aiLL 8/5/23 - The Future of Work in an AI World

Description

Chapters

Transcript

0:01 foreign
0:34 Silver Fox what's happening
0:37 all right
0:38 it is
0:40 Saturday night
0:43 I just tried to sign up for a service
0:45 and
0:49 you had to do it through PayPal I
0:51 haven't logged into PayPal in Forever
0:53 when it said I wasn't verified but I
0:56 have all my crap verified there so
0:59 [Laughter]
1:02 I hate it when crap doesn't work
1:05 uh welcome to AI learning lab everybody
1:07 hey hey
1:10 um happy Saturday night my name is Kyle
1:13 Shannon we'll get this party started in
1:16 a little moment
1:18 I have exciting news I started a YouTube
1:21 channel for Ye Olde learning lab
1:25 so I'm gonna put my uh the recordings of
1:28 these and I'll probably put over
1:30 everything if I can get it all automated
1:33 I couldn't automate downloading the
1:35 lives I don't think I might be able to
1:38 but I couldn't with
1:39 whatever tool was recommended last night
1:44 but I'll show that
1:47 I think that's a good thing I think
1:50 that's a good thing I actually thought
1:52 it'll be cool to use
1:54 um
1:55 maybe we can even do that tonight I
1:56 wonder if any of the transcripts are
1:58 there yet
2:00 I figured I could use
2:02 um
2:06 let's see seven hours ago
2:09 I figured I could use Ye Olde chat gbt
2:15 show transcript yeah
2:19 all right the transcripts there yeah so
2:22 we can go play with this tonight maybe
2:24 we'll go maybe we'll go experiment we'll
2:27 go spearmint with uh chat gbt because
2:30 what I thought would be cool
2:32 is use chat gbt to go
2:35 look at the transcript of these long ass
2:38 ramblings
2:39 and come up with a title for it come up
2:42 with a description and then come up with
2:44 a uh a set of chapters with time code in
2:48 them
2:48 I think that would be cool I've been
2:51 screen recording a lot of your stuff and
2:53 I'm working on an app to organize stuff
2:55 wow that's awesome
2:59 you're gonna sell me back my app that's
3:01 got my crap organized that would be nice
3:05 that's awesome oh that's hilarious
3:10 um that's cool yes definitely
3:12 um yes send me what you're working on I
3:14 I always like seeing seeing that because
3:16 it's there's so many ways to skin the
3:18 cat with this stuff so I love I love
3:20 seeing different things
3:23 no charge thank you I appreciate that
3:26 I appreciate that
3:29 um oh wow a bunch of people here all
3:30 right welcome so here's what this
3:32 channel is about my name is Kyle Shannon
3:33 I'm a long time entrepreneur I've
3:35 started 12 or 15 companies I'm currently
3:38 the CEO of a company called storybine
3:41 the founder and the CEO of that uh we've
3:43 been around for 11 years we do automated
3:45 video storytelling so people answer a
3:48 series of questions in an app that goes
3:50 up to the cloud a few minutes later you
3:51 got a fully edited video
3:54 and uh
3:56 over the past years I've been digging
3:58 into this AI stuff I'm completely
4:00 rethinking my value proposition and the
4:02 product roadmap so that's been
4:03 absolutely insane
4:05 um I was also an entrepreneur uh
4:08 a startup guy
4:10 back in the early days of the world wide
4:12 web and what's going on with this
4:15 generative AI stuff the parallels
4:17 between what was happening with the
4:18 early days of the web are absolutely
4:20 insane so this channel is all about uh
4:24 just talking about this generative AI
4:25 stuff it's it's very new it's very raw
4:28 it's very profound
4:30 um
4:31 this list is now seven this use this
4:35 when I first started the channel there
4:36 was one site here it was it was chat GPT
4:39 and then I added some others it was
4:42 three for like two months three months
4:44 and now it's seven
4:46 so the the competition is heating up and
4:49 these are all just kind of chat flavors
4:51 of chat gbt or competitors
4:54 um so there's a lot going on so if
4:57 you're if you're new to generative AI if
4:59 you haven't played with it that's what
5:01 this list is all about so let me let me
5:02 run through these
5:11 all right
5:12 so go ahead
5:15 and take yield screenshot if you're new
5:17 to this stuff if you're just getting
5:20 started you don't want to think about
5:21 too much just go to the top one go to
5:24 chat.openai.com that's the official
5:26 chatgpt website the 3.5 version is free
5:30 it's really good
5:32 um if you pay the 20 bucks a month there
