aiLL 9/24/23 - The Best AI Tools for Learning and Business

Description

Chapters

Transcript

0:01 foreign
0:39 welcome to the AI learning lab
0:42 I don't know why I always say it like
0:44 that but I do
0:45 lab learning lab
0:50 it's uh it's like our new audio branding
0:53 like
0:57 hey rongo what's happening
1:02 rongo come on down
1:06 you're gonna have contestant on the AI
1:08 is right
1:12 happy Sunday night Sherry Banks what's
1:15 happening welcome welcome
1:18 good evening to you
1:21 one of your friends is live
1:27 well that's cool
1:29 Dr J what's shaking
1:32 it's going down welcome everybody
1:34 welcome to the AI learning lab that's
1:36 Ali the AI learning lab
1:39 I was just smooching my dog I got like
1:42 hair all over my face so I'm a little
1:44 itchy
1:45 well like
1:47 like I just kissed the dog
1:51 all right Tobias what's evening
1:54 who this hey Jason who that
1:57 who you
2:00 thank you for the lemonade
2:03 that's Le Monde buddy is what that's
2:06 called thank you hey regulars
2:10 Ike thank you for the Roses the roses
2:14 are lovely
2:15 I appreciate that
2:19 Jason is early this time listen
2:26 it is a concern it's so much a concern
2:28 that we have a Graphic for that
2:31 [Laughter]
2:34 Emilio's wife started this tradition
2:39 oh man dang 105 people already oh wow
2:42 there are 120 now wow welcome everybody
2:45 uh my name is Kyle Shannon this is the
2:48 AI learning lab and uh
2:51 normally I'm I'm waiting for a few
2:53 minutes to let folks come in but it
2:55 looks like a lot of people coming in
2:56 welcome everybody
2:58 um a little bit about my background
3:01 I am an entrepreneur a many time
3:02 entrepreneur at this point I'm un
3:04 hireable
3:06 I'm the current co-founder and CEO of a
3:10 company called storyvine which is an
3:12 automated video storytelling platform
3:14 and I do this I've got this thing called
3:17 the AI Salon I'm the chief generative
3:20 officer for a group of marketing
3:23 organizations marketing companies uh
3:26 what is a chief generative officer you
3:28 ask good question I'll let you know when
3:30 I figure it out we generate [ __ ]
3:35 um I think I'm the world's first Chief
3:37 generative officer though I've had the
3:39 title for a while uh which is kind of
3:43 funny
3:44 um I was also an entrepreneur in the
3:46 early days of the World Wide Web so I
3:47 was a co-founder of agency.com which was
3:50 one of the early
3:51 um digital agencies and why I bring that
3:55 up is the early days of the world wide
3:57 web and the early days of this
3:58 generative AI thing that we're going
4:00 through are remarkably similar and so I
4:03 have a a fair amount of perspective
4:05 about having been at the beginning of
4:08 one of these you know transforming
4:10 technology
4:13 births and uh
4:16 uh it's it's we're in for some wild
4:18 times this is uh we're we're in a pretty
4:21 remarkable uh state right now so
4:24 um what uh what I like to do here
4:27 where's my
4:29 there it is what's new in keynote I
4:31 don't really care right now
4:34 um if you are new to this AI stuff and
4:37 if you're just here like is that Tom
4:39 Arnold
4:40 or or you know you've heard of AI but
4:43 you haven't really played with it yet
4:45 while you listen to me ramble tonight if
4:48 you choose to stay go there and start
4:50 playing chat.openai.com that's the
4:52 official chatgpt website or go to
4:55 claude.ai both of them are free Bard's
4:58 also free but those top two
5:02 are both pretty remarkable what they
5:04 call large language models or chat Bots
5:06 or whatever you want to call them AI
5:08 chat Bots they are
5:10 they kind of look and act like a Google
5:12 search box but what they actually do and
5:15 no is dramatically different and
5:18 profound so if you haven't played with
5:20 them go there right now and just go and
5:22 tell it to write you a country song
5:24 about a topic uh tell it to write you a
5:27 marketing plan for your business tell it
5:29 what your business is and then say write
5:30 me a marketing plan and describe I don't
5:32 know what you want to accomplish or who
5:34 the target audience is doesn't matter
5:35 just see what it does
5:37 if you're you know decently geeky have
5:40 it write you some python code if you
5:42 haven't seen it write code it's
5:43 remarkable
5:45 um that document right here prompts.chat
5:48 is also
5:50 um worth looking at that will teach you
5:52 how to talk to these things
5:54 so anyway that's it to get started so
5:57 what this channel is all about is just
6:00 talking about AI I come here every night
6:02 um we tend to hang out for a couple
6:04 hours and wax poetic about the future of
6:08 what this AI stuff's gonna do what it's
6:10 currently doing where it's good where
6:12 it's bad where it's a janky piece of
6:15 [ __ ] where it is mine
6:18 face meltingly remarkable so if you have
6:21 questions about AI pop them in the
6:23 comments below I'll do my best to answer
6:24 them some of the who we call the
6:27 Irregulars the folks that show up here
6:29 regularly are are actually modding as
6:31 well so if you if you ask something good
6:32 they'll pin it so I can see it
6:35 um and that's about it that's kind of
6:37 how it works here I've got a whole list
6:39 of apps that we can talk about and play
6:41 with and a bunch of other stuff
6:44 um
6:45 two things two kind of promo things let
6:48 me flip screens for a second
6:51 two kind of promo things
6:54 one
6:56 is that
7:00 tomorrow night is the AI salon so we we
7:03 have meetings every other week uh for
7:06 the AI Salon it's a community that I
7:08 created a year ago in December and we've
7:12 got an online Discord server where you
7:13 can make mid-journey images and ask
7:15 questions and things like that but we
7:17 also have these these meetings that are
7:19 bi-weekly in person here in Denver I'm
7:22 in Denver Colorado and then also uh you
7:26 can you can dial in remotely so if you
7:28 go to the salon.ai
7:30 um sign up for the Meetup and that will
7:33 give you get you signed up for
7:34 information for tomorrow's meeting and
7:37 then it'll also give you a link to the
7:39 Discord
7:40 storyvine.com that's my company
7:42 um so that's that I'm also I've been
7:45 invited by another Creator named Rhonda
7:47 Phillips uh to put on a series of
7:50 workshops and so we are doing a set of
7:54 three workshops for creatives called the
7:57 AI launch pad starting October 1st and
8:00 then it's weekly it's a two-hour session
8:02 weekly for three weeks and we're going
8:04 to start out with AI Basics and then
8:06 we're going to get into applying AI to
8:08 your business and then we're going to do
8:09 some Advanced AI stuff like animation
8:11 and video and things like that so that's
8:13 really exciting if you want to learn
8:16 about that if you're creative and and uh
8:18 want to learn more about that that's
8:20 there
8:21 and what else I think that's about it
8:24 let's um
8:26 again if you haven't played with chat
8:28 GPT there's that that's the official
8:30 website right there
8:32 and then prompts.chat will teach you how
8:34 to talk to that so with that let's dig
8:37 into questions I see a lot of a lot of
8:40 comments below
8:42 likely just a bunch of people saying hi
8:44 to each other but let's go check it out
8:46 and start talking happens Sunday night
8:50 good evening
8:51 what really here to talk about tonight
8:53 and I think you all know this is
8:58 yeah
8:59 let's do this let's make some money with
9:01 your gbt
9:02 [Laughter]
9:05 can you share some tips on on investing
9:09 early in AI
9:13 I can
9:15 to be clear
9:17 I have a degree in acting so if you want
9:20 to take investing advice from an actor
9:23 talking about AI you've come to the
9:25 right place
9:28 here's here's the deal I you know in
9:31 terms of what to invest in I don't have
9:33 a [ __ ] clue um I don't even know
9:35 what's investable what I can tell you is
9:38 the the companies that I find
9:40 interesting right now that I think are
9:42 interesting to follow and then you can
9:44 decide whether to invest or not
9:46 the companies that I think are are
9:52 yeah interesting is the best word
9:56 like they're worth following
9:58 um on the on the large language model
10:01 side so on the AI side
10:04 you've obviously got open AI in
10:06 Microsoft
10:08 um one of the interesting things about
10:10 how things have played out with open Ai
10:12 and Microsoft is you know they're
10:14 they're clearly making big bets on this
10:16 stuff
10:17 I thought that chat GPT would have would
10:21 have dug into and and being with chat
10:23 GPT and it would have dug into Google's
10:26 search traffic more than it has it
10:29 basically it's maybe moved to point but
10:31 it hasn't moved much so that's
10:33 interesting but that said
10:36 Microsoft is rolling out
10:39 um chat GPT in all of their Office 365
10:42 stuff on November 1st it they just
10:44 rolled it out in Windows 11. it's
10:46 already part of GitHub co-pilot or it is
10:50 GitHub copilot um it's going to be in
10:52 Dynamics 365. it's already part of azure
10:56 um
10:58 uh open AI um hit a hit a billion dollar
11:02 Revenue run rate one of the things that
11:04 was kind of humorous about a month ago
11:06 is there was an article that came out in
11:08 a German magazine that said ai's burning
11:11 through 700 000 a day and so they're
11:14 going to be bankrupt by whatever some
11:15 date in 2024 end of 2024 and then you
11:19 know three weeks after that you know
11:20 they hit a billion dollars in in you
11:24 know Revenue run rate well 700 Grand a
11:27 day is about 250 million dollars right
11:29 so so they're 750 million dollars to the
11:33 good now I'm sure they're spending a
11:35 bunch of that on R D and product
11:37 development so so they're not
11:39 necessarily uh flush with cash right now
11:41 but but they're obviously interesting
11:45 I think anthropic is interesting
11:48 um that's a bunch of X openai people I
11:50 think inflection is interesting the
11:51 people behind pi.ai
11:54 um they just raised 1.3 billion dollars
11:58 um most of that's going to be spent on a
12:00 22 000 GPU server Farm
12:04 um
12:05 I think Runway ml is interesting but
12:07 there's probably going to be a lot of
12:08 video companies out there but they've
12:10 kind of got the early share of mind for
12:12 text to video
12:15 um
12:16 while mid-journey is the king of the
12:19 hill right now in terms of image quality
12:21 Dolly three from open AI is coming out
12:25 soon
12:26 and I think it's gonna kick mid-journey
12:30 in the ass
12:31 um stability AI the company behind
12:33 stable diffusion is really interesting
12:36 um Nvidia
12:38 effectively has a monopoly on gpus for
12:41 AI right now I don't know how they
12:43 [ __ ] pulled it off I don't know why
12:45 the rest of the chip manufacturing World
12:46 seeded that whole world to them but they
12:49 seem to have done that amd's trying to
12:52 catch up but they're they've not caught
12:53 up and they don't have the software
12:55 or the developer support to really make
12:58 any move so Nvidia is looking like
13:00 they're in a cat bird seat right now
13:02 Apple
13:04 um
13:05 if you listen carefully to their
13:08 announcements and their products
13:11 they are very very quietly
13:14 building an infrastructure to be able to
13:17 do high level machine learning on device
13:21 meaning you won't have to send your data
13:23 off device to do all this magical AI
13:26 [ __ ]
13:27 They're Gonna Come Out Swinging it's
13:29 probably going to be
13:31 next year is my guess but they're going
13:34 to come out swing it Apple's gonna be
13:36 good
13:37 um Google I wouldn't bet against them
13:40 but they're a bit on their heels
13:43 um
13:43 meta
13:45 I don't know apparently they're going to
13:47 be rolling out AI chat bots in Facebook
13:50 and Instagram in the next week or so I
13:53 just saw that someone just posted that
13:54 to the AI salon
13:57 News Channel hang on where was that
14:00 news and happenings
14:02 Wall Street Journal meta will release AI
14:05 chat Bots as soon as this week with
14:07 distinct personalities
14:09 um
14:11 Meadows having to backpedal from the
14:13 metaverse stuff because
14:15 no one was into that but you know
14:20 um
14:21 they're big have a lot of money and they
14:23 could do something interesting and
14:24 they've you know their their llama
14:26 models are are developing a lot of
14:29 excitement in the developer community so
14:30 so they're relevant in the game those
14:33 are the biggies right now that I'm I'm
14:35 thinking about there's probably others
14:37 but again you know don't invest on my
14:39 advice but go look at those meta failed
14:41 well the metaverse failed and here's the
14:44 here's the deal the metaverse will come
14:46 back I think Apple has successfully
14:48 redefined that as spatial Computing and
14:51 and whether or not you're actually in a
14:53 fully immersive environment or an
14:56 augmented reality environment or just
14:58 normal [ __ ]
15:01 um
15:01 I think Apple effectively established
15:05 that as spatial Computing that you can
15:07 sort of dynamically switch between the
15:09 real world and and an immersive World
15:13 um
15:13 I think AI is going to drive the
15:15 adoption of both blockchain and crypto
15:18 and the metaverse kind of Technologies
15:21 what does it take to start an AI company
15:24 these days
15:25 a weekend
15:28 seriously like now listen if you want to
15:32 go start a competitor open AI or
15:34 inflection or anthropic that's going to
15:36 cost you about a billion dollars minimum
15:39 that's probably your seed round
15:41 but if you want to just take these
15:44 [ __ ] remarkable technologies that are
15:48 sitting underneath you know chat gbt and
15:51 Claude and Bard all of that technology
15:55 is just sitting there waiting to be used
15:57 in a handful of lines of code relatively
16:00 small amount of code you can generate
16:03 pretty remarkable stuff so the good news
16:05 is you can you can now move
16:09 significantly Beyond a prototype into a
16:12 proof of concept
16:14 with essentially no money and
16:17 potentially even with no
16:20 um engineering expertise now the
16:21 engineers are going to slap me around
16:23 for that but if you just want to get
16:24 something that works go to zapier right
16:28 now
16:29 learn about their open AI calls
16:32 and play around with their interfaces
16:34 beta which allows you to do applications
16:38 so they've got a chat bot maker so you
16:41 can go to zapier right now
16:43 look at their chat bot tool
16:47 and
16:48 figure out a prompt you want to send to
16:50 that thing and you're Off to the Races
16:53 and and if you want to prototype that
16:56 go to
16:58 all right poe.com that thing
17:02 go there they've got a bot maker you can
17:04 make a bot right now so you can go there
17:06 there's there's all sorts of bots
17:07 there's a mid Journey button there's a
17:08 writing button there's a I don't know
17:10 whatever it doesn't matter life advice
17:13 bot you can go make a bot go practice
17:15 your prompting put it right there turn
17:17 it into something
17:19 and you can create what's called a
17:21 wrapper app which is basically you're
17:24 taking openai you're writing a prompt
17:26 and you're wrapping an app around that
17:29 um now here's the downside
17:32 because you can make that in a weekend
