vibro Barefoot was first and zero
35:53 shoes was second
35:56 so
35:58 so I
36:00 I don't know what's going to happen with
36:02 SEO right the
36:05 I think you're right I think
36:07 search has to fundamentally change
36:09 because
36:11 like my experience has been
36:14 if I know that chat GPT is going to take
36:17 30 seconds to return a result
36:20 and I know that Google would return a
36:22 result in two seconds I'll wait for the
36:25 30 seconds for chat GPT just so I don't
36:28 have to dig through the [ __ ] SEO
36:30 injected crap
36:32 that you have to deal with at Google
36:34 so I don't know how Google
36:36 makes this transition
36:39 um it's interesting looking at how
36:41 Bing's doing it bing is sort of they're
36:43 sort of doing this this thing where
36:45 they're they're you know injecting
36:47 hyperlinks
36:49 um so it's not really clear what's paid
36:51 what's not they're not doing it as slick
36:53 as Bart is right now
36:56 um
36:57 yeah so so I I agree with you uh Marley
37:01 that that um
37:04 that SEO is going to change dramatically
37:06 and you know if I ran an SEO agency
37:09 right now I would be
37:12 um scrambling
37:14 scrambling to upskill
37:17 all of my uh all of my employees
37:21 to be full in on this you know I if I
37:24 had an SEO agency right now I would say
37:29 three weeks from now we're gonna have a
37:31 competition and whoever pitches me the
37:34 best idea
37:35 to put our agency out of business
37:39 we'll do that idea
37:41 or something like that like I would like
37:43 I would I would be radically pushing to
37:46 completely disrupt the whole model of of
37:49 SC of agencies in general
37:52 if if they don't aggressively start
37:55 disrupting themselves they're in for a
37:57 rude [ __ ] Awakening
37:59 and I'm not just I ain't just
38:01 speculating that out of my butt that's
38:04 some of my clients telling me that in
38:06 confidence
38:08 not for nothing but
38:13 if our agencies don't start delivering
38:15 [ __ ] faster at a higher quality and a
38:17 lower price they're gone
38:20 because the clients know that that's
38:22 possible now they didn't know it was
38:25 possible before they know it now
38:28 so what are those agencies gonna do
38:35 other than a lot of cocaine
38:38 I think companies will have to pay to
38:40 get featured on AI suggestions like what
38:43 Bard just did
38:45 I there's a part of me that feels yeah
38:48 it's gonna have to go somewhere in that
38:50 direction and
38:53 if they over inject so here's the
38:56 problem so
38:58 a week ago this this list had three on
39:00 it I could easily argue right now it
39:03 needs to have a sixth one now it has
39:05 five it needs to have a sixth one and
39:07 probably a seventh one
39:09 and probably within a month that list
39:12 should really be 10 companies Long
39:15 so
39:17 there's going to be a lot of competition
39:18 in this space so if the Googles of the
39:21 world start over injecting
39:25 add answers that are less than relevant
39:30 people aren't going to use them they're
39:31 going to run away because the minute
39:33 Google starts to feel like Google search
39:37 where it's just like it's just you know
39:39 like the signal to noise ratio on on
39:41 Google search because of all the SEO
39:43 optimization is just horrible you know
39:46 you have to you have to click through 30
39:48 different sites and 27 of them are just
39:51 [ __ ] SEO concoctions probably made
39:54 with AI from
39:56 six years ago you know
39:59 look at it all right
40:04 I'm 75 and love watching you teaching me
40:08 got me learning yes Daisy Sunshine
40:11 welcome
40:14 I I 75 yeah so you're you're probably
40:17 just Boomer I'm I'm Gen X I'm sort of am
40:20 i mid-gen x i don't know doesn't really
40:23 [ __ ] matter but welcome so listen the
40:26 fact that you're learning this the whole
40:27 point of what I'm doing here is just go
40:30 start playing with this stuff you don't
40:32 have to be technical