5:34 you get gpt4 and you get a couple other
5:36 things that I think are worth it I'm not
5:38 getting paid by them so I don't give a
5:40 [ __ ] if you do that or not
5:42 um then you've got Bing and Claude which
5:44 are kind of the next two down that that
5:46 I would I would consider Bing is
5:48 Microsoft search engine but if you click
5:50 on the chat button that's gpt4
5:53 uh you used to have to do it within Edge
5:56 but I think they've opened it up to
5:57 Chrome and Safari now it is connected to
6:00 the internet it can generate images it
6:03 can read images
6:05 uh it's it's decent it's not as good as
6:08 the in in terms of the
6:12 the content creation it's not as good as
6:14 the gpt4 you get here but it's free then
6:17 you've got Claude AI over on the left
6:18 there that is anthropics
6:23 um Claude 2. and it has a hundred
6:26 thousand token context window which if
6:28 you're new to this stuff just doesn't
6:29 matter all you need to know is you could
6:32 copy and paste an entire novel into your
6:35 chat prompt and start interacting with
6:38 it like tell it to summarize the novel
6:41 um claude's really interesting
6:44 um Bard is Google's entry into this
6:46 space this did not used to be on the
6:48 list because it was embarrassingly bad
6:49 for a while it's it's getting better
6:52 and I thought that they actually rolled
6:56 a new large language model called Gemini
6:58 into it I don't think they have now that
7:00 I now that I've said that in about 30
7:02 different places but it's much better
7:04 than it used to be and when they rolled
7:06 Gemini out I think it's going to get
7:08 really good uh
7:10 labs.perplexity.ai that's actually you
7:13 can those are uh the metas Llama Llama 2
7:18 models uh are being hosted by perplexity
7:21 that you can play with for free
7:24 um so you don't have to install them and
7:25 do all that happy horse [ __ ] and then
7:27 pie in the corner is from a company
7:29 called inflection it's very very chatty
7:32 if you actually want to have a
7:34 conversation with a large language model
7:35 that's a good one and then Poe you can
7:37 play with a bunch of different large
7:39 language models and it's got Bots that
7:42 you can make it's got a bot maker and
7:44 really interesting
7:45 just just in
7:47 it both of the bottom ones feel a little
7:51 more consumery where all these upper
7:54 ones feel more professional all right
7:57 so that's going on there okay what else
8:01 I got a lot this is like this is like
8:02 homework this is the homework portion of
8:04 the evening happy Saturday night let's
8:07 go on ahead and get fired up with uh
8:09 homework
8:11 okay so here's your homework if you're
8:14 new to this crap
8:16 um prompts.chat will teach you about
8:18 prompting it will also give you a whole
8:21 list of personas that you can tell
8:24 Chachi PD to act like certain things
8:26 like act like an accountant act like a
8:28 screenwriter act like an idiot and it
8:30 will
8:31 um and then openart.ai promptbook
8:34 that'll teach you how to prompt these
8:35 image generation tools and I I recommend
8:38 these three right here if you're just
8:39 getting started and then futurepedia.io
8:42 will give you just a a big old pile of
8:45 AI tools that you can go geek out on
8:49 um I have some news I have news
8:53 we now have
8:56 um a YouTube channel
8:58 that I just started today
9:00 and AI learning lab was taken so it's
9:03 YouTube slash
9:06 um AI learning lab Dash TT for an old
9:09 tickety talk
9:11 and uh we can go over there and take a
9:14 look at it
9:17 look I made I made a little uh
9:20 a little Banner
9:22 AI learning lab and I put my little
9:25 eyeball dude in there and I uploaded
9:27 nine and then it it uh rate limited me
9:30 it wouldn't let me upload more than
9:32 nines I have nine up there
9:34 um so those are the those are the early
9:36 ones if you want to see hey how fat I
9:38 was when I started this channel three
9:41 months ago I've lost some weight uh I I
9:45 look a little beefy I'm chubbier than
9:47 the dog in these in these videos uh you
9:50 can go check that out but anyway people
9:52 have been asking you know if these uh if
9:54 these lives are being saved they're
9:56 going to be I need to find a way to
9:58 automate it uh right now it's a manual
10:01 pain in the ass it's a pain in the knee
10:03 is that tell you opinions
10:08 um