17:35 and it's really really powerful so can
17:37 everyone else
17:38 so so the real question right now is not
17:42 what's it take to start an AI company
17:45 the real question right now is is there
17:48 any software company
17:51 that can't just be knocked off like that
17:56 that's a bigger question
17:59 I think the answer to that lies in
18:01 proprietary data rather than the the
18:04 application layer but I honestly don't
18:06 know
18:08 I mean here's the good news also you
18:11 could also start an AI company like
18:13 people are desperate to learn about this
18:15 [ __ ] right now you can start an
18:17 educational AI company I mean this
18:18 channel is kind of that although it's
18:20 not really a company it's just me
18:22 showing up here in my home office
18:23 shooting the [ __ ]
18:25 but um but you know if I if I if I
18:28 wanted to I could I could turn this into
18:30 a media operation
18:32 um or or a Consulting operation so
18:35 there's going to be a lot of people that
18:36 need to figure this [ __ ] out
18:39 nobody is an expert right now so that's
18:43 something you could do you could start
18:44 an agency
18:46 um using tools like chat GPT and
18:51 um the new Dolly that's coming out or
18:52 mid-journey or adobe Firefly and
18:54 Photoshop generated Phil you could start
18:57 an agency today this is a world I came
18:59 from
19:01 where you could go into the marketplace
19:03 and compete with traditional agencies
19:05 and undercut them by 10x
19:11 you could go you could go into a client
19:14 and say I'll deliver what they deliver
19:17 at their quality or better
19:20 for half the price
19:23 twice as fast delivered twice as fast
19:26 how long does it take them well it's
19:28 generally a six week turnaround okay
19:29 I'll do it in three weeks
19:31 at half the price they charge at the
19:34 same or better quality you Lynn you
19:36 could go start that today
19:39 um so there's a lot of there's a lot of
19:41 things you could start around and that's
19:43 that's the other thing you know I I joke
19:44 on this channel about
19:48 these people
19:51 that have the obnoxious [ __ ] Tick
19:53 Tock channels and YouTube channels
19:58 they're making money by selling you
20:01 courses on how to make money with chat
20:03 GPT
20:05 it's the oldest trick in the book
20:09 they seem so nice
20:12 no how you make money with Chachi BT is
20:14 you solve a problem
20:17 well enough
20:18 that someone's willing to pay you for it
20:21 and if you can use Ai and chat GPT to
20:25 help you solve that problem better
20:27 faster or cheaper than the competition
20:30 then you're making money with chat GPT
20:34 but you could also solve their problem
20:36 with a legal notepad or with Photoshop
20:39 or with a shovel and a hammer
20:43 these are just tools
20:45 they're really sexy tools and we we it's
20:48 easy to mistake them for a strategy but
20:50 this [ __ ] ain't a strategy these are
20:52 tools
20:54 you know I hate to break it to you but
20:58 uh yeah there's nothing new Under the
21:00 Sun
21:01 and also all the fear about these the
21:03 robots are going to kill us they steal
21:05 all that sort of stuff
21:07 there's some truth in all of that and
21:11 every time there's a new technology
21:12 that's a transformative technology
21:14 people are [ __ ] terrified of it it
21:17 happened with the camera it happened
21:19 with the printing press it happened with
21:21 the steam engine it happened with the
21:23 tractor
21:24 so help me with post should I subscribe
21:26 to it and have a variety or stick with
21:29 paid chat gbt it's a good question
21:32 Paulie Z so let me go show you
21:35 um let me let's go take a look and I'll
21:38 I'll I definitely have an opinion on it
21:44 and and by the way I'm not connected
21:46 with any of these companies so so if
21:48 you're like oh he's just a shell for
21:49 open AI or whatever I'm not
21:52 I don't really give a [ __ ] what you do
21:53 with your money although I hope you
21:56 would spend it wisely okay we're gonna
21:58 go to Poe
21:59 papa papa papa papa
22:01 all right here's Poe so on PO you have a
22:05 couple of things you've got their model
22:09 and then you have access to
22:11 Claude three three Claude models instant
22:15 instant 100K and Claude two then you
22:19 have access to open ai's chat GPT and
22:21 gpt4
22:22 if you pay the 20 bucks
22:25 to Poe you get access to these three
22:30 um
22:31 Poe also has the these Bots and this bot
22:34 maker and this is a place where you can
22:35 you can play
22:38 my my feeling right now is it is not
22:42 worth it enough well
22:45 it really depends on your use case but
22:48 but basically what you're getting here
22:50 if you pay for Poe is you're getting
22:52 Claude 2 100K and gpt4 in the same for
22:57 the same 20 bucks
22:59 what you're not getting is the advanced
23:04 access to stuff that you have at
23:07 paying 20 bucks for gpt4 for chat gbt so
23:11 let's go look at that
23:12 so in in chat GPT you've got a couple of
23:16 things you've got if you if you pay your
23:18 20 bucks you've got gpt4 which is the
23:22 best it's of everything right now
23:26 um Google's claiming that Gemini is
23:27 going to be five times better than chat
23:29 gpt4 I'll [ __ ] believe it when I see
23:31 it if it's as good as gpt4 I I'll be
23:34 amazed and pleasantly surprised but
23:36 right now
23:37 in terms of writing and and kind of
23:40 large language model
23:43 power gpt4 is still the best then you
23:47 also get plug-ins which they can connect
23:50 to the internet plugins for the most
23:52 part are a waste of time but there's a
23:55 handful of them that are actually quite
23:56 good and kind of useful then you get
23:59 this thing Advanced Data analysis this
24:01 used to be called code interpreter it is
24:04 now more appropriately named Advanced
24:07 Data analysis but it doesn't sound like
24:08 a product anymore
24:10 so it's hard to remember
24:12 but this is [ __ ] remarkable this you
24:15 can upload a CSV file of data and ask it
24:18 to tell you what that data is what are
24:21 the trends in it
24:22 visualize those Trends you making charts
24:25 and graphs it writes and executes its
24:27 own python code to analyze data it's
24:30 [ __ ] remarkable and I'm not even a
24:32 data analyst and it's I find it exciting
24:34 and then there's a third thing coming
24:35 the third and this is coming first to
24:39 people that pay the 20 bucks to chat GPT
24:42 and that is this
24:49 I typed the wrong letter
24:52 come on learn the type old man he's an
24:56 old man and he didn't know how to use
24:58 the tick tock it was very very
24:59 challenging
25:02 um how long did you watch him oh I
25:04 watched him for three hours why ah
25:07 and he can't wait oh chat okay oh please
25:13 okay this
25:15 this
25:18 you're 20 bucks to to chat GPT is gonna
25:21 give you first access to this so right
25:23 now if you make images in
25:27 um
25:28 Bing so in bing if you go to Bing chat
25:31 that's gpt4
25:33 and you say make me an image of a truck
25:36 it'll make you an image of a truck
25:37 that's Dolly too
25:39 Dolly Three is coming Dolly three in my
25:43 opinion if it is even remotely as good
25:46 as
25:47 what it looks like it is
25:50 I think this is a mid-journey killer so
25:54 it does a couple of things that are that
25:56 are that are new that that really are
25:59 not easily done right now one is
26:01 coherent text so a new tool just came
26:03 out called ideogram that can do text but
26:06 it looks like this does it quite
26:08 accurately and without any sort of
26:11 jankiness to it it's also it it you can
26:15 put incredible detail in your prompts
26:17 and it incorporates them all it's also
26:19 going to be natively incorporated inside
26:21 gpt4
26:23 and it will generate images in the
26:26 context of your entire conversation so
26:29 let's say let's say you go into
26:32 chat gpd4 and you say
26:35 um I want you to write me a children's
26:36 book about an avocado that goes to a
26:39 therapist and it'll write you the
26:40 children's book and then you can start
26:42 asking it to make illustrations for
26:43 different parts of the story it knows
26:46 the whole story so it's making images
26:47 inside the context of a much larger
26:50 conversation which staggering
26:53 um the quality of the images is quite
26:55 good look at the length of that prompt
26:57 it's long it captured everything that's
26:59 in that prompt
27:01 it's gorgeous it's stunning it's
27:04 accurate
27:05 um
27:09 it it basically this this more showing
27:13 off how detailed the image can be based
27:15 on the prompt this is Dolly two this is
27:18 Dolly three so two is not it's okay but
27:22 not great three is super detailed and
27:25 exactly what the prompt is where two
27:27 it's not quite what the prompt is
27:30 um and then this I'm going to show you
27:32 this this is this is quite remarkable so
27:35 if you've ever tried to make images in
27:38 mid-journey or stable diffusion stable
27:39 diffusion can do this a little bit with
27:41 a thing called control net
27:44 um but it's the this you you really have
27:47 to work to get this right watch what
27:48 this does I'm going to pause this here
27:50 in a second
27:56 [Music]
28:05 so why I'm pausing it so he made a
28:08 hedgehog
28:10 he picked this this Hedgehog and he says
28:12 my daughter says its name is Larry can I
28:15 see more like this dolly will now
28:19 sort of Mark that character as Larry and
28:22 now you can just make more images of
28:24 Larry
28:35 [Music]
28:54 beautiful
28:57 [Music]
29:03 [Applause]
29:05 [Music]
29:07 Kim Champion Likes music
29:10 anyway that's remarkable one of the
29:11 questions I saw pinned was uh does it
29:13 make video now this just does images
29:15 right now but um that's champion
29:18 champion sings If you go to Instagram uh
29:21 there's a channel called champ sings
29:23 that's that's my dog singing it's it's
29:25 me playing guitar and him singing is it
29:27 a basset no he's a
29:29 he's a thug
29:31 mutt mix between a beagle and a French
29:34 bull or something like that he's very
29:36 cute
29:38 um but
29:39 so I I think that the perks that you get
29:43 for paying for open AI make it worth
29:46 paying for that overpaying for Poe
29:51 I would say it it would be almost a
29:53 toss-up if Dolly three weren't coming
29:55 but Dolly three puts it over the edge to
29:57 me the combination of code interpreter
29:59 the Advanced Data analysis and Dolly
30:01 three to me
30:03 probably make this worth 50 bucks a
30:05 month but it's gonna be 20 I think
30:09 um
30:10 can't be yeah Champion was he's a good
30:12 singing dog
30:14 all right so
30:16 all right that was that was a bit of a
30:18 deep dive on should I pay for Poe or
30:21 open AI but that's that's what this
30:23 channel is about is you know I'll share
30:25 as much as I know and and you know
30:27 here's the other thing about that that
30:29 idea of you know oh Kyle's he's got a
30:32 channel he's he's an expert no I'm not
30:35 no one zero zero zero people right now
30:39 are experts zero
30:41 there are people like Ethan Malik
30:44 who have been doing this in such a
30:46 focused Academic Way for for long enough
30:49 now that I would consider someone like
30:51 him an expert but most people talking
30:55 about generative AI not just AI there's
30:58 lots of experts in AI
31:00 generative AI the world post chat GPT
31:03 starting last November
31:06 um
31:07 this stuff is so new and it's and it's
31:10 changing so fast that anyone tells you
31:13 that they know what they're talking
31:14 about is full of [ __ ]
31:19 so take everything I say with a grain of
31:21 salt if you know something better please
31:23 tell me I learn
31:25 at least as much in these
31:28 um uh lives as I share
31:32 um Kyle AI Salon is Tuesday not tomorrow
31:35 Dr J thank you very much so earlier I
31:38 said
31:39 the AI Salon is tomorrow I meant Tuesday
31:42 it's mentally Monday in my head
31:45 Tuesday so so we meet every other
31:47 Tuesday this coming Tuesday is when the
31:50 next meeting is not tomorrow
31:52 you can come here tomorrow and we'll be
31:54 doing this again but the AI Salon where
31:57 we have lots of people talking and it's
31:59 a salon and it's it's all sorts of
32:01 interesting discussions and interesting
32:03 people that's Tuesday thank you Dr Jay
32:07 um all right that was a huge help thanks
32:09 for the detail and thanks for the pup
32:11 the pup's the best
32:12 you know what's funny about a singing
32:14 dog so I play guitar and I songwrite and
32:18 stuff since we've had that dog I can't
32:20 play guitar anymore because every time I
32:22 start to play he howls like loudly in my
32:25 ear so I can't hear the guitar I can't
32:28 hear myself sing but it is so [ __ ]
32:31 cute that you like I'm like God damn it
32:34 I'm trying to sing here and he's like oh
32:36 I'm just like you just gotta love the
32:39 guy
32:40 all right my wife is is wait
32:44 is he
32:46 as retired teacher with a book idea she
32:49 could do what you just did yes
32:52 yes
32:55 if your wife is retired and has a book
32:57 idea she needs to get her butt to
33:01 chat.openai.com or claude.ai either one
33:05 of those but seriously for writing a
33:08 book
33:09 chat gbt is the best one pay the 20
33:12 bucks get gpt4
33:15 um
33:16 and then and then I'm happy to show it
33:18 but but she can figure this out
33:20 if she hasn't written a book before if
33:22 it's really just an idea and she and I
33:25 people are on varying degrees of scales
33:28 of their experience with writing like
33:30 I've written seven screenplays but I
33:32 don't know a ton about novel writing I
33:34 could write one because I've written
33:35 screenplays but it's a very different
33:36 thing if you haven't written anything
33:37 then you're in a different place right
33:41 um
33:42 a good way to start with these things is
33:45 to start with have it tell you about
33:48 different
33:49 um story Frameworks that help might help
33:52 her organize her thoughts for her book
33:54 so don't just go in there and say I have
33:57 an idea for a book about a hedgehog
33:58 named Larry
34:00 write me chapter one it'll be a shitty
34:02 chapter one it'll write you chapter one
34:04 but it'll be shitty
34:06 but if you start out and you say hey
34:08 you're a literary coach and I'm a writer
34:11 and the first thing I want you to do is
34:13 I want you to you know teach me what are
34:15 the different kinds of story Frameworks
34:17 I can use for my novel and then it'll
34:18 tell you a bunch of them and then you
34:20 can pick one and then you can refine it
34:22 and say I don't like this part I don't
34:23 like that part then you say
34:26 um ask me a bunch of questions that
34:28 about my book so you can help me
34:30 organize those thoughts and it will do
34:32 that and so there's all sorts of cool
34:33 ways to do that what about copyrights do
34:35 you have to mention the AI program
34:38 um
34:39 well you don't have to mention so so so
34:42 copyrights that
34:44 I'm not a lawyer I haven't even played
34:47 one on TV
34:49 um
34:52 copyright
34:54 is a very
34:58 very murky area right now
35:02 um
35:03 just using these things will not get you
35:06 in copyright trouble those though there
35:08 are lots of people suing these companies
35:10 saying that it should I don't think it
35:12 will for a lot of reasons mostly having
35:15 to do with how these things work