40:34 you can do recipes and gardening and
40:36 teeth have it teach you
40:38 Mesopotamian pot making it doesn't
40:40 [ __ ] matter it knows it all
40:43 so just go start talking the one thing
40:46 that's not up here that I would
40:47 encourage you to play with if you
40:48 haven't is go get the PI app pi
40:52 or go to pi.ai or heypi.com
40:56 that's inflections uh personal assistant
41:01 is what they're calling it but the IOS
41:04 app in particular you can just put it in
41:06 voice mode and just start talking to it
41:08 so just like you talked to Alexa and you
41:11 know the iPhone one I don't want to
41:13 trigger it here
41:14 um you just start talking to it like
41:17 that except it knows everything and it's
41:19 connected to the internet and it's
41:20 really friendly
41:23 and it's and it's a wild experience
41:25 because we're so used to those chat Bots
41:28 sucking
41:30 that when you have one that doesn't
41:34 it's strange
41:36 I mean it's it's not quite like you're
41:38 in a movie because there's there's still
41:40 this weird delay there's like a
41:43 three quarters of a second delay like
41:45 it's just enough that you're like
41:48 and
41:51 but once you sort of like mentally
41:54 absorb that and sort of recognize
41:56 that'll get better in the future
41:59 it's pretty trippy
42:01 it's pretty trippy
42:08 when AI brings everyone at the same line
42:11 who needs money in oh wait when everyone
42:14 brings every when AI brings everyone at
42:17 the same line oh up to the same line who
42:20 needs money anymore big changes coming
42:22 yeah this is we we are entering an age
42:25 of prosperity that's going to be
42:27 preceded by an age of
42:30 Panic disruption
42:33 so so yeah like I
42:36 so so kind of this is what it I I do not
42:40 have a degree in economics so I don't
42:41 understand economies other than at the
42:44 most broad level
42:46 you have this many people working this
42:48 many hours
42:49 to create this much economy
42:52 you now have tools that for all intents
42:55 and purposes are going to give us but
42:57 somewhere between 25 and 50 percent
43:00 productivity increases right
43:02 so with this amount of people working
43:05 you can either
43:08 keep your GDP the same and reduce the
43:11 number of people working so this amount
43:13 of people are not working or you
43:16 increase the prosperity
43:18 and if you've got a government that
43:20 doesn't have their head up their ass
43:21 say one not in America
43:25 then you can take a big chunk of this
43:27 and
43:29 pump more Prosperity into the system
43:33 um
43:35 so yeah
43:36 what does that look like I don't know
43:39 but every oh I like I can't think of a
43:43 job that's not going to be somehow
43:45 directly impacted by by these
43:47 Technologies so the disruption is just
43:50 going to be insane and it's going to be
43:53 scary and some people are not going to
43:54 adapt and others people are and the
43:56 whole purpose of this channel is get in
43:59 front of this [ __ ] so that you don't get
44:02 blindsided
44:04 um do me a favor we'll do I'll do
44:05 another sort of 15 minutes here hit the
44:07 hit the like button and if you would
44:09 share the live so we can just see if we
44:11 can get some more people in here
44:14 um wishing you all a good night I know
44:16 it's late I gotta go to bed too
44:19 Night Joy
44:21 I'm in Virginia ha ha cool
44:24 which killer I don't know what I was
44:26 referring to
44:28 all the GPT sites
44:31 charging for the use oh are they no so
44:37 um chat gbt you can use 3.5 for free you
44:41 have to pay for gpt4 but that's the best
44:44 GPT for it you also get code interpreter
44:46 which is insanely powerful
44:48 Bing is free
44:50 it's gpt4 but it's sort of Microsoft's
44:53 flavor of it so it's a little tame
44:57 um but it can read images and it can
44:59 create images and it's free so cool
45:02 claude.a AI is free
45:05 it's got a hundred thousand token