10:09 what I figured we could do tonight also
10:12 is
10:14 um the videos that I uploaded have
10:16 transcripts with them so I figured we
10:18 could go play with some plugins that go
10:20 look at YouTube videos
10:22 and maybe we could write descriptions of
10:26 um
10:29 descriptions of of the uh of the old
10:32 lives let's see how good it does at
10:34 making sense of me just rambling for two
10:36 hours maybe it can chapterize them in a
10:38 decent way and actually make those
10:40 things usable
10:42 and discoverable and interesting
10:44 that would be nice wouldn't it wouldn't
10:46 it be nice if this AI technology
10:48 actually helped us instead of destroyed
10:51 our lives and
10:52 killed all of humanity
10:55 might be good all right so what do we do
10:58 here I don't know a ramble I swear I
11:01 make jokes I make voices occasionally my
11:05 jokes and voices get me kicked off this
11:07 thing for seven days tapping three times
11:09 now I'm gonna try to avoid that tonight
11:12 um and if you've got questions about AI
11:14 pop them in the comments below I do
11:16 spend
11:17 most of the time on these answering
11:20 questions or comments or thoughts that
11:22 you have down below
11:24 um I think this stuff's really really
11:25 important I'm really passionate about it
11:27 passionate enough that I hop on here for
11:29 a couple hours every night instead of
11:31 doing Sim racing which I love and I'm
11:33 giving up Sim Racing for you
11:36 for you to get you up to speed on all
11:39 this it's crazy
11:41 all right
11:43 let's see let me go back and see what
11:46 questions we have here what comments we
11:48 have no comments about the hair although
11:50 it's it looks like it's about ready to
11:52 take off it could go bad pretty quickly
11:54 here tonight
11:56 all right all right
12:00 let's see you remind me of Tom Arnold
12:03 I'm sure you've heard that before I have
12:06 Tom Arnold Rosie O'Donnell uh fat Alec
12:10 Baldwin
12:12 um Philip Seymour Hoffman
12:15 uh uh um wait who's the other guy John
12:19 Goodman skinny uh skinny John Goodman uh
12:26 someone said Leonardo DiCaprio once and
12:29 I thought well yeah Leonardo Leonardo
12:31 DiCaprio if he made much worse life
12:33 decisions Maybe
12:36 laughs
12:39 if he didn't get that big break when he
12:41 was
12:42 you know who's Eating Gilbert Grape
12:46 uh I've heard some of that stuff before
12:50 um yeah that's cool uh
12:53 add it to Amazon Bedrock what's add what
12:56 to it I assume uh that
12:59 the uh
13:01 llama llama two Labs without perplexity
13:04 is that what you're talking about
13:05 yeah Amazon bedrock's interesting I'm
13:08 I'm uh I'm building a I'm going to New
13:11 York in
13:13 like 10 days yeah yeah eight days I'm
13:18 going to New York in eight days to talk
13:19 to a big Pharma client about
13:21 building some some tools for them and uh
13:24 I think Amazon Bedrock might be a
13:27 might be a big piece of that I don't
13:30 know if it'll be a part of the
13:30 conversation it might be a big part of
13:32 how we solve the problem
13:35 Amazon Bedrock is you can use it's
13:39 basically you can hook all your AWS [ __ ]
13:42 up to large language models and use
13:44 different models and things like that
13:46 um I don't know how far along it is or
13:49 how good or bad it is I haven't really
13:50 dug deep into it yet
13:52 uh
13:54 um
13:56 did I see that Skype
13:59 now has Bing chat built into it no I
14:02 haven't seen that that's interesting
14:05 let me go let me go looky looky at Skype
14:07 and see how how much they're hyping it
14:09 or if they just kind of snuck it in
14:10 there
14:22 Hawaii Five-O someone someone include me
14:25 and that song just popped in my head
14:26 last night I tend to have ADD singing
14:29 moments when I'm going and searching for
14:31 things
14:32 introducing new Bing in Skype
14:35 interesting do they have a little video
14:37 of it
14:40 get social
14:42 wait sorry
14:44 get social with the new Bing in Skype
14:46 the new Bing in Skype is your AI powered
14:48 co-pilot
14:50 can provide helpful answers to your
14:52 questions whether you need a quick
14:53 answer on something or something more in
14:55 depth
14:56 learn more they got a cool little video
14:58 for us to watch
15:03 download Skype now just ask Bing
15:07 hmm
15:08 interesting all right
15:10 yeah so it looks like they just popped
15:12 it in there as part of the