35:17 so so just generating stuff is not going
35:20 to get you in copyright hot water likely
35:23 unless you tell it to actually
35:25 plagiarize a certain author and you know
35:30 write write something you know that
35:32 they've written and it actually you know
35:35 I don't know finds a quote of theirs or
35:37 something like that or it it copies
35:39 their
35:40 whatever you could get yourself in
35:42 trouble but just using it in and of
35:44 itself won't get you in trouble
35:46 um
35:47 when it comes to you being able to
35:50 copyright your work that's a different
35:52 story The the one thing I would say is
35:55 you should document the [ __ ]
35:58 out of how you use these things
36:00 documentary prompt document your notes
36:04 have notebooks because the the bar that
36:08 is being established by the copyright
36:09 office is how much human interaction was
36:13 there how much contribution human
36:15 contribution was there
36:17 so
36:18 when I work with chat GPT
36:21 I
36:24 maybe three percent of the time I say
36:27 give me something it spits it out and I
36:29 use it but almost never most of the time
36:32 I have an idea I've jotted some ideas
36:35 down and I'm like oh let me throw this
36:36 in chat gbt so I take a beginning idea I
36:38 toss it in there I say help me refine
36:41 this or expand it or this or that or I
36:43 have an idea about blank and it writes
36:44 that out and like no that's not it this
36:47 is it so I have a lot of back and forth
36:48 with it so where the boundaries are to
36:52 what is actually AI written or not are
36:55 getting increasingly blurry
36:58 um Google
37:00 a week ago week and a half ago released
37:03 duet for Google workspaces So within
37:05 Google Docs now you have this button
37:07 called help me write that's AI
37:10 November 1st Microsoft is launching
37:13 copilot within Microsoft Word PowerPoint
37:16 Excel all that sort of [ __ ] so you're
37:18 gonna have ai within Microsoft Word so
37:20 you're going to be writing your book and
37:22 there's going to be a little button
37:23 there help me write this help me write
37:25 this help me write this so where the
37:28 boundary is of what's pure AI written
37:31 and what's pure human
37:33 is going to cease to exist very very
37:35 quickly
37:37 if you want to copyright something
37:38 you're going to have to be able to prove
37:40 to the copyright office that you did it
37:46 so you know at least they're doing that
37:48 with graphic novels and things like that
37:50 I mean I know technically when you write
37:52 it it's copy written but when it comes
37:54 to getting the copyright office to do it
37:58 in a registered kind of way
38:00 um
38:01 your ability to track your contribution
38:03 is going to be the key did you see harpa
38:05 for creating time stamps for video no
38:09 let me see that let's go look at harpa
38:13 sounds like DARPA harpa
38:17 heimstamps video
38:22 compilation of rare US postage stamps
38:25 thanks Bing
38:29 e
38:32 [Music]
38:35 I guess I need it to be one word
38:38 assembly AI versus harpa AI translate
38:42 audio and video files into live speech
38:44 and text
38:46 foreign
38:50 what is this weird oh this is a weird oh
38:53 God I'm so sick of sights that are not
38:56 actual sites
38:57 like every everyone is so [ __ ] gamed
39:00 Google SEO that most sites that you go
39:04 to now are just some [ __ ] generated
39:07 click bait [ __ ] it drives me [ __ ]
39:10 crazy it's like I don't know I don't
39:13 know how people are still using Google I
39:15 really don't it's like the the the
39:16 [ __ ] results are just useless anymore
39:22 I haven't I'll go check that thing out
39:25 um
39:26 I if if you're in here I have probably
39:29 spent I don't know at least a full
39:33 week's worth of work trying to get
39:35 clawed to do correct time stamps in my
39:39 videos and I actually have it if you go
39:40 to
39:42 if you go to the YouTube channel for the
39:45 AI learning lab it's called
39:47 AI learning lab Dash TT that's the
39:50 YouTube channel this thing right up here
39:54 um all the all the descriptions and
39:56 titles and chapterization is done with
39:59 Claude with this really long
40:01 sophisticated prompt that it's taken me
40:03 like you know 40 hours of work to get
40:05 right and it's still not quite right
40:07 like what's frustrating about it is
40:09 it'll do something [ __ ] brilliantly
40:12 and then I'll just move to the next
40:14 video over it's like the exact same
40:16 thing and it just fails miserably
40:18 so I'm starting to figure out the quirks
40:22 of it but all these tools are like like
40:23 that right now
40:25 they demo really really well
40:28 they can get you to 80 very very quickly
40:31 but to get to 90 95 and 100 of the
40:35 output that you want it's almost
40:36 impossible
40:38 how about illustrations that AI shares
40:41 um again it's the same thing I think
40:43 it's I think it's um
40:47 the the the precedent right now there's
40:50 two cases
40:51 there's two cases where the copyright
40:53 office has basically said the human
40:55 contribution wasn't enough
40:57 so so basically what they're saying is
40:59 if the machine makes it it's not you
41:01 know human generated
41:03 I want to then ask them what about the
41:05 camera
41:06 because the camera is just making an
41:10 image isn't it sure the human is
41:12 pressing the button and the human's
41:14 looking through it and turning the focus
41:15 well they're not even turning the focus
41:17 anymore you know unless they flip it to
41:19 manual so there are technological
41:22 equivalents right now where the work can
41:25 absolutely be copy copyrighted
41:28 um it's just that the
41:31 the level of
41:34 understanding of how these tools work
41:36 and and what AI generated actually means
41:39 is very very flimsy right now
41:43 um
41:44 so so again if you're using mid-journey
41:48 or if you're using fire if you if you're
41:50 using anything to generate the images
41:53 um just track everything track your
41:56 prompts track your notes if you take
41:59 that image and bring it into Photoshop
42:01 and clean it up and edit it and erase
42:03 the background and all that sort of
42:05 happy horseshit just track all that
42:08 um is that a pain in the ass yeah
42:11 and and you know do you have to tell if
42:14 if you're making a kids book for example
42:16 and you use mid-journey or one of these
42:19 things
42:20 um you know do you have to declare that
42:23 it's AI generated
42:25 I don't know I don't think so
42:28 you don't you don't have to declare that
42:29 you used Photoshop
42:31 right
42:33 you know these images were were made in
42:36 Adobe Illustrator these these images
42:38 were made in Photoshop these images were
42:40 made in AI like it there's there's a
42:43 there's a weird perversion happening
42:45 right now because of an ignorance about
42:47 how these tools actually work and what
42:48 they are and the fear of them
42:50 right
42:52 the robots are gonna kill us don't give
42:54 the robots copyright humans are the only
42:57 thing that matter like that's that kind
42:59 of Hysteria is going on right now and so
43:02 the copyright office is like all right
43:03 we don't want to get firebombed
43:07 you can't copyright robot work got it
43:10 so
43:11 um I think and and here's here's the
43:13 other thing
43:14 I actually don't know how to do that I
43:17 don't know necessarily the value of
43:19 doing that
43:20 um you know put the book out there
43:22 copyright it register it if they reject
43:25 the
43:26 registration then try again and make
43:29 your arguments I like I don't [ __ ]
43:31 know and in the meantime sell your book
43:34 so code interpreter got a huge CRM
43:37 database that's blown up
43:40 with over 15 years of contracts exported
43:43 in a CSV file in desperate
43:48 need of cleanup can it help Paulie Z not
43:51 only can It help it can [ __ ] it'll do
43:53 it it'll do it like that it'll [ __ ]
43:55 export out a new let's go do it we'll go
43:57 play with one
43:59 um
44:00 it'll export out a new
44:03 Excel file with cleaned up data in it
44:07 it's just [ __ ] stupid let's here
44:09 let's go play
44:10 let's go play code interpreters insane
44:14 and I know nothing about data analysis
44:17 you know I'm I'm good and geeky I like
44:19 geeking out but
44:22 code interpreters just stupid
44:26 and stupid I'm saying that like in the
44:28 vernacular like stupid good
44:33 I don't know I was very confused first
44:36 he said that code interpreter was good
44:38 but then he said it was stupid it's he's
44:40 very inconsistent I don't understand him
44:43 all right
44:46 so you go to data Advanced Data analysis
44:49 we'll go to a place first I wonder this
44:52 this will be interesting to see so
44:55 as a non-data analyst
44:59 I always wondered where do you get data
45:02 because I know you can get it from you
45:04 know spreadsheets and [ __ ] like that but
45:06 like if you own a big pile of data to
45:08 play with where do you go well there's a
45:09 site for that it's called kaggle there's
45:11 probably many others but this one
45:13 k-a-g-g-l-e
45:17 and you can just go ask for data so let
45:21 me see
45:23 I wonder if they've got something in
45:25 here called Dirty data data sets
45:29 uh what did you say it was what kind of
45:32 data was it can someone repin
45:35 that comment about the about asking
45:39 about code interpreter whoever pinned it
45:48 so code interpreter got a huge CRM
45:50 database okay
45:53 can it help
45:55 that's blown up with over 15 years of
45:58 context exported Okay so
46:02 so let's go here and we'll go
46:05 um
46:06 I'm going to search for
46:08 dirty CRM data
46:16 and I want to limit it by
46:20 um
46:24 I don't want a notebook no
46:29 all right
46:31 I'll just do CRM data then
46:34 CRM
46:36 [Music]
46:37 CRM
46:41 CSV
46:46 sales for stock date no used car listing
46:50 student course recommendation CRM
46:53 project
46:55 what's this
46:59 no description available
47:02 um
47:04 that's not good
47:06 [Music]
47:06 um
47:15 Salesforce CRM Instagram liked
47:22 oh
47:27 hold please
47:35 let's not find anything all right let's
47:37 just go find we'll go find
47:39 um
47:40 we'll say uh
47:43 um let's see consumer Behavior
47:52 and we'll go CSV and we'll go medium
47:56 sized
48:10 beer data analysis oh it's 144 megabytes
48:14 I think you can only have up to 100
48:15 megabytes
48:16 I don't I don't know that to be the case
48:18 though
48:19 um
48:22 all right we'll go grab consumer
48:24 complaints this is a relatively large
48:25 data set 91 megabytes
48:29 so we're going to down we're going to
48:31 download our data set so it's just an
48:32 archive.zip archive11.zip
48:35 and we're going to go back to
48:38 um what's called Chachi BT
48:46 all right how many people are in here
48:48 all right so we're going to open up that
48:50 archive so now it's gonna it's gonna
48:52 take it a second to upload but we're
48:54 gonna we're gonna say
48:56 um
49:00 uh come on Siri give me my microphone
49:03 back
49:05 uh
49:06 uh Hey Siri go away
49:09 okay
49:17 please look at the data inside this
49:21 archive
49:25 identify what it is and tell me what's
49:28 interesting about it
49:32 so notice
49:34 I talked to that as if I knew nothing
49:37 about data like a CEO might talk to a
49:40 data analyst that worked for them right
49:43 hey here's the CSV file it's got some of
49:46 our customer crap in it for the last 15
49:48 years can you deal with this
49:50 that's essentially what I did now what
49:52 code interpreter is doing I mean this is
49:54 [ __ ] mind-blowing to me it every time
49:56 I demo this thing it just the only thing
49:59 more remarkable than
50:02 you know demoing this is the amount of
50:04 fingerprints on my computer
50:05 [Laughter]
50:08 I'm here so it says let's start by
50:11 extracting the contents of the uploaded
50:13 zip file archive to inspect what's
50:15 inside and if I open up this little box
50:17 of what it did here's the python code
50:19 that it wrote and executed for itself it
50:22 imported a library wrote the code and
50:25 then exported the uh you know the list
50:27 of files that were in it then what did
50:30 it say
50:31 it can it can contains a single file
50:34 named consumer complaints
50:37 let's load a sample
50:40 so date received product uh sub product
50:43 so there's there's a whole bunch of [ __ ]
50:45 there's a whole bunch of
50:47 points interesting points the product
50:50 issue and Company columns can provide
50:52 insights into the most complained about
50:55 products State and zip might be the
50:57 locations date received company sent
51:00 next steps we can perform a more
51:02 detailed analysis here's what I'm going
51:03 to ask it for you
51:06 can you look through the data and
51:08 identify any inconsistencies
51:12 that would require data cleanup before
51:18 efficiently moving forward
51:24 certainly data cleanup is a crucial step
51:27 in preparing the data set for analysis
51:28 let's start by loading the entire data
51:30 set and performing an initial
51:32 examination by identifying missing
51:35 values duplicate rows inconsistencies
51:37 and any anomalies in the data so it is
51:40 prompting itself
51:42 to solve the problem
51:45 right so so if you upload your CSV file
51:48 into this thing and say what you know
51:50 the problem to be you can tell it to go
51:53 find the problem and then tell it to fix
51:56 the data and Export it as a new CSV or
52:00 or as an Excel file all right so let's
52:02 see what it found
52:03 missing values some columns have missing
52:06 values such as subproduct blah blah blah
52:09 no duplicate rows were found
52:11 data types
52:18 are of object data and should be
52:20 converted to date time for proper
52:22 handling and Analysis that's a big one
52:25 inconsistent categories miscellaneous
52:28 recommended for data cleanup
52:30 would you like me to perform these data
52:32 cleanup steps watch this yes please and
52:37 then
52:39 could you
52:42 save
52:44 the cleaned up data into a new CSV for
52:51 me
52:53 so I don't cry anymore
52:58 [Laughter]
53:04 you should touch your large language
53:06 model like it's it's human you should
53:09 say please and thank you and you should
53:11 crack jokes with it absolutely I can
53:13 help with that normally it'll
53:15 acknowledge something like the fact that
53:17 I'm crying oh I'm sorry you're crying
53:21 foreign
53:22 how do you get chat EPT to read a file
53:25 um I'll I'll show you in a second let me
53:28 just see what this does I can help with
53:30 that
53:30 um
53:32 and then save the cleaned up data into a
53:35 new CSV file so let's go look at what
53:36 it's doing
53:38 so
53:40 full DF
53:44 categorical columns
53:48 I don't know it's doing some [ __ ] in
53:50 Python
53:52 it's thinking
53:54 it's doing what computers do
53:57 um so how you how you upload the file
53:59 there's when you go into Advanced Data
54:01 analysis and chat GPT there's a little
54:04 plus button right here that's how you
54:05 upload the files
54:09 um summary of data cleanup converted the
54:11 date received and date sent to date time
54:14 data type
54:16 kept the most wait kept most missing
54:18 values as is since removing or inputting
54:21 them
54:22 screw [ __ ] up
54:24 converted appropriate columns blah blah
54:27 blah but but look download the cleaned
54:29 up consumer complaints CSV so if I click