45:07 context window which doesn't mean [ __ ]
45:10 to human beings
45:12 but it just thinks 75 000 words you can
45:15 put a 75 000 word prompt in claude.ai
45:19 so like uh what is it uh what's what's
45:23 the the novel
45:25 I don't know
45:26 like a 50 000 word novel you can just
45:28 pop that in there just paste it copy and
45:31 paste your whole novel or upload up to
45:33 five PDFs up to 10 megabytes each and
45:37 then it'll just deal with it so that's
45:40 free pose free it's got three models in
45:46 there gpt4 and the and the
45:49 two clod models that are that are
45:52 subscription only but all the rest of
45:54 it's free and it's got free Bots and
45:56 then bard.google.com is free most of
45:58 them are free
45:59 free pray pray
46:03 I wonder what he's selling
46:06 what are your qualifications Mr AI lab
46:09 learning huh
46:11 what have you done here
46:14 you have a degree in this I don't think
46:16 so so you're obviously selling something
46:18 or you wouldn't be here just doesn't
46:20 make sense
46:22 is that a chemtrail anyway
46:25 just went all uh Jim Carrey there for a
46:28 minute sorry sorry about that
46:35 yes we don't need to go through all
46:37 sites to get the info we are looking for
46:41 yeah exactly exactly and that's Google's
46:44 business model
46:46 Google's business model
46:49 the thing you want it ain't
46:52 they're business models to deliver you
46:54 40 things you have to click on to
46:57 eventually find the thing you want
46:59 and I guarantee you they've got the data
47:02 science to know exactly where to place
47:04 the thing that's of value to you so that
47:07 you click on eight different things
47:09 before you get to it I guarantee you
47:11 they have that down to a [ __ ] science
47:17 ask yourself
47:20 if these guys if Google invented the
47:23 Transformer which is the thing that
47:25 empowers all of these tools
47:29 why weren't they first out of the gate
47:33 why did they wait until chat GPT went to
47:36 100 million users in six weeks before
47:38 they panicked actually that's not fair
47:40 they panicked after five days when it
47:43 went to a million users
47:46 that's when they they it was declared
47:48 there was a red alert at Google
47:51 which is ah oh [ __ ] boys the uh the uh
47:55 that thing that's the thing we invented
47:57 did I where
48:00 is it hot in here
48:09 because they knew it would [ __ ] up their
48:12 business
48:15 so that's what they're dealing with
48:17 but they're Google they got a lot of
48:20 smart people they got a lot of money and
48:22 Gemini the model that's sitting under
48:24 that it doesn't suck
48:27 this wasn't on my list a week ago
48:29 because it it was laughably bad it ain't
48:31 anymore
48:33 amazing
48:35 so website visitors will drop
48:37 dramatically yeah I I think you know
48:41 I I think what ends up happening with
48:44 this with this new kind of search
48:47 is that search engines don't go away
48:49 there there are some times where you're
48:51 just like I need to Google it and and
48:53 you just you kind of know how Google
48:54 works and you know it's just going to
48:56 pop up the weather it's going to give
48:58 you the answer to that math problem or
49:00 whatever it is right
49:01 or or you need a list of these specific
49:04 sites and you know how Google delivers
49:07 them
49:08 um so I don't think it goes away but I
49:10 think it drops dramatically in
49:12 importance and and what Google's gonna
49:14 have to toy with
49:16 is how do they take this new mode of I
49:19 just want the answer rather than I want
49:21 to [ __ ]
49:22 spend the next hour looking for it like
49:25 a goddamn scavenger hunt scavenger
49:27 torture hunt
49:30 like they have a [ __ ] a shitty product
49:32 right
49:33 Because the Internet is so complicated
49:37 and before you add large language models
49:39 you had to use your actual human brain
49:41 to like click on a thing
49:44 go out to that site use your dang eyes
49:48 or your dang ears and listen to it and