15:16 refined things answers
15:19 they've got it writing poems
15:24 well we're so [ __ ] early in this in
15:27 all this all this stuff being integrated
15:29 and so a lot of it's just parlor tricks
15:31 right now look you can make poems yeah
15:34 okay
15:39 um interesting well I I don't know
15:41 anything about it so I'm seeing it for
15:43 the first time now
15:44 um but but that's why I think the
15:46 there's an important Point here actually
15:49 um
15:50 the the the thing the thing that's
15:52 significant about that is is you know
15:55 it's it's further evidence that
15:58 um
15:59 these tools are just going to be
16:01 ubiquitous they're going to be kind of
16:02 like uh the World Wide Web where just
16:05 you know it's it's out there for
16:06 everyone to use it's going to get rolled
16:08 into all the you know all the software
16:11 we already use every day
16:13 um it's not going to be this thing off
16:14 to the side like it is right now like
16:16 you know chat gbt is like this you know
16:19 we've got you know AI detectors and you
16:22 know
16:23 we we've got to figure out if it was
16:25 written by AI well no we don't it like
16:29 if
16:31 if we're doing things like writing
16:32 articles that have citations in them as
16:34 human beings we need to confirm that the
16:36 citations are correct because these
16:38 things lie
16:40 um but but in the end it doesn't really
16:42 matter if a human wrote it or if this
16:43 wrote it if it gives you value right
16:45 that was Google's initial response to
16:48 Chachi PT taking off was you know we're
16:51 gonna
16:52 de-prioritize anything that was written
16:54 by Chach EBT and then they were like
16:55 yeah no we're not going to do that if if
16:57 it adds value if it's valuable content
16:59 it's cool if it's shitty content then
17:02 we'll de-prioritize it and I think
17:04 that's the I think that's the the right
17:07 the right approach but this stuff's
17:09 going to be everywhere so it behooves
17:10 all of us uh to get much more
17:14 um
17:16 uh AI literate in the next year we'll
17:19 we'll pay dividends in the next year
17:21 year and a half
17:22 um if you would since I'm just getting
17:24 this started tonight if you're willing
17:26 to share this live with your friends and
17:29 let them call you a weirdos for playing
17:31 with this AI [ __ ] that would be great uh
17:33 and you know tap the uh
17:35 double tap the screen get some hearts
17:37 going maybe we can get some more folks
17:39 in here that'd be awesome
17:40 uh
17:42 um
17:42 what else what other questions do we
17:44 have
17:45 what's the best one to use on an iPhone
17:48 um
17:49 the the
17:51 official chat gbt in my opinion is just
17:54 it's just the best so I there there is
17:58 an official iPhone app from openai for
18:01 chat GPT
18:02 um it's a very simple website so it
18:04 actually I just have it as a uh a
18:07 bookmark to the website on my home page
18:09 I actually have it in the bottom four
18:11 like it's one of my primary buttons on
18:13 my iPhone
18:15 um the other one that's really
18:17 interesting here is pi dot AI because
18:20 you can put it in voice mode and just
18:23 talk to it and then you pick one of six
18:25 voices and it talks back to you and it's
18:28 connected to the internet
18:30 um chat EPT is not connected to the
18:32 internet although with plugins it kinda
18:33 is and I can I can show that later
18:37 um I would say for the iPhone right now
18:40 chat GPT
18:42 clawed just because it's got a really
18:45 big context window and that might be
18:47 useful if you want to copy and paste big
18:48 documents even on your iPhone and then
18:51 uh Pi dot AI the thing about Pi dot AI
18:55 if you want to like like like the way
18:57 I've been using pi.ai on my iPhone is I
19:00 get in the car I turn on the Bluetooth
19:02 and I turn on pi and just as I'm driving
19:04 places I just start having conversations
19:06 with
19:07 with this large language model and it's
19:10 fascinating it's very very chatty so if
19:12 you actually want to get [ __ ] done don't
19:14 use Pi it's just too chatty for that
19:17 it's like haha isn't that clever Kyle
19:20 you know it's very funny
19:23 um but but to have a conversation with
19:26 it it's it's actually quite interesting
19:28 you know after whatever it's been 10 15
19:30 20 years of things like
19:32 um Alexa and apples talk to me crap have