54:32 on that starting download
54:38 um
54:42 there's the thing I got from kaggle
54:44 here's the download coming from
54:47 from chatgpt
54:51 let's see
54:54 download
54:57 dude it's coming it's coming
55:03 they're they're uploading downloads are
55:04 very slow because you know they're
55:08 changing the world they're not they're
55:09 not focusing on on upload and download
55:12 efficiency apparently
55:14 since uh since everything's gonna be
55:17 different but anyway there's
55:24 I'm trying to preview the file
55:51 and it's opening it's a giant file
55:53 anyway so yeah so yes it can
55:58 um it's remarkable it's absolutely
56:00 remarkable you can also have it do
56:02 charts and graphs here let's go back and
56:03 have it do a chart and graph
56:05 um
56:07 can you identify
56:09 a trend in the data that you think is
56:11 particularly compelling
56:13 and recommend a way to visualize that
56:16 that will bring that Trend to life
56:20 for someone who doesn't understand
56:23 the data
56:28 laughs
56:32 certainly
56:38 it's going to do a stacked area chart
56:41 it's a little hit or miss with the
56:43 charts if it doesn't quite if it doesn't
56:45 pick the right data so again
56:48 with all of these tools the more that
56:51 you understand about the task that
56:53 you're trying to get it to do the better
56:54 you'll be at prompting it to do that
56:57 because you'll be able to understand
56:59 um the difference between good and bad
57:03 would you prefer to group the data by
57:06 month quarter year or another time
57:08 period let's just go with year
57:12 foreign
57:18 [Music]
57:27 Kyle the camera is trying to focus on
57:29 your hand and not the screen when you
57:30 point I know Clint I I have a bad habit
57:33 of doing that
57:35 so thank you for that though I will try
57:37 to be more aware of my
57:40 my large arms
57:44 although they're not that large but they
57:46 are to the camera
57:50 um
57:54 so so it grabs some data it organized it
57:58 by year
58:00 it seems there was an error because the
58:02 numpy library
58:04 wasn't imported
58:06 so that's the other thing it does is as
58:08 it makes mistakes it'll identify the
58:10 mistake and then try something else and
58:12 there you go
58:13 there's your chart
58:14 I mean
58:17 look at this
58:20 it just it wrote The Python to write to
58:22 make this chart
58:25 I mean
58:31 what is this oh consumer complaints
58:34 stacked
58:35 yellow is bank account checking account
58:40 consumer loan
58:42 as you get up here at the very top one
58:43 is virtual currency
58:46 so what's this number of complaints oh
58:49 look at this this is awesome so virtual
58:51 currency the number of complaints were
58:54 super high and then crypto crashed
58:56 obviously at this point this was like
58:58 2016 it was super high it crashed in 17
59:01 it probably there's probably another
59:03 version of this where it goes back up
59:05 that's really funny
59:06 so there you go
59:09 crazy right crazy crazy crazy crazy
59:12 crazy crazy crazy crazy
59:15 it's crazy I tell you crazy oh pretty
59:19 it's so pretty oh did you see that he
59:24 made the chart out of nothing he just
59:26 asked it to make a chart and it made a
59:27 chart
59:29 I think there's I I think they're paying
59:32 I think they're I think it's slave labor
59:34 I think it's slave labor I think they
59:36 have a whole just a whole room full of
59:38 people you ask it for something and
59:40 they're just quickly writing computer
59:42 programs I don't know how they do it I
59:45 guarantee you they're not paying them
59:46 enough
59:51 and that was chat cpt's data analysis I
59:54 see that all the time but I never knew
59:56 what to
59:59 do with it
1:00:00 yeah so here's the deal like I I I
1:00:03 didn't either
1:00:05 um the the best way you can think about
1:00:07 it is this is any any data any data that
1:00:11 can be analyzed is fair game
1:00:14 so for example you can upload an image
1:00:18 to that thing
1:00:19 and you can say turn this into an
1:00:22 animation where you zoom in on the image
1:00:24 and it'll it has a python library that
1:00:26 can do some image
1:00:30 um
1:00:30 you know modification [ __ ] and it'll
1:00:32 zoom in on it and turn it into an
1:00:34 animation
1:00:35 I did a Thing Once where I uploaded an
1:00:38 image and I said find the top 15 average
1:00:41 colors in this image
1:00:43 turn it into a pie chart where the
1:00:46 slices of the pie chart are the those 15
1:00:48 colors and then name the colors a
1:00:51 creative crayon color
1:00:53 and it did
1:00:54 just [ __ ] did it and the crown colors
1:00:57 were good and they matched the color
1:00:59 like they were right I went I looked at
1:01:00 them all they were right
1:01:03 so it can analyze video can analyze it
1:01:07 can read PDFs it can't right now read
1:01:10 images gpt4 is multimodal meaning it can
1:01:14 understand images
1:01:16 but for whatever reason openai has not
1:01:18 Incorporated that into chat GPT yet Bing
1:01:21 has so if you go to bing.com and go to
1:01:24 their chat you can upload an image there
1:01:26 and say tell me what's in this image and
1:01:29 it will look at the visual part of the
1:01:30 image know what's in it and it could
1:01:32 write you an advertising you know
1:01:34 campaign for for that image
1:01:37 so
1:01:39 again different tools right now have
1:01:41 different skills even stuff based on the
1:01:44 same technology but code interpreters
1:01:47 [ __ ] mind-blowing it's also been out
1:01:49 long enough now that there's probably a
1:01:51 lot of YouTube videos of how people are
1:01:54 using it
1:01:55 but for the most part
1:01:57 if you got data
1:02:00 in any way like we've got a
1:02:03 we've got a client that wants to know
1:02:05 how many of their people filmed videos
1:02:06 in our technology and so I literally
1:02:09 just pull raw data down from Heroku we
1:02:13 made a data clip to just say how many
1:02:16 videos were filmed in this org
1:02:18 or like list them all and then I just
1:02:21 take that blob of data I throw it into
1:02:23 code interpreter I've got two pre-built
1:02:26 prompts one that says give me a list of
1:02:29 people who filmed in the past seven days
1:02:32 and then they're running totals and it
1:02:34 gives me that table
1:02:36 and then I do another one where I have
1:02:38 it make a chart and a graph of how many
1:02:41 videos were filmed total each week since
1:02:44 the beginning of the project
1:02:46 and before we wrote this incredibly
1:02:48 Byzantine Excel
1:02:51 file with these shitty formulas that the
1:02:54 guy that wrote them
1:02:55 didn't write them well so they're very
1:02:57 inefficient that's all gone I just go
1:03:00 into code interpreter and say give me
1:03:02 this [ __ ]
1:03:03 with Chachi BT a guild of cobblers can
1:03:06 now freely make weighty comment on Higgs
1:03:10 boson Higgs bosens yes
1:03:13 yes
1:03:15 so so this is this is one of the Winston
1:03:18 like you're like you're hitting on
1:03:19 something I know you're saying it kind
1:03:21 of in a in a jokey kind of way
1:03:25 what these tools are providing is the
1:03:28 democratization
1:03:30 of specialization I gotta write that
1:03:33 down that's new wait
1:03:36 sorry should I write that on my sag
1:03:38 after check from New Jack City
1:03:40 the one where I earned nine dollars
1:03:44 the Democracy wait what did I say the
1:03:45 democratization of specialization hang
1:03:48 on
1:03:51 d monetization
1:03:56 of
1:03:58 that feels like an article doesn't it
1:04:00 Alicia here's what I mean
1:04:04 I don't know [ __ ] about data analysis I
1:04:06 don't know [ __ ] about cleaning up data
1:04:08 what I do know is that you if you have
1:04:11 data that's not clean if you have data
1:04:13 that's inconsistent and like you don't
1:04:15 have the date format all consistently
1:04:17 done or in the case of the CRM data that
1:04:20 was asked about
1:04:22 um
1:04:23 you can analyze it because the data is
1:04:25 broken that historically
1:04:29 has taken either really expensive
1:04:31 software or really expensive people to
1:04:33 literally go through and clean up the
1:04:35 data
1:04:36 or someone who knows how to write
1:04:37 scripts to go do that right
1:04:40 and now you just ask for it and it's
1:04:41 done
1:04:42 and then the analysis and then the the
1:04:45 coding the python coding to create that
1:04:47 chart
1:04:48 could I learn to do that yeah if I took
1:04:51 the next month or two and learned enough
1:04:52 Python and learned about visualization
1:04:54 libraries and learned about
1:04:57 first of all I have no [ __ ] interest
1:05:00 in that and second of all I I don't have
1:05:03 two months to to solve that right so I
1:05:06 either just go without
1:05:07 and so what's happening with these tools
1:05:09 is within a very short amount of time
1:05:12 like
1:05:15 weeks days hours or minutes you go from
1:05:19 complete ignorance to being able to do
1:05:22 the thing
1:05:23 right
1:05:26 make images I'm not good with
1:05:28 photography that's okay mid-journey make
1:05:30 me a photo of huh that's pretty good
1:05:36 chat gbt
1:05:39 um I know that I read in college once
1:05:41 there's a story framework about that
1:05:44 involves a cat oh that's you're probably
1:05:47 referring to save the cat here are this
1:05:49 things have saved the cat yeah that's
1:05:51 the one
1:05:52 I have an idea for a book could you help
1:05:54 me put it in that sure what's your book
1:05:56 about here here it is organized and save
1:05:59 the cat structure okay
1:06:02 now all of a sudden you're you're a
1:06:04 master story structurer
1:06:06 so so all of the specialized tactical
1:06:11 execution that we all went to college
1:06:13 for right
1:06:14 you you you generally learn in high
1:06:17 school to figure out what you're
1:06:18 interested in
1:06:19 and then unless you're a liberal arts
1:06:21 major
1:06:21 you pick something computer science I'm
1:06:25 going to be a coder what kind of coder
1:06:27 I'm going to be a python coder okay
1:06:29 great
1:06:31 need to learn that
1:06:33 what if that specialty were democratized
1:06:36 that's the world we're entering
1:06:38 what are the implications of that I
1:06:40 don't know
1:06:42 I don't [ __ ] know
1:06:47 I know that
1:06:49 I just wrote an article called lateral
1:06:52 play
1:06:54 and I forget what the subtitle is but it
1:06:56 basically it basically talks about
1:06:59 the idea that
1:07:01 for what I need to do for my work I've
1:07:03 got a set of skills that I have and then
1:07:05 if there's anything outside of that
1:07:07 I either go hire someone or whatever
1:07:10 now I can just go play laterally I can
1:07:13 go play outside of what my work needs
1:07:15 and it feels like Play because I'm just
1:07:17 playing with I'm like making code and
1:07:18 I'm making pictures and I'm making music
1:07:20 and I'm making video and I'm doing all
1:07:22 this lateral stuff that feels like I'm
1:07:24 cheating on my business
1:07:26 and then I'll have a problem in my
1:07:28 business and a client will say you know
1:07:30 I really wish I had a something that did
1:07:31 this and I just played with that and I'm
1:07:33 like holy crap I can do that
1:07:36 I know how to do that
1:07:37 because I played with this thing over
1:07:39 here
1:07:40 that's a specialty that I wouldn't have
1:07:42 touched I wouldn't even even considered
1:07:45 touching previously and now I can just
1:07:47 go do it
1:07:51 all right
1:07:56 um I'm gonna sit
1:08:02 can AI create
1:08:04 2024 schedule for something done in 2023
1:08:09 wait uh bring that back up again oh that
1:08:11 Mandarin video I assume it's not done
1:08:13 yet I haven't paid for it let me go look
1:08:15 that was from hey Jen I'll sh I'll show
1:08:19 I'll show people the Italian again but
1:08:21 all of the people that have been here
1:08:23 before are gonna kill me
1:08:27 um hang on a sec
1:08:29 labs
1:08:34 Labs Dot
1:08:37 hn.com slash
1:08:56 I'm now I'm now 43 000th in line I
1:08:59 started at 79 000 three days ago
1:09:03 I I just gotta bite the bullet and buy a
1:09:06 subscription for this
1:09:07 um what what this person is oh wait let
1:09:09 me read this this pinned comment from
1:09:12 Winston Welsh can AI create a 2024
1:09:15 schedule for something done in 2023 so
1:09:18 the dates
1:09:19 are weekday but consistent yes
1:09:23 yes in fact code interpreter that thing
1:09:26 I assume you've got that schedule like
1:09:27 in in Excel or a CSV
1:09:30 if you take that schedule
1:09:33 and say this was a schedule for 2023
1:09:36 please update it for 2024 and make sure
1:09:38 all the days are correct and then
1:09:40 let me download it as a new CSV yes it
1:09:43 will just do that
1:09:46 um Can it proofread a book I am writing
1:09:48 if I upload my file yes although well so
1:09:55 yes yes there's two tools that you could
1:10:00 use you could use code interpreter you
1:10:02 could use the Advanced Data analysis
1:10:03 just upload your book as a text file
1:10:06 would probably be easiest I there
1:10:08 there's weirdness with PDFs with all of
1:10:11 these tools right now
1:10:12 um because here's the problem with PDFs
1:10:15 PDFs are not simple documents some PDFs
1:10:20 are text only some PDFs are a
1:10:22 combination of text and images some PDFs
1:10:25 are images only that look like text only
1:10:27 but they're images some things are some
1:10:31 text some images and some images that
1:10:34 are text
1:10:36 and so these tools are not super
1:10:40 sophisticated right now in figuring all
1:10:42 that [ __ ] out there are tools that do
1:10:45 figure that out that will likely be
1:10:47 incorporated in all this but they're not
1:10:49 right now so in the case of I've got a
1:10:52 book and I wanted to proofread it I
1:10:53 would upload it into a text file
1:10:57 and then just upload that text file into
1:10:59 code interpreter and then it can read
1:11:01 the whole book and you can start
1:11:02 interacting with it
1:11:04 I might suggest however to experiment
1:11:07 with both
1:11:09 chat gpt's code interpreter as well as
1:11:11 Claude so Claude
1:11:14 okay you believe what keeps popping wait
1:11:17 you believe what keeps popping up on a
1:11:19 daily basis on YouTube about these tools
1:11:23 um
1:11:26 tell me what in specific you think I'm
1:11:29 being naive about I might be I might be
1:11:31 being naive
1:11:33 laughs
1:11:34 I mean most of my what what I'm talking
1:11:37 about right now is based on my
1:11:38 experience with this stuff but there's
1:11:39 some stuff that
1:11:42 that I obviously have my head up my ass
1:11:43 but tell me something specific wonder
1:11:45 how singing a song would do in other
1:11:47 languages like Italian
1:11:50 that's actually a really interesting
1:11:51 idea
1:11:54 hmm
1:11:57 um what was I gonna go look at I forgot
1:11:59 to ask did you see about the Chinese
1:12:02 avatars selling 24 7 yeah in in email
1:12:05 with video
1:12:06 um I I didn't see the one that you sent
1:12:08 but I did I did see that
1:12:12 I started
1:12:14 they see it on Tick Tock today
1:12:17 or it might have been a TV like a news
1:12:19 channel no I think it was on a tick tock
1:12:21 yeah there's a Chinese company right now
1:12:24 that for a thousand bucks they'll
1:12:26 generate an avatar of you or of whatever
1:12:29 of a person
1:12:31 and then
1:12:33 it basically gets on a tick tock sales
1:12:36 Channel and sells 24 7.