49:50 read the site and go yep that's oh no no
49:53 that's not what I want click back to the
49:55 click innate
49:58 that's a shitty product that's a shitty
50:00 experience now
50:02 was it the absolute
50:05 best thing we had for two decades yes no
50:09 one came close
50:11 Alta Vista was the search engine to beat
50:14 all search engines and Google came up
50:16 with their [ __ ] algorithm and just
50:19 everything else just went bye-bye
50:23 and they had it for two decades and
50:25 literally overnight
50:28 Chachi PT
50:30 blew that up so
50:34 um
50:36 so I think search Still Remains I think
50:38 websites still remain I think the
50:39 importance of them goes down
50:41 dramatically
50:46 I think we're entering I actually just
50:48 wrote an article if you go to my
50:50 LinkedIn Kyle Shannon on LinkedIn
50:52 I wrote an article today
50:55 um about I think we're at the dawn of of
50:59 of an era of hyper personalization
51:03 where rather than me going to a website
51:06 that someone else designed
51:08 the website is going to be dynamically
51:11 generated based on what I need in that
51:13 moment
51:15 and and that's kind of what these things
51:17 already are
51:18 right
51:19 because they've been trained on all of
51:21 the internet and you're like gee
51:24 I wonder what it would look like if
51:27 Einstein and Stephen Hawking had a
51:30 conversation
51:32 you know
51:33 what would they say to each other
51:36 and it goes out and it finds everything
51:39 that's been written about them
51:41 and it imagines it in its probabilistic
51:45 way and it shows you that and that's
51:48 kind of like a dynamically generated
51:50 website now right now it's just text but
51:53 it's not just text right
51:54 it'll deliver markup it'll deliver code
51:58 if you watch code interpreter do what it
52:00 does it's writing and executing code and
52:03 generating images and handing them back
52:05 to you on a platter
52:10 so isn't that what a lot of websites do
52:14 yeah you mean ones that have been around
52:16 for a decade that they spent a decade
52:18 figuring out how to write code to hand
52:21 you charts on a platter
52:23 and so it's just dynamically doing that
52:25 right now yeah it is
52:28 so code interpreter
52:30 and the plugins start to be a glimpse
52:33 into
52:35 rather than giving it a prompt and
52:37 telling it exactly what I want I kind of
52:39 give it a goal
52:41 and have a general vague idea of what I
52:44 want and then it goes gotcha boss I'm on
52:48 it
52:49 hey let me write up some code to go try
52:51 this oh that didn't work she sucks let
52:53 me try again
52:56 that seemed to work let's go look at
52:57 what I created hmm
53:00 there seems to be some missing data here
53:01 let me write some code to fix that
53:06 there it's fixed
53:08 you said you wanted to it to be
53:10 interesting visually let me go ahead and
53:12 make you a spider graph
53:16 here's your spider graph
53:19 I mean this is [ __ ] nuts
53:22 I'm working on something sort of an idea
53:25 rather than a business to tackle the SEO
53:27 matter oh that's cool I'll pitch it to
53:29 you awesome would love to hear it sounds
53:31 cool I'm 59 and this is all Greek to me
53:35 well I'm 58 and this is all Latin to me
53:41 um but I listen every night hoping to
53:43 latch onto something so listen press dog
53:48 just
53:49 pick one actually you know what Prez dog
53:52 if you've got an iPhone even if you
53:54 don't if you don't if you don't have an
53:55 iPhone
53:57 go to pie
53:59 dot a i p i dot a I or if you've got an
54:03 iPhone Go download the PI app pi
54:06 from inflection
54:09 put it into voice mode and just start
54:11 talking to it
54:15 don't put pressure on yourself that you
54:17 have to learn AI
54:19 you don't we don't
54:23 I think we need to understand I I think
54:26 the sooner we can understand the tools
54:27 and what they do the the bigger
54:30 Advantage we have but we don't need to