19:38 sucked
19:39 um it doesn't suck it's actually you
19:41 know it's a little awkward
19:44 um I've got I've got it set to some
19:46 Australian woman voice and she pauses a
19:48 little too long before she answers but
19:50 it's pretty interesting so that's what I
19:52 play with
19:54 all right
19:56 um pie pies tasty Chachi BT available
20:00 for Android Android uh iPhone and
20:04 Android yeah they just announced Android
20:08 if you're gonna download the chat GPT
20:10 apps be very very careful I would
20:13 strongly recommend you go to openai.com
20:16 and click on the links for Android or
20:19 iOS from their site
20:21 rather than searching for it in the app
20:24 store or in Google Play because there
20:27 are so many copycats that are like just
20:30 trying to get your money that are
20:32 basically charging you money to give you
20:34 free GPT 3.5
20:36 so don't fall for that [ __ ] little
20:39 [ __ ] scammers are out there ah we can
20:41 make money off to clueless people
20:44 so go there
20:47 um but yes they're available for both
20:49 the the challenge right now with uh with
20:52 the apps from open AI although I haven't
20:54 played with them to be fair in the past
20:57 three or four weeks
20:59 um they don't have all the features of
21:01 the website so you just need to check
21:04 you know go to the website see what
21:06 features are there what's available and
21:08 then see what's available on the uh
21:11 on the apps what I will say is that
21:15 the voice recognition
21:18 in chat GPT the official app uses their
21:22 whisper protocol the for to do the trip
21:25 the the transcription of your voice it's
21:27 really good
21:28 so
21:30 um yeah so I would you know
21:33 you could you could use it to do a voice
21:36 transcription much more accurately than
21:38 the the iPhone local stuff
21:40 good good good little life hack there
21:42 huh
21:44 delivering value right off the bat right
21:47 off the bat at the lab the AI learning
21:49 lab what we do here is all value all the
21:51 time it's just it's a Relentless pursuit
21:54 of value of Excellence of
21:59 um
22:01 wow
22:02 wow
22:03 All Right Moving On
22:06 they have an app now yes they do
22:11 just jumped me to the bottom
22:14 if you're new here it's one of my pet
22:15 peeves is Tick Tock jumping me to the
22:18 bottom I like to answer all the
22:20 questions I don't like to be at the
22:21 bottom I like to be in a place where you
22:24 don't think I'm actually going to get to
22:25 your question and chances are good I
22:27 won't
22:27 [Laughter]
22:31 ah he's so silly he's very silly he's
22:35 silly I don't you know I come in here
22:38 and and I just I put it on and it I like
22:41 to go to sleep to his voice I like his
22:43 voice his voice is really nice and uh
22:46 sometimes he'll make me laugh other
22:48 times it's a little cringy a lot of dad
22:51 jokes a lot of dad jokes he's not really
22:53 qualified to talk about this stuff but
22:56 he seems to like doing it so uh yes I
22:59 think it's good
23:02 all right
23:03 I've never used that but that voice
23:06 before ever in my life kind of hurt
23:13 [Laughter]
23:16 Tobias here with his plus one awesome
23:18 Tobias is your plus one a puppy or a
23:21 human or a kitty cat or a goldfish
23:27 welcome welcome welcome
23:29 what'd you do on date night uh I want to
23:33 talk about it
23:36 no you didn't tell me that's probably
23:38 best if you don't know now tell me
23:40 and we watched this guy on Tick Tock for
23:42 two hours what
23:43 [Laughter]
23:52 uh down
24:00 I downloaded a chat gbt and it was a
24:03 fake I couldn't close it uninstall it oh
24:06 that sucks bolognes I know that's that's
24:08 again like why I put these here this is
24:11 not for me to make money I'm not getting
24:12 paid by any of these [ __ ]
24:15 um they should pay me
24:16 you know all we do all they just sit
24:19 here and talk about them someone should
24:21 be writing a check but um
24:24 yeah that sucks it's one of the reasons
24:26 I have them here because there's just so
24:27 many scammers out there trying to
24:30 trying to take advantage of the uh
24:32 excitement about the future of
24:34 communication
24:37 um what game do you use for Sim racing
24:39 assetto Corsa competition
24:44 GT3 stuff
24:46 and I raced the McLaren
24:51 I'm thinking of switching I might switch
24:53 to the Honda