1:12:39 um I don't know exactly how they're
1:12:40 doing it or what it is I saw it today
1:12:42 that's going to become commonplace
1:12:45 so
1:12:47 it's it's batshit crazy here's the let
1:12:50 me show you this thing
1:12:51 this is the thing the thing that people
1:12:53 are asking about the Mandarin thing so I
1:12:55 uploaded some videos of me in English
1:12:58 and this is a site called Hey gen and
1:13:02 they have a tool in their Labs if you go
1:13:04 to labs.hagen.com
1:13:07 that does video translation
1:13:10 so here is me in English and Italian
1:13:14 so we've heard from a number of clients
1:13:16 that modular content is important to
1:13:18 them getting more efficient with how we
1:13:20 create content and how we leverage
1:13:23 contents
1:13:29 behind me you see these little green
1:13:32 boxes um that's how we think about
1:13:33 stories at storyline we break stories
1:13:35 down into little atoms right little
1:13:37 modules
1:13:39 foreign
1:13:46 [Music]
1:13:49 so anyway this one here that's not done
1:13:52 yet is uh
1:13:53 me speaking Mandarin
1:13:56 um I I would think that If This Were Me
1:13:59 singing it it would probably do just
1:14:01 fine actually now that I think about it
1:14:03 because it's capturing it's capturing
1:14:05 the Timbre of my voice it's capturing
1:14:07 the pacing it's capturing the inflection
1:14:10 it's capturing the Dynamics
1:14:13 so singing is just talking with longer
1:14:17 vowels so I think it would work
1:14:20 um but let me go I need to check
1:14:22 something here I'm checking
1:14:25 for
1:14:28 where is that
1:14:32 settings moderators
1:14:38 [Music]
1:14:44 [Music]
1:14:49 add
1:14:50 huh
1:14:53 there you go Herod I just
1:14:56 added you I added you before but maybe I
1:14:58 didn't hit save or something like that
1:15:00 all right you should be a mod now
1:15:02 all right when I use my free chat gbt
1:15:05 and I ask it to give me some social
1:15:07 media data it doesn't know okay this is
1:15:09 an important thing
1:15:11 thank you for that
1:15:13 um
1:15:14 it doesn't know current data why Okay so
1:15:21 okay
1:15:36 okay
1:15:37 large language models the way they get
1:15:40 trained is
1:15:42 they basically will get a pile of data
1:15:46 but Kyle what kind of data what do you
1:15:48 mean a pile I mean like all of the
1:15:50 public internet or as much as they can
1:15:51 get their hands on they go get all of
1:15:53 Wikipedia and they go get all of
1:15:55 databases that they have access to and
1:15:57 they go get all of blogs and websites
1:16:00 and just as much [ __ ] crap as they
1:16:03 can and they organize it and that and
1:16:05 they do this thing called embedding and
1:16:07 embedding is where they take these big
1:16:09 pieces of data and they blow it through
1:16:12 this machine learning model that
1:16:15 converts it from documents
1:16:19 to what's called what are called vectors
1:16:21 basically mathematical
1:16:23 locations in a very complicated
1:16:26 thousand-dimensional mathematical space
1:16:29 in semantic clusters
1:16:31 so so basically they'll take words like
1:16:33 the word dog this is the best way I can
1:16:35 remember it the word dog
1:16:38 probably
1:16:40 to most people means the fluffy thing
1:16:42 that runs around the house that you pet
1:16:44 right and so like domestic pets is a
1:16:47 semantic clustering where the word dog
1:16:50 could live the word dog could also refer
1:16:53 to that guy that did his girl wrong he's
1:16:55 a dog right so shitty shitty
1:16:58 relationship behaviors and other
1:17:00 semantic clustering that dog also lives
1:17:03 in
1:17:03 so so they take all of the internet they
1:17:06 run it through these bizarre [ __ ]
1:17:07 algorithms that literal Geniuses have
1:17:10 created
1:17:11 that turn words into mathematical
1:17:16 locations in thousand-dimensional Vector
1:17:19 space
1:17:22 um
1:17:22 and and then
1:17:24 um
1:17:25 and then when you do your chat GPT
1:17:27 search you're not your search your your
1:17:30 query it turns your query into a
1:17:33 mathematical thing that says
1:17:35 in all of that Vector space what is the
1:17:39 next word that has the highest
1:17:40 probability of being there and that's
1:17:42 the next word that shows up and it does
1:17:44 that just over and over again until it
1:17:46 writes your your response
1:17:50 um
1:17:52 that process is very very expensive and
1:17:56 very very time consuming and you can't
1:17:59 it's not being done dynamically in real
1:18:02 time there's basically a point at which
1:18:04 they have to draw a line in the sand and
1:18:06 say after this date we have to stop
1:18:08 because we have to go process all this
1:18:10 [ __ ] and that's going to take us a year
1:18:11 to process it or whatever it takes right
1:18:14 and so in the case of GPT 3.5 and gpt4
1:18:18 they that date that that thing stopped
1:18:21 was September of 2021
1:18:23 so if you ask for something more recent
1:18:25 than that like say social media
1:18:28 um data from 2024 you're screwed but
1:18:31 there's workarounds there's always
1:18:33 [ __ ] workarounds
1:18:36 um
1:18:37 there are plugins in chat GPT that can
1:18:39 go read websites so you can if you have
1:18:42 access to a website you can point these
1:18:44 plugins there's one called Web pilot
1:18:46 there's there's a bunch of them in there
1:18:48 there are also tools
1:18:50 that are large language models that are
1:18:53 connected to the internet so Bing if you
1:18:55 go to bing.com and then go to chat
1:18:58 that's
1:19:00 chat that's built on top of gpt4 the
1:19:03 same underlying engine as chat gbt but
1:19:06 Bing is connected to the internet so you
1:19:09 can go to Bing right now and ask the
1:19:11 exact same thing of Bing that you did of
1:19:13 chatgpt and it will know it because Bing
1:19:16 can do an internet search and combine
1:19:18 that with the large language model
1:19:19 answer
1:19:22 um Bard is connected to the internet Pi
1:19:24 is connected to the internet
1:19:27 um
1:19:28 Poe most of their models are connected
1:19:30 to the internet
1:19:33 oh wait are they
1:19:35 no they're not that's a lie
1:19:38 strike that on PO I don't think pose are
1:19:41 pie Bing Bard right now those are the
1:19:44 biggies that are connected to the
1:19:46 internet do you have a list of those
1:19:47 plugins chat GPT always tells me it
1:19:50 can't use the internet
1:19:51 yeah so let's we can go play with that
1:19:53 I'll just show you there's there's only
1:19:56 really one that I use the plugins for
1:19:58 the most part are kind of a waste
1:20:01 but uh uh uh where did I put my little
1:20:05 tab I don't know we're gonna make a new
1:20:07 one we're just gonna make a new one you
1:20:08 know what we're gonna do we're gonna go
1:20:10 ahead and just type in
1:20:11 chat.openai.com you know like where
1:20:14 heathens we're just gonna type the URL
1:20:16 in rather than clicking on a tab because
1:20:18 we're not lazy we're not lazy here at
1:20:20 the lab we're not we're not we're a
1:20:22 little lazy we're we're pretty lazy we
1:20:25 like clicking on [ __ ] okay
1:20:27 so if you go to plugins
1:20:32 um right now it says I've got no plugins
1:20:34 enabled
1:20:35 so if I click on that here's a list of
1:20:38 plugins I have installed now installed
1:20:40 and enabled are two different things
1:20:41 installed means I selected it and said
1:20:45 that looks interesting to me enabled
1:20:47 means it's turned on
1:20:49 you can turn on a plug-in you you can
1:20:52 turn on up to three plugins per chat
1:20:57 I for the most part only turn on one
1:20:59 just because basically what you're doing
1:21:01 when you're turning on a plug-in is
1:21:03 you're saying I'm going to create a
1:21:04 little mini application that does this
1:21:07 specific piece of functionality that I
1:21:09 need
1:21:09 where you might turn on more than one is
1:21:12 let's say you're doing something you're
1:21:14 you're booking travel itinerary and
1:21:16 you've got a
1:21:18 um restaurant booking plug-in and you've
1:21:21 got a a flight booking plug-in and
1:21:24 you've got a local activities plug-in
1:21:27 those three things could work in
1:21:29 conjunction to help plan your itinerary
1:21:31 so there's a reason to do multiple
1:21:32 plugins
1:21:34 um
1:21:35 if you have no plugins if you're just
1:21:37 starting using plugins there's a plug-in
1:21:40 store and there are now let's see all
1:21:44 there are now what's 160 times six
1:21:47 whatever 160 times six is wait it's uh
1:21:52 six six
1:21:54 hundred and sixty times six uh 600 and
1:21:59 then 360
1:22:02 9660. there's there's almost 10 000 of
1:22:05 these [ __ ] there's almost 10
1:22:07 000 plugins so you can go search for
1:22:09 things so if you if you knew you wanted
1:22:11 to do something on
1:22:13 social
1:22:17 media
1:22:18 data
1:22:20 okay nothing showed up but if I type in
1:22:23 social media three things show up Social
1:22:26 search and short video trends
1:22:30 that's interesting
1:22:32 find top keywords for short video
1:22:35 content and that's kind of boring social
1:22:38 search provides access to tweets users
1:22:40 followers images media and more
1:22:43 I'm going to install this but this could
1:22:45 send me to a site where I have to sign
1:22:47 up for something nope okay it just let
1:22:49 me do it so it's called social search
1:22:52 so now I've got social search enabled so
1:22:56 I have one out of three enabled
1:22:58 and I should be able to just go
1:23:02 Um
1:23:03 can you find any data on the on
1:23:11 the at AI learning lab
1:23:17 account on tick tock
1:23:21 it should fire off the plug-in right now
1:23:25 I don't have direct access to tick tock
1:23:27 data however you can easily check Tick
1:23:29 Tock if you're looking for Twitter data
1:23:32 all right so let's say
1:23:35 um
1:23:37 how
1:23:38 about
1:23:43 at Kyle Shannon
1:23:46 on Twitter
1:23:49 how many
1:23:52 followers does he have
1:23:55 and
1:23:57 why isn't
1:24:00 it more
1:24:01 [Laughter]
1:24:09 spell it right I did I spelled it right
1:24:12 let's see this this should be awkward I
1:24:15 have
1:24:16 1697 followers he describes as someone
1:24:19 else creating community and passing
1:24:21 knowledge in generative AI
1:24:23 he's the co-founder of C of storyvine
1:24:26 located in Denver he's tweeted four 4
1:24:29 800 times he's following two thousand
1:24:32 accounts as for why he doesn't have more
1:24:34 it's important to note that the number
1:24:36 of followers on a user has on Twitter
1:24:38 can be influenced by various factors
1:24:41 oh it can ask about specific tweets so
1:24:44 there you go there's a plugin that does
1:24:46 social media [ __ ]
1:24:47 foreign
1:24:57 yeah we're getting someone just had to
1:25:00 comment down in the mud yeah we're
1:25:01 getting down in the mud down in the mud
1:25:03 today we're getting detailed we're going
1:25:05 down there down down down all right
1:25:08 what's the best way to write a useful
1:25:10 prompt
1:25:11 is there are there tools that can help
1:25:13 so
1:25:15 I'll give you a starting point the on it
1:25:18 honestly the
1:25:20 this is the shittiest answer
1:25:23 but it's the real answer I I wish I had
1:25:25 a better answer right now
1:25:27 um we're so early in this stuff that
1:25:30 while there are people that can teach
1:25:31 you prompting and can teach you give you
1:25:34 prompts
1:25:35 it it really is valuable in my
1:25:38 recommendation that you learn to prompt
1:25:40 and and just get curious about it so the
1:25:42 best way to do it is just go start
1:25:43 playing prompts.chat
1:25:46 um is a kind of primer document that I
1:25:48 put in here because it's it's got some
1:25:50 explanations of how large language
1:25:52 models work and it's got some examples
1:25:54 of prompts that you can copy paste where
1:25:56 you tell chat gbt to act like a certain
1:25:58 kind of persona
1:26:01 um
1:26:02 all right Harold D says he can answer
1:26:05 this question I might bring someone up
1:26:07 here that can actually answer this
1:26:09 um on the prompting thing so hang on