54:32 learn AI because these tools are going
54:33 to be [ __ ] everywhere
54:36 so go get the PI app
54:38 and just tomorrow like I did it in my
54:40 car like this was mind-blowing to me
54:42 like I just like I don't know it was
54:45 like on Twitter or something like that
54:47 someone posted a video of them talking
54:48 to this app and I'm like oh let me go
54:50 try this thing it's probably [ __ ]
54:51 stupid right that's where I start all of
54:54 my software transactions is this is
54:56 probably a piece of [ __ ] and then if it
54:58 proves otherwise then I'm like oh okay
55:00 fine well it but if I push it a little
55:03 hard it's probably then it's a piece of
55:04 [ __ ] oh fine you know that was where
55:06 chat GPT the more I pushed it the better
55:08 it got
55:09 and I was like
55:11 everything's different
55:14 so get Pi installed
55:16 either hook it up to your car
55:18 or just if you're like sitting out on
55:20 the porch drinking coffee
55:22 just start talking to it
55:25 ask it about
55:26 any news that happened asking about
55:29 local sports teams
55:31 um
55:33 asking about
55:35 some blight you have in your garden ask
55:37 it about does it have a suggestion for a
55:40 recipe for dinner tonight
55:42 um
55:43 I'll give you a quiz about
55:46 something that interests you
55:48 um
55:49 have it come up with a game ask it to be
55:52 a dungeon master Play Play It doesn't
55:54 [ __ ] matter
55:56 but just like talk with it
55:59 for 30 minutes
56:01 just let your let the conversation
56:03 ramble and
56:06 I think I think
56:09 the idea of play right now is actually
56:12 really important
56:13 because as important as all this [ __ ] is
56:15 these tools are very very primitive for
56:18 from what they're going to be
56:20 so if you want to play with them early
56:23 just start playing and I think I think
56:25 Pi is there's something really playful
56:29 about how how that voice has been
56:31 trained to interact
56:33 like like I asked it
56:35 I asked it to create a game for me how
56:38 how well I understood Transformers or
56:40 something like that
56:42 and it started out and said oh that's an
56:45 excellent idea I can definitely do that
56:47 you know why don't you tell me a little
56:48 bit about what you already know and I
56:50 told it a couple of things like oh
56:52 that's very Advanced you're very white
56:56 had no idea I was talking to someone
56:58 that was you know knew this stuff and
57:01 then it told me a little something and I
57:03 said I said well I don't understand this
57:05 part that's an excellent question and
57:07 it's like I'm like oh well thank you I
57:10 that's an excellent question if I do say
57:12 so myself so just go play with that and
57:15 just if you can get yourself into
57:19 into a mode where you have fun with this
57:22 stuff and take the pressure off feeling
57:23 like you have to learn it
57:26 because that's there's nothing shittier
57:28 like you can tell by how not in shape I
57:32 am you know every time I tell myself I
57:34 gotta work out my my you know visceral
57:40 reptile brain reaction is go [ __ ]
57:42 yourself
57:45 and and so if if it's feeling like that
57:47 then just find a way to make it play and
57:49 I think pie could be a good a good tool
57:51 for that
57:54 all right Daisy's on it I don't know
57:56 what I told her to do but awesome well
57:59 your storyline title may be made up now
58:01 my storyline title is founder and CEO
58:04 that's real it's my
58:06 um my content Evolution title Chief
58:08 generative officer but it's fire and it
58:11 should be under your pick on the website
58:12 no it's it's a chief generative officers
58:15 is uh
58:17 uh content Evolution but thank you
58:20 that's funny
58:21 yeah I know I liked it when when we came
58:24 up with it I'm like oh that's pretty
58:25 good
58:26 because generative can mean many things
58:29 it could mean generative AI it could
58:31 mean generating Revenue as in you can