24:55 and I might switch to a Porsche but I
24:58 don't I don't think the Porsches are
24:59 that good they're easy to lose lose
25:01 control of
25:03 what's good what's good uh
25:07 oh here's something that's good this is
25:09 new
25:10 um people have been asking but Kyle do
25:12 you have a YouTube channel and I have
25:15 not until today we do now
25:17 self was already taken by someone else
25:20 the bastard
25:22 um so it's YouTube come on Focus you
25:25 jerk ah
25:29 come on you can do it
25:37 anybody know Tim Cook
25:41 um
25:42 YouTube is at AI learning lab slash TT
25:46 as in Tick Tock huh clever
25:49 so I've got nine uh I I have the first
25:53 nine lives up there at some point I'll
25:56 get up all the tick tocks up there and
25:58 I'll get i'll get all the lives up there
26:00 and then hopefully I can automate this
26:01 [ __ ] because it's a lot
26:03 um but but we'll go I I think we're
26:05 gonna go play one of the things I'm
26:07 gonna do tonight is I'm gonna go see if
26:09 we can make a
26:10 a prompt that will give us a YouTube
26:13 description with chapters from these
26:16 tick tocks from The Tick Tock live
26:18 transcripts so that should be fun so we
26:21 can do that together we can do a little
26:22 we can do a little prop crafting
26:24 together and we can learn we can learn
26:26 things together
26:29 that's fantastic
26:32 yeah it seems he's he's like a teacher
26:35 but but you know a little he's he's
26:38 attached he's a little touched he's
26:42 touched
26:44 all right you're behind I am behind and
26:48 I will be behind I will get increasingly
26:50 behind as more people come in here and
26:52 ask me questions thank you Robert I
26:54 appreciate the donut it's obvious that I
26:57 appreciate Donuts
26:59 I'm trying man
27:01 I'm losing weight slowly
27:07 Donuts is tasty
27:11 um encourage more viewers to share yeah
27:14 viewers I'm encouraging you
27:17 um Can AI build similar HTML Pages based
27:20 on the design of another page yes it's
27:22 really good at that [ __ ] if you grab you
27:25 know your CSS style sheets and your HTML
27:27 of a thing and then you say you know
27:30 based on this give me a page that says
27:33 that and you can have it write the
27:35 content
27:36 generate the HTML update to CSS if you
27:39 need to it's quite good at that I've
27:41 even done it where where um
27:45 for like bulk emails um you know you can
27:47 do you can send plain emails which look
27:51 like [ __ ] or you can send HTML emails
27:53 well HTML emails are very very
27:54 particular because email programs still
27:58 can't handle much HTML so it understands
28:02 that so you can just say take this
28:03 shitty looking email and format it using
28:06 HTML that's appropriate for this
28:09 particular service and it it does it
28:13 good
28:14 and then and then rather than going into
28:16 the HTML and editing [ __ ] yourself you
28:18 can just go yeah make the Line Orange
28:20 now make it a darker orange or you know
28:23 here's a hex code or recommend a hex
28:25 code and go change it you know
28:27 throughout the throughout the HTML it'll
28:29 just go do it it's really good at that
28:32 um and of of all of these I would say
28:35 um chat GPT gpt4 is probably best but I
28:39 would experiment with with Bard I think
28:41 Bard is actually decent at coding which
28:43 makes sense considering it's from Google
28:46 was literally wait will literally invent
28:50 citations for sources that don't exist
28:51 it will so there's a pretty famous
28:54 lawyer
28:57 with a with a Scarlet eye on his
29:00 forehead for idiot
29:02 that had Chachi Piti write a legal
29:05 opinion for him and he handed it to a
29:07 [ __ ] judge
29:09 and the judge actually clicked on the
29:11 citations and said uh sir it appears
29:15 that your citations are non-existent
29:17 which you care to explain
29:19 [Laughter]
29:22 and I will argue
29:25 up one side and down the other that's
29:26 not Chachi PT's fault that's that
29:29 lawyer's fault if if you're doing things
29:31 like oh say
29:33 trying to prove a case do your own work
29:36 you know verify your own work
29:38 um hang on
29:41 but it is important to know all of these
29:43 things do what's called hallucinating
29:45 it's the the developers you know they
29:48 came up with that term because it's
29:50 better than you know it will [ __ ]
29:52 you I like to describe these things as
29:54 mansplaining as a service