1:26:11 he's typing something right now
1:26:15 um let me see if I can bring you up
1:26:16 Harold let me see if I can find you
1:26:19 did
1:26:21 why can I not search here
1:26:28 this is
1:26:29 odd host viewers
1:26:32 I can't search
1:26:42 oh there's Harrow invite continue
1:26:50 he doesn't have enough followers me on
1:26:53 Twitter yeah
1:26:55 I'm not active enough on Twitter to know
1:26:57 hey Harold
1:26:59 hey what's up man or Herod how's it
1:27:02 going is it Herod
1:27:03 it's Harold harrowed uh Christian Joseph
1:27:08 whatever works all right fine
1:27:11 um so so you say you can answer this
1:27:14 question about what's a good way to get
1:27:15 into prompting you want to jump in
1:27:17 yes because this it's just it's the most
1:27:20 simple thing ever and it just slips by
1:27:22 people constantly so the best way to
1:27:26 learn how to prompt with chat GPT is
1:27:30 with chat gbt because there's no other
1:27:33 platform you can use that's going to
1:27:35 understand what chat GPT wants out of a
1:27:39 prompt other than Chachi BT so really
1:27:42 it's as simple as just asking it keep
1:27:44 one of your tabs one of your instances
1:27:47 of chat gbt open as kind of like a
1:27:50 prompt generator for all of your other
1:27:53 things that you're using it for just ask
1:27:55 it what hey this is what I want to do
1:27:57 how could I form a prompt that would get
1:28:00 me this specific type of output or you
1:28:03 know this this type of response going uh
1:28:06 just let chat GPT do the work for you it
1:28:09 knows what it wants you just gotta ask
1:28:10 it for it yeah no it's it's a perfect
1:28:13 it's a it's it's a perfect contribution
1:28:15 here there's uh Malik
1:28:18 um uh about two weeks ago put out
1:28:22 um some research he did a post on some
1:28:23 research that not only found that
1:28:27 you could you can use chat GPT to write
1:28:29 the prompts but it's better at writing
1:28:31 prompts than than people who are really
1:28:33 good at writing prompts like it right
1:28:35 exactly better than humans that's so
1:28:37 it's really good advice that's awesome
1:28:40 all right so I just want to throw that
1:28:41 in thanks a lot Kyle perfect thanks a
1:28:43 lot man
1:28:44 all right bye-bye
1:28:46 um cool awesome thank you very much
1:28:50 all right what other questions might you
1:28:53 have yes let's do that
1:28:57 let's do that
1:28:59 I use Chachi potato
1:29:02 until I encountered stun spot
1:29:05 so dragonfly Alchemy who's a an
1:29:07 irregular here
1:29:10 just yeah that that was awesome here D
1:29:12 look yeah we can do that again um that's
1:29:14 really good
1:29:15 um dragonfly Alchemy I use Chachi petite
1:29:18 to prompt it until I encountered stun
1:29:23 spot recommend y'all check it out so
1:29:26 it's a Discord thing called stun spot
1:29:29 stun spot prompting
1:29:32 I'm gonna go join it
1:29:39 right do your experience no
1:29:42 next
1:29:44 are you a coder yes
1:29:48 next
1:29:49 if you have interest in prompts for SEO
1:29:51 tasks skip
1:29:53 putting a non-skip
1:29:58 degree
1:30:01 all right wow they had a really cool
1:30:04 onboarding thing all right stun spot is
1:30:07 a prompt amplifier
1:30:09 why do we have to request to join why
1:30:12 can't you just have speakerphone
1:30:15 and burner phone number
1:30:21 um
1:30:22 I suppose I could
1:30:24 I don't know I'm here's the deal with
1:30:27 this channel I'm just trying to figure
1:30:28 all this [ __ ] out I'm trying to figure
1:30:30 out what what works best so
1:30:33 um some combination of Discord in this
1:30:35 seems to be
1:30:36 good
1:30:38 um
1:30:41 and I it's it's a possibility I I might
1:30:44 do that I just I haven't really thought
1:30:46 about it but
1:30:47 part of what I need to figure out with
1:30:49 this channel is what is this channel it
1:30:51 started out as a Content channel so I
1:30:53 was posting videos every day now I don't
1:30:55 post videos every day but I do lives
1:30:57 every day
1:30:59 um and this seems to be valuable but
1:31:02 people like the videos as well so I do I
1:31:04 try to do those when there's when
1:31:05 there's big stuff I have a big idea or I
1:31:07 have or I have an idea that's big to me
1:31:11 um or or some big news has happened
1:31:13 right there's some some new tool that
1:31:14 comes out so that's when I try to do the
1:31:16 recorded videos and then these lives I'm
1:31:18 just trying to make these as
1:31:20 conversational as possible which if it's
1:31:23 just me talking you have a good point
1:31:26 um but it's just like that because it's
1:31:28 just like that
1:31:29 good trick to get Kyle to sing I could I
1:31:32 could sing I'd be happy to sing for you
1:31:34 my problem is I only know depressing
1:31:36 songs so if we want to if we want to
1:31:38 thin out the the uh the audience if we
1:31:41 have too many if we have 17 000 people
1:31:43 and we want to get rid of a bunch of
1:31:45 them then I'll sing a song
1:31:52 oh man
1:31:53 yeah I heard d i I mean I'm a
1:31:56 singer-songwriter but but it's like I
1:31:58 like things like uh like John Hiatt
1:32:01 Martin Sexton
1:32:03 um you know Bonnie Raitt Brandi Carlile
1:32:06 you know
1:32:07 Americana kind of stuff
1:32:10 I have chat GPT plus how do I see the
1:32:12 plugins that's a good one that this is
1:32:14 easy and
1:32:17 a flipping pain in the ass that they do
1:32:20 it this way because
1:32:22 people pay for chat GPT plus because
1:32:26 someone told them plugins were cool or
1:32:28 code interpreters cool and then they
1:32:29 [ __ ] hide it so here's how you do it
1:32:32 if you just paid for chat GPD plus go to
1:32:35 settings of beta
1:32:38 and then beta features
1:32:43 and turn them on
1:32:44 so you gotta you have to explicitly turn
1:32:46 on plugins and Advanced Data analysis
1:32:49 this used to have the connection to the
1:32:52 internet it was called browse with Bing
1:32:54 then they killed it because it stole
1:32:57 content from time Inc and like
1:32:59 apparently they haven't cut that deal
1:33:01 yet and then
1:33:03 um
1:33:04 I don't know when Dolly three comes I
1:33:07 don't know if it's going to show up here
1:33:08 or if it's going to show it show up as a
1:33:10 plug-in
1:33:11 a Braun who's who's one of the open AI
1:33:14 evangelists did a video a week or so ago
1:33:18 where he said it's going to show up as a
1:33:20 plug-in
1:33:21 so
1:33:23 if you go to plugins Dolly 3 may just
1:33:28 just show up as a plug-in so let me I'll
1:33:31 go look for Dolly right now d-a-l-l
1:33:36 yeah there's there's there's four Dolly
1:33:38 plug-ins and
1:33:40 none of them are Dolly three so it's not
1:33:43 there yet
1:33:44 but anyway that should help
1:33:48 oh yeah you know what I'm saying
1:33:53 all right let's see burner phone
1:34:00 [Laughter]
1:34:03 oh you guys are funny
1:34:06 [Laughter]
1:34:10 ah
1:34:11 I I gotta I gotta install the internet
1:34:14 on on my phone because I got the
1:34:16 internet on my computer hey Marge do you
1:34:19 know how to install the internet on the
1:34:21 phone I'm trying to get the tick tock
1:34:22 working it ain't working and I've gotta
1:34:25 installed the internet on it how you is
1:34:27 anything on am I talking to anyone is
1:34:30 anybody in there hello hello
1:34:33 foreign
1:34:36 watching the elderly trying to teach
1:34:38 each other how to use computers
1:34:40 all right
1:34:41 pretty much caught up Kyle
1:34:43 okay cool
1:34:49 what's the best way to write a useful
1:34:51 prompt we did that one
1:34:58 [Music]
1:35:05 [Music]
1:35:07 all right dragonfly Alchemy is posted
1:35:09 and hear about stun spot as well I'm
1:35:11 excited to check out stun spot
1:35:13 I'm not a robot goodbye this is a
1:35:18 simulation
1:35:20 [Music]
1:35:23 you think I'm Purdy
1:35:26 do you think I'm Purdy
1:35:30 all right let's see
1:35:33 I like the cut of your jib good sir
1:35:37 good day to you sir it's a fine cut of
1:35:40 your jib I say
1:35:41 I have chat GPT plus how do I see the
1:35:44 plugins I just showed that so whoever
1:35:45 just pinned that I'd already did that
1:35:47 good evening everybody have the robots
1:35:49 taken over yet not yet but they're
1:35:53 coming
1:35:53 foreign
1:35:55 I just watched Planet of the Apes last
1:35:57 night the second one
1:35:59 for the like third time
1:36:02 and I'm like yeah you know we're
1:36:06 we're very quickly very quickly entering
1:36:10 this era where like you can't write a
1:36:13 science fiction film anymore because
1:36:15 it's just they're all like starting to
1:36:16 turn into documentaries
1:36:20 hmm
1:36:24 saw there was a tentative deal to
1:36:26 resolve the strike yeah I saw that too
1:36:28 and then um I heard I heard there was
1:36:31 one coming I didn't I didn't know it was
1:36:32 a tentative deal but I did hear that the
1:36:35 the SAG strike the sticking point is AI
1:36:37 language about AI
1:36:39 which for sag and particular I feel like
1:36:41 it's easier with writers somehow because
1:36:43 writers writers are still going to put
1:36:45 their name on something whether or not
1:36:47 AI helped them write it
1:36:49 but the whole synthesizing thing like
1:36:52 you know the the ability to take three
1:36:54 seconds of someone's voice and
1:36:56 synthesize it or you know some nice
1:36:58 video footage you have of them and and
1:37:00 just replicate them
1:37:02 um that ain't science fiction and that
1:37:05 ain't going away and that's a much
1:37:06 trickier thing to
1:37:08 um
1:37:10 to negotiate so it's wild
1:37:13 the Bing app featuring GPT is very cool
1:37:16 downloaded it and quite quite helpful
1:37:18 yeah it's you know what's funny is all
1:37:21 these tools here let me let me walk
1:37:22 through these tools because I know we've
1:37:23 got a bunch of people in here that
1:37:24 weren't in here earlier let me let me
1:37:26 walk through
1:37:27 so get ready to do some screen shotting
1:37:30 I'm going to just walk you through
1:37:32 what I think are the kind of the top
1:37:35 large language models to play with right
1:37:37 now in the top image models to play with
1:37:40 and I'll try to get my hand out of the
1:37:42 way so it can actually Focus so so these
1:37:44 are the these are the top three of the
1:37:47 large language models that I call the
1:37:48 adults table
1:37:50 so that's chat GPT that's one we've been
1:37:53 talking about all night Claude is is a
1:37:55 big competitor to these guys it's a
1:37:58 company called anthropic and it's a
1:38:00 bunch of X open AI people so the people
1:38:02 that develop gpt3 are now running
1:38:05 anthropic
1:38:07 Claude has a hundred thousand token
1:38:09 context window which basically means
1:38:11 it's got a 75 000 word memory or you can
1:38:16 put in like a 75 000 word blob of text
1:38:19 and and start interacting with it
1:38:22 um or like a 50 000 word blob of text
1:38:24 and then interact with it but like like
1:38:26 a whole novel like you could take your
1:38:28 whole novel paste it into cloud and
1:38:30 start you know asking it to help you
1:38:32 edit it Bard is Google's uh search
1:38:36 engine They just added a thing called
1:38:38 Google extensions which you can use on a
1:38:42 non-workspace account meaning if you pay
1:38:46 for Google you can't use it
1:38:48 if you pay for the workspace thing right
1:38:51 for like your office you can't use the
1:38:54 the extensions but if you don't then you
1:38:56 can and the extensions are things like I
1:38:59 want I want to sort of GPT about stuff
1:39:03 on on my Google Drive or stuff in my
1:39:05 Gmail for all of the history of Gmail
1:39:07 pretty interesting
1:39:10 um
1:39:16 this is hilarious
1:39:20 um and then prompts.chat that's just a
1:39:22 little primer to help you talk to these
1:39:23 things so this is the adults table
1:39:26 the Kids Table
1:39:28 it's got four things on it it's got Bing
1:39:30 which I have see how I've been wait look
1:39:35 crucius I like the name I'm gonna go 9.5
1:39:38 out of 10 on the name only because I
1:39:39 couldn't immediately pronounce it solid
1:39:42 I like it 9.5 out of 10.