58:34 make money
58:35 [Music]
58:37 foreign
58:38 all right
58:40 what do you think of mid-journey mid
58:42 journey is the [ __ ] best
58:45 mid journey is by far the best of the
58:47 image generation tools
58:50 here's some caveats
58:53 it still only works in Discord
58:56 which is stupid
58:59 they need to build a [ __ ] website
59:01 there's only 11 of them there apparently
59:03 like the the quality of that tool for
59:05 the
59:07 relatively zero amount of people that
59:09 work there's pretty amazing I would
59:11 imagine they've got more now but
59:12 whatever
59:14 um it's also been trained up on lots of
59:17 copywritten images so both it and stable
59:20 diffusion
59:21 and and Dolly to a degree although did
59:24 you hear open AI just cut a deal with
59:26 Shutterstock
59:28 so a lot of these tools have been
59:29 trained up on copywritten
59:32 information and we live in a litigious
59:34 Society
59:36 and so I think there's gonna be a ton of
59:37 lawsuits so I I think a risk with
59:40 mid-journey is that it gets you know
59:42 sued off the face of the planet but
59:43 we'll see
59:45 um stable diffusion
59:47 has a similar issue but it's been open
59:49 sourced so because it's been open
59:51 sourced it's by far the most flexible of
59:54 the tools and it's quickly catching up
59:55 to Mid journey in terms of quality
59:57 and leonardo.ai
1:00:00 is a web front end for stable diffusion
1:00:04 that's really good and they just
1:00:06 launched version three and it's got all
1:00:08 sorts of prompt assisting and you can
1:00:10 train your own images and
1:00:12 the the one downside of Leonardo is you
1:00:14 kind of have to do [ __ ] in public which
1:00:16 which I don't like but it's a really
1:00:17 good site and you don't have to learn
1:00:18 anything technically
1:00:21 um Dolly which is the open AI image
1:00:24 generation tool that's the one that's
1:00:26 behind Bing is by far the shittiest
1:00:28 but it's built into Bing so you can just
1:00:30 go to Bing and say make me an image and
1:00:32 it makes images and you know they're
1:00:34 sometimes okay
1:00:35 and then the final one on the image
1:00:37 Generation stuff is Adobe Firefly and
1:00:40 generative fill within Photoshop
1:00:43 um
1:00:44 they're good but not great some of their
1:00:47 interface stuff is better than the other
1:00:49 tools like just the ability to click on
1:00:51 different styles and it just updates the
1:00:53 images
1:00:54 um they've got a
1:00:57 like a text thing where you can make
1:00:59 fancy text like you can say fuzzy text
1:01:01 of the word fuzzy and then the letters
1:01:03 become fuzzy in particular fonts
1:01:07 um
1:01:07 the big deal with Adobe is that they've
1:01:10 trained their model on public domain and
1:01:15 um Adobe stock images that they own all
1:01:17 the rights to
1:01:18 and so they're very aware of artist
1:01:21 rights they're going to allow artists to
1:01:23 opt out of being trained they're going
1:01:24 to allow artistops into being trained
1:01:27 they're going to figure out attribution
1:01:28 and they're going to figure out
1:01:29 compensation so
1:01:31 I I'm
1:01:34 um
1:01:35 I'm a big advocate for Adobe right now
1:01:37 the fact that they're moved as fast as
1:01:39 they have and they're this thoughtful
1:01:41 about the major issues that have the
1:01:44 entire art community [ __ ] pissed off
1:01:47 um and they're gonna win them back over
1:01:49 I think that's brilliant but
1:01:52 back to your original question mid
1:01:54 Journey still makes the best images
1:01:57 hey I took your advice and I went live
1:01:59 chatting with AI I think people liked it
1:02:02 thanks that's awesome man
1:02:05 p-i-p-a-p-i Pi dot a i and me or BFFs
1:02:09 now that's awesome that's really cool
1:02:14 um
1:02:15 oh you played a friend of a killer oh
1:02:17 what was the question what killer it was
1:02:19 Adam
1:02:20 forget what his name was he was so he
1:02:23 was some dude that uh shot a guy in the