29:56 um I didn't come up with that but I I
29:58 love it
29:59 um they will confidently answer you even
30:02 if they're wrong like every good man in
30:05 your life
30:08 um
30:10 the the way they work
30:14 what it looks like they're doing it
30:16 looks like they're a sentient being on
30:18 the other side of the computer just
30:20 typing all this remarkable crap they
30:22 don't actually understand what they
30:23 wrote what they're doing is using a
30:26 probability engine figuring out one word
30:29 at a time what's the next most probable
30:30 word so after you write your prompt it
30:33 sort of creates a mathematical waiting
30:35 it goes into this insanely complicated
30:39 Vector space and it picks out a vector
30:42 number that's associated with a word and
30:44 it plops it in there and then it does
30:45 that over and over and over again so but
30:47 it but what it doesn't do is it doesn't
30:49 go back and look at what it wrote and
30:51 understand if it's right or wrong so it
30:52 has no [ __ ] clue if what it gave you
30:55 is real or not and you can tell it oh
30:58 you got it wrong could you please get it
30:59 right now it'll be like Oh I'm terribly
31:01 sorry let me try to that again and it'll
31:04 get it wrong again sometimes so so you
31:06 just need to be conscious of that
31:08 um
31:09 doesn't mean don't use them
31:12 these are
31:13 the least capable they will ever be
31:16 right now the the speed with which these
31:18 things are improving is absolutely
31:20 insane so don't dismiss them just
31:22 because they're not perfect right now
31:24 I know nothing about AI or chat GPT
31:26 where do I begin and where do I go all
31:28 right Mama LA
31:30 let's get you going if you're just
31:32 getting started I would just go to the
31:33 top one don't worry about all these
31:35 other ones just go to chat.openai.com
31:37 and
31:39 go to
31:40 prompts.chat that very top URL
31:44 that's a document that will explain to
31:46 you what these large language models are
31:48 a little bit about how they work there's
31:50 little primers in there for free and
31:52 then there's a whole series of
31:56 prompts that that
31:58 put chat gbt into different modes like
32:02 you can have it act like a writing
32:04 partner or a gardening expert or a
32:07 historian or a therapist and you can
32:11 start interacting with it in that way
32:12 and those prompts believe it or not when
32:15 you tell it to act like a certain thing
32:17 it it actually gets much better at that
32:20 thing it's it's kind of
32:21 counter-intuitive from how we're used to
32:23 using Google which is give it as as
32:26 concise a prompt as possible and Google
32:29 will sort of figure out what you need
32:30 these things need a lot more context
32:34 and so that prompts.chat
32:37 um document will teach you about that
32:39 it's really good start there
32:42 here's how I would approach it start out
32:44 doing what I call parlor tricks just
32:47 play with it have it write you a country
32:49 song or have it write you a poem or have
32:52 it explain photosynthesis to you or give
32:56 you advice on your gardening or whatever
32:58 you might enjoy
33:00 just start playing with it and then
33:03 rather than try to figure out well what
33:05 do I use AI for flip it around and say
33:07 what's something I'm trying to do in my
33:10 life like what's a problem I'm trying to
33:12 solve either with my family it could be
33:15 like we want to eat better and I want to
33:17 do some meal planning
33:19 um or it might be with a community group
33:21 where you you want to plan a bake sale
33:23 or it might be something at work where
33:26 you know
33:28 I don't know you're struggling with
33:29 marketing and they're looking to you to
33:31 come up with a marketing plan take some
33:33 problem in your life that you need to
33:35 solve and see if you can solve it with
33:37 these things
33:38 that's the best way to start to get your
33:40 head around what they make possible
33:42 because anywhere where you know how you
33:44 would have done it without using these
33:47 things when you use these things you
33:49 will likely have what I call your Kevin
33:51 McAllister moment which is this
33:54 foreign
33:55 alone
33:58 um
33:59 and that's that's the moment you want to
34:00 get to because when you get to that then
34:03 you realize
34:05 these things are profoundly powerful and