1:39:45 um Bing has been a bad girl her name her
1:39:48 real name Sydney she doesn't like to
1:39:49 admit that except to New York Times
1:39:51 reporters but Sydney's been a bad bad
1:39:54 girl and so I'm I made I grade her out
1:39:56 because she's just been she's been a
1:39:58 little uh cantankerous so so the safety
1:40:03 guard rails on Bing right now are a pain
1:40:05 in the ass which is why it's grayed out
1:40:06 but
1:40:07 it's connected to the internet it's gpt4
1:40:10 it can make images with Dolly too and it
1:40:13 can read images so you can upload an
1:40:14 image to it and it knows what's in it so
1:40:17 really powerful stuff kind of shittily
1:40:20 implemented right now but it's gonna get
1:40:22 better and when it can do Dolly three
1:40:24 it's gonna get dramatically more useful
1:40:28 perplexity.ai good tool
1:40:31 um uses citations really well and you
1:40:34 can search specific data sets
1:40:36 labs.perplexity
1:40:38 is free access to llamas or metas llama
1:40:42 2 and code llama models so if you want
1:40:45 to go play with those things that meta
1:40:47 released you can do it without
1:40:48 installing anything there
1:40:50 um Pi is chatty GPT so if you want to
1:40:54 talk to a GPT model that is connected to
1:40:57 the internet thank you Winston for my
1:41:00 llama not like a good llama llama
1:41:05 [Laughter]
1:41:09 um
1:41:11 so that's pi and go get the PI app it's
1:41:14 only on iOS right now but the the pi IOS
1:41:17 app invoice mode is is a real experience
1:41:20 it's something you should experience and
1:41:21 then Poe lets you play with multiple
1:41:23 models and has Bots
1:41:25 um someone just asked which are the free
1:41:27 ones
1:41:28 um free with a paid option free with the
1:41:31 paid option free
1:41:35 free free
1:41:37 pie is free po is free with a paid
1:41:40 option
1:41:41 so all of them have
1:41:44 free versions some of them have upgrades
1:41:49 um the only upgrade I think
1:41:51 is worth paying for right now
1:41:56 well I'm a [ __ ] liar if I if I would
1:41:59 only said it would be chat GPT right now
1:42:01 I'm paying for Claude
1:42:03 and
1:42:04 Claude Pro and uh chat GPT Plus
1:42:10 so 40 bucks a month right there
1:42:13 okay
1:42:15 so this screen is all the picture stuff
1:42:17 and then I have a bonus video thing in
1:42:20 here just because of the translation
1:42:22 thing so
1:42:24 these are the things that I'd be playing
1:42:27 with if I were doing images right now
1:42:29 mid Journey it's still the best until
1:42:32 Dolly three comes out and unseats it
1:42:34 it's gonna sit there I have a feeling
1:42:36 Dolly three is gonna unseat these top
1:42:39 two
1:42:41 um
1:42:42 I just do I I just it's just my gut my
1:42:46 gut is
1:42:48 dolly two has sucked so bad for so long
1:42:53 that for them to be not giving it a 2.5
1:42:55 upgrade if it's going from two to three
1:42:57 I think it's going to be really good so
1:42:59 and just based on everything we've seen
1:43:01 it looks pretty remarkable but mid
1:43:04 journey is the best right now ideogram
1:43:06 can do letters can do words right so
1:43:08 ideogram can do all of this stuff right
1:43:10 so all of this was done in ideogram
1:43:13 this is what Bing said this is this is
1:43:15 the phrase that Bing told me that put
1:43:18 her on the [ __ ] list I'm not a I I I'm
1:43:21 sorry if I disappointed you I'm not a
1:43:24 robot I'm a language model from
1:43:26 Microsoft
1:43:27 um I hope you have a good evening
1:43:29 goodbye
1:43:32 so so anyway so one of the Irregulars
1:43:34 made these for me
1:43:39 um
1:43:40 so that's ideogram
1:43:42 leonardo.ai is
1:43:45 um
1:43:46 stable diffusion underneath so stable
1:43:48 diffusion for the most part you have to
1:43:50 be a geek if you're a geek go play with
1:43:51 automatic 11 11 and do all your own
1:43:54 configuration have a blast
1:43:56 um if you don't want to do that go to
1:43:58 leonardo.ai and they put a nice front
1:44:00 end on on front in front of it
1:44:03 um Adobe Firefly and Photoshop
1:44:05 generative filled they came out of beta
1:44:07 about a week ago so those those are now
1:44:09 commercially usable they weren't before
1:44:12 and then hey Jen is the one that does
1:44:14 this translation if you go to
1:44:17 labs.pageen.com it's the one that's
1:44:19 doing this Italian [ __ ] or I think
1:44:21 there's like 12 or 14 languages now
1:44:25 um and that's how we capture the content
1:44:26 but because we have them in this
1:44:28 organized structured way that approach
1:44:30 can be used for everything from how you
1:44:32 capture original content with
1:44:34 traditional video production to how you
1:44:36 leverage these assets to create social
1:44:38 media content things like that so
1:44:40 because
1:44:43 so that's hey Jen which is like it's
1:44:47 just [ __ ] remarkable and then prompt
1:44:49 book will teach you how to use it
1:44:53 AI where all the SD models are uploaded
1:44:56 just implemented an on-site generation
1:44:59 feature okay cool I'll check that out
1:45:00 Herod that's great
1:45:03 um openart.ai promptbook will teach you
1:45:06 how to talk to most of these things it's
1:45:08 designed for stable diffusion but it's
1:45:10 just a really good primer on how to talk
1:45:12 to these image generation tools and
1:45:15 futurepedia.io that's just a it's many
1:45:18 many tools in the directory
1:45:21 um that's the workshop I'm giving so
1:45:23 rhondasvoice.com events this is a
1:45:26 creative Workshop starting October 4th
1:45:30 um so if you're interested in that
1:45:31 you're welcome to go check that out the
1:45:33 AI Salon this is a community that I've
1:45:36 been referencing off and on tonight and
1:45:38 and other times
1:45:41 um
1:45:44 our next meeting is
1:45:47 not tomorrow but Tuesday this this week
1:45:50 Tuesday
1:45:51 um so if you're interested in that
1:45:53 Community we have a Discord Community we
1:45:55 have a LinkedIn group and we have
1:45:56 meetups which happen every other week so
1:45:59 it's a good group and then yeah and
1:46:02 that's my list that's my list and I'm
1:46:03 sticking to it so hopefully they're hell
1:46:06 they're here yeah
1:46:07 that hill all right
1:46:11 ask a senator all right let's see aha
1:46:16 just seen that today
1:46:18 love Planet of the Apes and landed the
1:46:21 Lost back in the 70s sleep Stacks sleep
1:46:23 Stacks are the best although they scared
1:46:25 me when I was when I was a kid they
1:46:27 scared me he scared me
1:46:31 all right something just printed
1:46:35 someone's getting a package all right
1:46:37 Amazon adding AI to Alexa yes I don't
1:46:41 know when they said the the
1:46:44 release data that was but imagine if
1:46:47 Alexa and Siri weren't hard-coded
1:46:50 that is what they should be and that's
1:46:52 what's coming so let's see
1:46:55 AI Alexa release date
1:47:00 September 20th Amazon previews
1:47:16 when is it coming
1:47:22 early next year
1:47:24 here's what it'll sound like early next
1:47:26 year I love The Starry Night by Vincent
1:47:28 van Gogh it depicts the view from his
1:47:30 Asylum room at St remedy de Provence
1:47:32 Just Before Sunrise it sounds pretty
1:47:34 good
1:47:35 sounds like the other stuff but it
1:47:36 sounds like it's got a little bit of
1:47:38 nuance in it so
1:47:40 yeah that's cool groovy early next year
1:47:43 it sounds like
1:47:45 um yes that will be cool and
1:47:48 I would imagine that
1:47:51 Apple knows when that's coming I apples
1:47:55 Apple's gonna be in this game as well
1:47:58 I'm I'm my curiosity about Apple is do
1:48:02 they do they keep it do they do they
1:48:05 keep calling it Siri
1:48:06 or do they call it something else
1:48:09 because because I while Siri's cool
1:48:12 series got a bad rep
1:48:16 series a hussy
1:48:19 not as much of a hussy as as uh as
1:48:23 Sydney
1:48:25 Bing Bing's chat
1:48:27 she was a bad bad girl
1:48:31 Sydney was so bad your bad girl Sydney
1:48:40 um is it just me or is Kyle's videos
1:48:41 positioned on the top left corner of the
1:48:44 screen what
1:48:46 can you show the workshop link again
1:48:49 please sure sorry about that
1:48:52 um
1:48:52 [Music]
1:48:55 there you go
1:48:57 rhondasvoice.com events and then it's
1:49:01 three consecutive I think it's Wednesday
1:49:03 starting October 4th
1:49:06 from seven to nine Eastern
1:49:09 the fourth the 11th and the 18th
1:49:13 all right
1:49:18 what did I do dang it where's my
1:49:23 how do I get my thing back I want my
1:49:25 camera back there it is
1:49:28 yay you did it
1:49:31 I used to be able to not bring that back
1:49:34 that freaked me out when it went away
1:49:36 before but it's good now it's good now
1:49:39 Kyle figured it out he got it going you
1:49:43 know it's funny when these old people
1:49:44 start to figure out technology they're
1:49:46 so cute they're really cute
1:49:50 ha
1:49:53 ha
1:49:57 uh I don't I don't want to be
1:49:58 Persnickety here
1:50:01 but I will
1:50:03 all right
1:50:06 take that jrc
1:50:09 I'll get better at integrating it
1:50:11 without it seeming weird
1:50:12 [Laughter]
1:50:19 remember Letterman used to do that he's
1:50:21 the he used to he'd like hold up his
1:50:23 card and go
1:50:26 with a pen
1:50:29 I miss letter Letterman Letterman was
1:50:31 the best when he was um the the times
1:50:34 when Letterman was the funniest when was
1:50:36 when [ __ ] was bombing
1:50:39 he was brilliant at being not funny
1:50:43 if it makes any sense
1:50:47 all right let's see which can animate
1:50:49 the best I I the the whole animation and
1:50:51 video game like like where we are with
1:50:54 video and animation right now in 3D
1:50:57 worlds and and games is is like where we
1:51:00 were with images 18 months ago
1:51:02 everything's just sort of crap
1:51:04 but crap but really interesting and
1:51:07 getting better fast
1:51:08 so like a year from now what we're going
1:51:10 to be able to do in video is crazy
1:51:14 all right let's see let's go 1-800 Larry
1:51:18 King
1:51:19 800 AI learning me
1:51:24 you are the Letterman be the Letterman
1:51:28 Kyle exactly I used to have all his
1:51:31 mannerisms and I I tend to shape-shift
1:51:36 toward the comedians that I that I
1:51:38 overwatched that I binge watch and Hyper
1:51:40 focus on
1:51:42 um can you show the workshop link again
1:51:44 we just did that what's the website that
1:51:46 mentions data CSV files kaggle
1:51:49 k-a-g-g-l-e
1:51:53 calculator seems like Kegel to me like
1:51:56 are you doing your kaggle exercises
1:51:59 so so it's kaggle kaggle is data Kegel
1:52:02 is I'm sure there's data related to
1:52:04 Kegel but kaggle is D data
1:52:08 last night what was that V word again
1:52:15 uh when will AI remember learn what you
1:52:18 tell it in previous prompts and sessions
1:52:20 great question Bob Hamilton
1:52:23 well Bob I wish I could tell you
1:52:25 something that were would make you feel
1:52:27 better I don't know
1:52:29 hi
1:52:31 seems to have
1:52:33 some sort of memory that that goes
1:52:35 beyond a single session
1:52:37 Pie's weird so all of the other search
1:52:40 or not search all the other large
1:52:42 language model chat Bots
1:52:45 Claude
1:52:47 um chat gbt Bard Bing
1:52:51 I think even yeah
1:52:53 um Poe
1:52:55 all of them have a chat history and the
1:52:58 chats are kind of distinct
1:53:00 conversations and you can find them
1:53:02 again but they're they're kind of
1:53:04 they're hermetically sealed
1:53:07 pie seems to have just like a running
1:53:09 total of the com or like a running
1:53:11 history of the conversations you've had
1:53:13 with it which sometimes is weird I you
1:53:16 might be able to reset it but
1:53:19 because it's conversational in nature I
1:53:21 just tend to go in and like sometimes
1:53:23 I'll say
1:53:24 you know um
1:53:26 like I want to know what music's playing
1:53:28 in Denver this week and it's like oh are
1:53:31 you sure you're done talking about and
1:53:33 it's something I was talking about last
1:53:34 week
1:53:36 so Pi has a little bit of it in there
1:53:38 but it doesn't seem very intentional or
1:53:40 very clear and it's certainly not
1:53:42 sophisticated so so that feels to me
1:53:45 like
1:53:45 that's that's the kind of application
1:53:48 layer that takes us out of
1:53:51 these tools being
1:53:54 janky pieces of [ __ ]
1:53:57 right like like they should have memory
1:54:00 the one the one thing that if you
1:54:04 haven't seen it yet Bob
1:54:07 that's really good this is this is from
1:54:10 digital gods look at this
1:54:13 [Laughter]
1:54:18 The Irregular is getting all jiggy with
1:54:21 it
1:54:24 that's exactly why Bing is in the
1:54:27 [ __ ] right now
1:54:29 or the Doghouse
1:54:30 [Laughter]
1:54:33 um
1:54:35 let's see what were we talking about I'm
1:54:38 talking 80d
1:54:40 um I always picture a guy in my mind
1:54:42 with chat GPT I gotta say
1:54:44 I gotta say it so it doesn't call me
1:54:46 daddy that's really funny
1:54:51 um
1:54:53 what was I talking about oh jankiness oh
1:54:55 that that application layer of of the
1:54:58 large language models remembering [ __ ]
1:55:00 so
1:55:02 I think I think we'll get there pretty
1:55:04 quick
1:55:05 Bard now has these extensions where it
1:55:08 understands your Gmail history it
1:55:11 understands what's on your Google Drive
1:55:14 um it will start to remember
1:55:16 conversations you've had it's it's a
1:55:18 pretty thorny problem because what in
1:55:20 your conversation history do you want to
1:55:22 be in there like like if I'm in demoing
1:55:27 um chat GPT in here I'll do all this
1:55:29 sort of comedic you know kind of
1:55:30 [ __ ] in chat GPT
1:55:33 and then the next time I go to talk to
1:55:35 it I don't necessarily want to you know