1:02:26 back my my line on the America's Most
1:02:29 Wanted episode was
1:02:30 yeah about this gun thing
1:02:34 that was the scene before he shoots the
1:02:36 guy in the back
1:02:38 we're we're all a little worried about
1:02:39 this gun thing
1:02:42 and that guy was on the lamb for like 25
1:02:44 years and it was funny Steve got
1:02:46 recognized in airports and stuff like
1:02:47 that
1:02:49 aren't you that killer
1:02:52 he's like no I was just on the show
1:02:55 uh
1:03:00 conversed with Poe for the first time
1:03:02 today and it created a two-week low-carb
1:03:05 meal plan with a shopping list awesome
1:03:06 the bear see this is that see you got to
1:03:09 start with this [ __ ] you start with this
1:03:11 [ __ ] and then you're like oh that's
1:03:12 pretty good and then it's just good you
1:03:14 go deeper and deeper
1:03:18 it's just crazy
1:03:21 I'm going to get lunch thanks Kyle see
1:03:23 you next time peace out apocrypha
1:03:27 New York Times article today speaking on
1:03:29 the idea that AI may be the end of
1:03:31 Wikipedia
1:03:32 that's interesting
1:03:35 yeah yeah I can see that because
1:03:45 because these large language models are
1:03:48 trained on all of the information and
1:03:50 Wikipedia is a bunch of
1:03:54 things written by people who have
1:03:56 limited information even if it's like
1:03:58 anal obsessed I'm gonna spend three
1:04:01 months researching
1:04:03 this Wikipedia post
1:04:07 and they're brilliant at it
1:04:09 they're not going to be as comprehensive
1:04:10 as this thing being trained on all the
1:04:13 [ __ ]
1:04:14 and being able to dynamically pull that
1:04:16 stuff together
1:04:18 that's really interesting yeah I didn't
1:04:21 read that I think that's uh
1:04:30 it's just it like
1:04:33 the depth of what's gonna change is
1:04:37 staggering
1:04:40 like things that I mean could you
1:04:42 [ __ ] imagined
1:04:44 put yourself a year ago last October
1:04:48 could you have imagined that anyone
1:04:51 would not Google on their heels in the
1:04:54 search game
1:04:56 and and
1:04:57 that they would do it in like a month
1:05:01 six week period
1:05:06 no
1:05:08 and that's what happened and so like
1:05:10 that's the first Canary in the coal mine
1:05:13 is
1:05:14 what do you mean this thing can do that
1:05:17 what's that mean for Google that was the
1:05:19 first thing everyone said what does this
1:05:21 mean for Google
1:05:22 uh
1:05:24 any good
1:05:28 it's just amazing all those sites listed
1:05:31 in order of preference for you
1:05:33 um
1:05:42 I'd probably flip Bing and Claude but
1:05:44 for the most part yeah
1:05:47 and quite frankly Google
1:05:50 is
1:05:52 I see
1:05:54 when I put Google at the bottom there
1:05:57 I didn't realize that they had
1:05:59 completely swapped out the large
1:06:01 language model like that's a that's a
1:06:03 different large language model
1:06:05 so that probably deserves a spot higher
1:06:07 in the list but I haven't played with it
1:06:09 enough
1:06:11 and and they're all kind of different I
1:06:12 kind of feel like
1:06:14 like Bing is good for more corporate
1:06:16 kind of stuff
1:06:18 chat GPT with code interpreters just in
1:06:20 a class of its own gpt4s in a class of
1:06:22 its own Claude is a very viable
1:06:25 competitor to gpt4
1:06:29 yeah
1:06:30 something like that
1:06:33 the more triggering you are the more
1:06:35 viewers you get add a triggering title
1:06:37 to the image sad but true
1:06:42 a triggering title out to that image
1:06:47 what like gender affirmingcare.com
1:06:55 I'll see my 87 year old mom at the end
1:06:57 of August she's excited for me to
1:06:59 introduce her to AI she saw 60 minutes
1:07:02 yeah when it starts getting on 60
1:07:03 Minutes this happened in the early days
1:07:04 of the web right it started getting on
1:07:06 60 Minutes they started doing all those
1:07:08 pieces
1:07:12 all right
1:07:14 I got five more minutes I'll do a couple