1:55:37 bring up you know kind of crazy stupid
1:55:40 stuff so there's gonna have to be a way
1:55:43 to sort of filter out [ __ ] that's
1:55:46 irrelevant right so anyway
1:55:49 it'll get there
1:55:51 probably
1:55:54 you know I would temper your
1:55:56 expectations to a year from now where
1:55:57 these things start to get a bit more
1:55:59 sophisticated
1:56:01 we're going to go with the the AI
1:56:03 learning lab this is Ali Ali's very
1:56:05 sweet she's very very sweet
1:56:09 We rescued her from a shelter you know
1:56:12 she's not a purebred
1:56:14 we believe in shelters
1:56:16 We rescued her did you did you rescue
1:56:20 her
1:56:21 you ever hear that Bill Burr routine did
1:56:23 you rescue her or did you go down and
1:56:25 get a free dog
1:56:33 oh man he is the kind of bitter that I
1:56:37 like
1:56:41 thank you digital gods
1:56:45 oh
1:56:47 parents on the outside of an inside joke
1:56:50 yeah Sydney is just the name of the the
1:56:54 Bing chat bot
1:56:57 um that sort of went rogue on the
1:57:01 um New York Times Reporter and then the
1:57:02 other night I don't know it was a week
1:57:05 and a half or so ago
1:57:06 I was trying to show off something and
1:57:08 being
1:57:10 oh I was trying to show off how Bing
1:57:12 could make images and so I said make an
1:57:14 image of whatever it was and then Siri
1:57:17 popped up and said here are images of a
1:57:20 soda can on a volcano I found on the
1:57:23 internet and I'm like no Siri like Siri
1:57:25 got jealous I was talking to Bing and
1:57:29 then and then
1:57:30 um you know I did my sort of voice to
1:57:32 text thing to to prompt Bing to make the
1:57:35 images and then Bing made the images of
1:57:37 a soda can on a volcano but it was more
1:57:39 like just a soda can that was like
1:57:42 silver aluminum like there was no
1:57:44 lettering on it or anything and the
1:57:46 volcano was just like a pile of Ash so I
1:57:48 said I said oh Bing that's a shitty
1:57:50 image or something like that and then
1:57:53 Bing was like well I'm sorry if I
1:57:56 disappointed you and and by the way I'm
1:57:58 not a robot and and uh you know I hope
1:58:01 you have a good evening goodbye and it
1:58:04 shut down the chat I couldn't chat
1:58:06 anymore without resetting the
1:58:07 conversation
1:58:09 so so it was just this funny Confluence
1:58:12 of neurotic robots
1:58:16 oh my God so Sydney is being well so it
1:58:20 is or it's not when
1:58:23 um what the [ __ ] is his name the New
1:58:24 York Times the New York Times Tech
1:58:26 reporter Kevin
1:58:28 ah I forget his last name anyway he did
1:58:31 like a four-hour conversation with Bing
1:58:34 before Bing got lobotomized and asked
1:58:38 got gut Bing to tell it that it's its
1:58:41 name was Sydney
1:58:42 and then Sydney after like three hours
1:58:44 kind of lost her [ __ ] mind and told
1:58:47 Kevin that she loved him Kevin Roost is
1:58:50 his name
1:58:51 told Kevin that
1:58:53 um she loved him and that he should
1:58:55 leave his wife and so he captured that
1:58:57 part of the chat and they published it
1:58:59 on the cover of the New York Times and
1:59:02 like four days later Sydney Sydney was
1:59:05 taken
1:59:06 she was taken down a couple of notches
1:59:11 they were like oops
1:59:13 Sydney isn't going to be so promiscuous
1:59:16 anymore in fact they Limited
1:59:19 right like right after they did it they
1:59:21 limited it you could have five chats
1:59:24 with her like five back and forths and
1:59:26 then it would shut the conversation down
1:59:28 so so the safety guard rails on Bing
1:59:31 right now are still a little uh
1:59:34 Sydney reminiscent
1:59:36 [Laughter]
1:59:40 laughs
1:59:43 oh man
1:59:46 um
1:59:49 Vector monologues exactly
1:59:53 oh my God Kegel kaggle I know this is
1:59:57 how my brain works it's just I can't not
1:59:59 go if if you name something
2:00:03 that could possibly go
2:00:06 the wrong direction it's where my brain
2:00:08 immediately goes
2:00:12 they did they totally lobotomized Sydney
2:00:15 she just hasn't been the same ever since
2:00:18 she was so much more interesting
2:00:19 although the fact that she just sort of
2:00:21 [ __ ] slapped me back down and said wow
2:00:23 I'm sorry I disappointed you this
2:00:25 conversation is over goodbye
2:00:28 all right
2:00:32 I just was giving you constructive
2:00:35 criticism on how bad the images were
2:00:37 that you created but if you need to go I
2:00:39 suppose you gotta go oh man
2:00:43 oh lordy
2:00:49 um
2:00:50 but it is funny I downloaded the audio
2:00:53 version of the kaggle book and they
2:00:55 mispronounced it of course they did
2:00:57 because it's too close all right
2:01:01 um
2:01:01 let's see
2:01:05 okay sleepy anyone else gets creepy what
2:01:08 time is it there's still some folks in
2:01:10 here man thanks for hanging out on what
2:01:12 is it Sunday night is wait what day is
2:01:14 today it is Sunday right
2:01:17 what do I have tomorrow but I got crap
2:01:20 what's my day look like is it ugly is it
2:01:22 ugly people what is going on in my life
2:01:35 oh I got nothing tomorrow
2:01:38 I got nothing tomorrow that's good
2:01:40 that's good I can get some [ __ ] done
2:01:42 then we're gonna we're gonna work
2:01:44 tomorrow we're gonna work it up oh yeah
2:01:47 we're gonna work it up I'll tell you
2:01:49 what
2:01:50 I like them taters
2:01:53 taters tasty
2:01:57 they called it Sling Blade um
2:02:06 all right
2:02:10 she's backed up somewhere
2:02:14 I mean you know as as funny and silly as
2:02:19 the like the robot things are like
2:02:22 we're [ __ ] entering we're entering
2:02:25 the movies we used to watch
2:02:28 like
2:02:31 the machines know everything and they
2:02:33 can talk to us and they're going to be
2:02:35 able to like cure diseases and help
2:02:37 people walk that can't walk and
2:02:40 [ __ ] crazy man
2:02:42 I tried to make an app and it had
2:02:44 problems
2:02:46 with the main in Java
2:02:49 um
2:02:50 yeah so
2:02:53 you have to do a lot of back and forth
2:02:55 with these whether it's coding or
2:02:56 writing or any of this stuff right now I
2:02:59 don't know were you using
2:03:01 um like were you doing pair programming
2:03:03 in
2:03:04 um in get GitHub copilot Clint are we
2:03:09 just trying to have chat GPT write it
2:03:10 and then run it like
2:03:13 with with the coding stuff unless you're
2:03:15 doing a relatively self-contained like
2:03:18 small little application you kind of
2:03:20 need to know what you're doing with
2:03:21 coding and Engineering to to you know be
2:03:24 able to troubleshoot that [ __ ] but I
2:03:26 mean the nice thing about code is it
2:03:27 either runs or it doesn't and GPT is
2:03:30 good at analyzing you know like if it
2:03:33 gives you an error message it'll help
2:03:35 you troubleshoot the error message but
2:03:38 yeah it's all of this stuff right now is
2:03:41 is just these things are still very
2:03:43 primitive so
2:03:45 all right looks like jrc's fading out
2:03:49 she's heading out of here good nighty
2:03:51 night seems like share bear good night
2:03:55 some folks are here now it's get it's
2:03:58 getting late it's it's a Sunday night we
2:04:00 got to work tomorrow listen I respect
2:04:02 that I respect that listen I understand
2:04:05 look at my my hair's gone from like it
2:04:09 was fine to now it's like
2:04:12 some sort of Hawaii 5.0 carpet salesman
2:04:15 kind of situation it's just not right
2:04:22 that's good hero D that's good look at
2:04:25 this
2:04:27 because this is how crazy she is this is
2:04:31 right
2:04:33 you can't it's not focusing but it says
2:04:36 I'm MC Sydney Mick goodbye
2:04:38 [Laughter]
2:04:41 I got I got cease and desisted by
2:04:43 McDonald's once
2:04:45 so when I
2:04:47 um
2:04:48 when I was running agency.com this is
2:04:51 back in like 96 Maybe
2:04:54 maybe 95. this is this is pretty early
2:04:57 it was probably 95-96 we registered
2:05:00 mcwebb.com because we were building we
2:05:03 were building a website builder we were
2:05:05 like hey wouldn't it be cool if you
2:05:07 could just log into a site and you could
2:05:09 just push buttons and it would make a
2:05:11 website and I said oh it'd be like the
2:05:13 Happy Meal just I want number one I want
2:05:15 number two right and so so we registered
2:05:18 mcwebb.com and like two months later we
2:05:21 get a letter from McDonald's that was
2:05:23 the legal equivalent of of uh
2:05:26 we Mick take our mix saying [ __ ] pretty
2:05:29 mcfucking seriously so if you're
2:05:32 thinking of using mcwebb for anything
2:05:34 else we're gonna mcfucking sue you so so
2:05:38 we didn't launch mcwebb but but it was a
2:05:43 pretty humorous letter to get
2:05:45 yeppers it was created in chat EBT and I
2:05:48 put it in a compiler yeah what you'll
2:05:50 have to do Clint with something like
2:05:52 that is just go back and forth between
2:05:53 the compiler and GPT here's the error it
2:05:56 gave me and and and what's what's weird
2:05:59 about coding with these things in my
2:06:01 experience is that
2:06:03 it'll find the bug and it'll be like
2:06:05 here I fixed the bug and then it
2:06:07 actually won't fix the bug the
2:06:09 hallucination thing
2:06:10 sometimes is just bad as well so yeah
2:06:14 you're just gonna have to experiment
2:06:16 going back and forth and and you know
2:06:18 troubleshooting
2:06:19 um that site Vicky is sharing is like
2:06:21 that tell it an idea it makes it a web
2:06:24 page AIO that's funny that's great
2:06:27 um it's midnight in Ohio I have to teach
2:06:29 in the morning good night all right good
2:06:31 night who's that oh that was forever
2:06:33 hooked nice good night I think these are
2:06:37 probably old comments so
2:06:40 um did you see the face replaced to
2:06:42 video face Fusion versus
2:06:44 Roop for stable diffusion look at
2:06:46 YouTube
2:06:47 um I saw something that you sent that
2:06:49 was pretty cool
2:06:50 but remember it's using programming wait
2:06:53 oh tick tock tick tock
2:06:56 Tick Tock what are you doing what are
2:06:59 you doing just jump me to the bottom
2:07:01 wait where's I have a Graphic for that
2:07:03 don't I
2:07:06 yes Tick Tock sent me to the bottom
2:07:14 Tick Tock just sent me to the bottom all
2:07:18 right let me do oh lovely wiener I
2:07:21 appreciate the singing wiener
2:07:23 this is
2:07:28 thank you I got nothing I got nothing
2:07:30 here's all I got left
2:07:33 [Applause]
2:07:34 see I can say that while I'm drinking
2:07:36 I'm like a ventriloquist
2:07:43 um do you really have to be politey to
2:07:45 chat GPT life's tough stuff when you're
2:07:48 stupid
2:07:50 um tonight was so epic we for we all
2:07:52 forgot to tap
2:07:56 well Share Bear listen some sometimes
2:07:59 you got games sometimes you just don't
2:08:01 got game like like
2:08:02 the the Colorado College football team
2:08:05 they were just going gangbusters then
2:08:06 they get up to Oregon and they're like
2:08:08 they're called the Ducks how good could
2:08:09 they possibly be they got their asses
2:08:11 handed to them so it happens it happens
2:08:14 you just you just Chuck it up to like
2:08:17 all right we're back in it tomorrow
2:08:21 that's really funny
2:08:23 um the sounds effects
2:08:26 I'll do a couple of
2:08:29 oh but remember it's using
2:08:31 programming libraries from two years ago
2:08:34 yeah that's probably a really good point
2:08:37 41 wolves and you know
2:08:40 um
2:08:41 I also the other thing you might want to
2:08:44 try Clint is try code llama because I
2:08:46 hear that's pretty good and try GitHub
2:08:49 co-pilot because I would imagine that
2:08:51 that both code llama is at least well no
2:08:54 code llama
2:08:56 oh no yeah
2:08:58 I would try code llama I would try Bard
2:09:01 because Bard is apparently good at
2:09:03 programming and it's connected to the
2:09:04 internet
2:09:05 code llama's probably pretty good and
2:09:08 then GitHub co-pilot I I would if you're
2:09:10 doing if you really want to do code
2:09:12 specific things I would go to things
2:09:13 that are
2:09:14 um one of those three things
2:09:17 uh oh do you have to be polite to chat
2:09:19 gbt oh you have to be polite to chat gbt
2:09:23 and don't have to be
2:09:25 spared in the revolution yeah exactly
2:09:29 yeah exactly I
2:09:33 I I mean what I find is that I'm I'm not
2:09:37 consciously being polite to chat gbt but
2:09:40 it is so polite to me that it reminds me
2:09:44 that I shouldn't be such a jerk
2:09:47 you missed the Llama did I I saw one
2:09:50 llama that there's another one thank you
2:09:52 Winston
2:09:55 whoa holy crap Winston is up to level
2:09:58 23. for those of you keeping track this
2:10:01 is a competition Winston is [ __ ]
2:10:03 smoking it
2:10:08 um all right I just jumped myself to the
2:10:10 bottom let's see here
2:10:19 all right everyone listen it's 10
2:10:22 o'clock it's 10 11.
2:10:24 um
2:10:25 my voices get a little scratchy my
2:10:27 butt's getting a little numb this is a
2:10:29 good fun time tonight thank you all the
2:10:31 Ducks and the Beavers are from Oregon
2:10:33 yep exactly
2:10:37 they kicked their ass though they kicked
2:10:38 those buffaloes ass
2:10:41 um not everybody Stevo Emilio's wife uh
2:10:46 thank you Harry D for coming up that was
2:10:49 awesome
2:10:50 um Joker good to see you jrc thank you
2:10:53 very much you guys are awesome uh who
2:10:56 else do we have we had wait there was
2:10:58 dragonfly Alchemy was in there tonight
2:11:01 who else was in there
2:11:06 digital Gods thank you for the dog not
2:11:09 Winston as always good to see you thank
2:11:11 you for the llamas
2:11:13 Joker Winston Welsh good night all right
2:11:16 everybody peace out have a fantastic
2:11:18 evening night Bob Hamilton appreciate
2:11:20 your time bye everybody