1:07:16 more questions I still hope this will
1:07:18 increase empathy among people I I think
1:07:20 it will yawn I think it this stuff's
1:07:23 actually really good at empathetic cues
1:07:27 it it responds in nuanced ways
1:07:31 um pie in particular Pi dot Ai and the
1:07:34 PI app
1:07:36 is very very empathetic like that's been
1:07:38 programmed like they want that to be
1:07:41 your friend
1:07:43 and
1:07:45 I think it's training I think it's
1:07:47 training I think
1:07:48 if you've used these things for a while
1:07:50 it I mean unless you're a weirdo like I
1:07:53 find myself using please and thanks and
1:07:56 good job
1:07:57 when when these things do good hey could
1:08:00 you please do this and as I'm writing
1:08:02 the word please I'm like I don't need to
1:08:04 write the word please I'm like oh [ __ ]
1:08:05 it like it's nice to me I'm gonna be
1:08:07 nice to it and you know parents people
1:08:09 have told me like it's better if it gets
1:08:12 better if you do that [ __ ] I don't know
1:08:14 if that's true
1:08:15 but um
1:08:17 but I I I think these things I think
1:08:20 we're going to get increasingly
1:08:21 comfortable with conversational
1:08:23 interaction with the machines
1:08:26 and the Machines
1:08:28 are going to model for us
1:08:31 how not to be such trashy [ __ ]
1:08:33 [ __ ] to each other
1:08:37 and we might get better I think that's
1:08:39 right
1:08:40 what's pie sorry I came late Matt p i
1:08:43 dot a i is the website or Pi the app
1:08:49 Kyle you're on late tonight I had a
1:08:51 salon an AI Salon tonight so I got home
1:08:54 late we went out for drinks after the
1:08:56 salon and then I got home late and then
1:08:59 I had dinner and then I watched a little
1:09:01 TV show
1:09:02 hi Jack if you're wondering
1:09:06 and then I got on the old tick Tock
1:09:12 um how happy are you with your
1:09:13 screenplay
1:09:15 AI Edition
1:09:16 and is it lengthier
1:09:20 um I haven't I haven't written any
1:09:21 screenplays I just had it summarize some
1:09:23 [ __ ] for me so I haven't done a full I
1:09:26 haven't done a full writing yet so I
1:09:27 don't know it's probably lengthier the
1:09:30 the lengthier thing is an issue
1:09:32 um
1:09:34 you gotta you gotta tell it to there's a
1:09:36 there's a phrase you can use
1:09:38 um limit Pros
1:09:40 that will make it talk less so I would I
1:09:43 the other thing you can do the bear is
1:09:46 um
1:09:50 give it examples of your writing it's
1:09:52 really good at mimicking style so if
1:09:54 you're finding that it's giving you [ __ ]
1:09:56 that's just a little too flowery and my
1:09:58 problem is it's not even flowery it's
1:09:59 just like it adds words for no [ __ ]
1:10:02 reason
1:10:03 um give it example dialogue in
1:10:06 particular and it'll probably help
1:10:07 because it it seems these tools seem to
1:10:11 be really remarkable at ingesting stuff
1:10:13 analyzing it mimicking it
1:10:16 you know it's it's just really good at
1:10:18 that stuff
1:10:20 um if fun travels to another country
1:10:23 does that make one that's that's good
1:10:25 joke there's Thelonius
1:10:27 my mom I would say it out loud but I'm
1:10:29 just afraid of like what's gonna get me
1:10:31 kicked off [ __ ] Tick Tock again uh my
1:10:34 mom is 71 and thinks AI is cool awesome
1:10:36 love your mom I'm now addicted let's
1:10:39 chat what's that what did you say
1:10:42 to the let's chat shortcut currently
1:10:45 asking it to be Michael Kane what's the
1:10:48 let's oh the let's chat shortcut on the
1:10:50 iPhone oh that's cool yeah I mean yeah
1:10:54 those shortcuts are cool and Pie's cool
1:10:56 there's there's something about it it is
1:10:58 it's kind of addictive
1:10:59 my mom and I write great songs with chat
1:11:02 that's awesome
1:11:04 all right
1:11:06 listen I gotta go everybody it is 11 49
1:11:10 here in Denver Colorado hope you're all
1:11:13 well
1:11:14 enjoyed it tonight it was a fun group
1:11:15 peace out everyone have a good one