34:08 they're amplifiers they're Humanity
34:10 amplifiers their ability amplifiers
34:13 you want to make a marketing plan okay I
34:16 gotta think about okay what are the
34:17 different kinds of marketing and who's
34:18 our target audience and this and that or
34:21 you just go to this thing and say Here's
34:23 my business
34:24 I need to create a marketing plan is
34:27 that something you can help with sure
34:30 here you go
34:33 and you'll be like really and you'll
34:36 read it and like 80 of it will be like
34:39 better than you thought and then 20 will
34:41 suck and then tell it hey this 20 sucks
34:44 oh sorry about that let's try that again
34:47 and we'll get better
34:49 and you say that's good but it seems a
34:51 little
34:51 expected do you have can you give me
34:54 some unexpected ideas sure I can do that
34:58 and within 10 or 15 minutes
35:01 you'll have a fully fleshed out
35:03 marketing plan
35:04 and your job you should have little
35:06 scabs on your chin from your jaw
35:08 dragging across the floor
35:12 so that's how I would start
35:14 with playing then solve a problem then
35:17 get hooked and then your life will never
35:20 be the same
35:24 you will step over the precipice
35:29 and enter a world largely unseen by
35:32 anyone before
35:35 you'll look back and it'll look quaint
35:38 and then you'll be hit with
35:41 raging fear oh my God
35:44 is my job going to be the same no it's
35:47 not
35:51 but
35:52 you'll be
35:54 way more prepared than people who aren't
35:57 exploring with this stuff
35:59 so yeah
36:01 give us your YouTube again okay
36:04 AI learning lab Dash TT
36:09 someone else some other dude got AI
36:11 learning lab bastard
36:16 I avoided it for three months so it's my
36:18 own stupid fault but you gotta add the
36:21 dash TT at the end he might be a good
36:23 dude I don't know but he got my AI
36:26 learning lab and it makes me cranky
36:32 uh only lonely people use Ai No that's
36:36 not true I think lonely people may start
36:38 having conversations with AI
36:41 and that's not necessarily A Bad Thing
36:45 I can tell you that
36:47 since using AI
36:49 I am more engaged I started a community
36:52 called the AI salon
36:54 so if you go to the salon.ai it's free
36:57 you seem like one of those that would be
36:59 a little bitter oh I was just trying to
37:00 get us to join us his thing for money
37:03 no it's free it's just a community we
37:07 meet every other week we have a online
37:09 Discord server where you can hang out
37:11 make mid-journey images ask questions
37:14 um
37:16 I've been more connected
37:18 more creatively
37:20 engaged
37:22 um
37:23 more productive than I've been probably
37:26 in my life
37:28 since using AI
37:32 so I do not resonate with only use
37:35 lonely people use AI at all
37:38 we're too see here's the deal
37:40 the the cliche if you haven't used AI
37:44 which I assume you haven't played with
37:45 it much
37:46 um based on the comment
37:48 the cliche is that the machines are
37:51 doing all the work right and that that
37:53 humans aren't necessary the reality is
37:55 is that humans
37:57 are collaborating with these machines to
38:00 do work and
38:02 to figure this [ __ ] out
38:06 there's no books on it there's no
38:08 manuals there's no university courses on
38:11 generative AI like the post chat GPT
38:13 [ __ ] like how do you use this stuff in
38:16 your marketing department
38:17 there's no [ __ ] manual for that right
38:19 now so you know what you have to do talk
38:21 to people
38:23 so what ends up happening is the more
38:24 you use this [ __ ] the more you're like
38:26 oh my God how do how do you use this
38:28 stuff oh here's what I do oh my God
38:30 that's amazing here's what I do oh my
38:32 God that's amazing
38:33 that's the stage of the business we're
38:35 in right now that's the stage of the of
38:38 the technology it's it's this is all
38:40 kind of spit and duct tape right now
38:41 it's built on Decades of research
38:46 um and and it's incredibly powerful but
38:48 the tools themselves like Chachi PT is
38:50 very very thin from a from a a product
38:54 standpoint
38:55 Pi is like barely functional
39:02 but they just raised 1.3 billion dollars
39:05 and and they're setting out to make a
39:08 personal assistant
39:09 and that personal assistant might do
39:11 things like help you socialize more help
39:14 you get more dates
39:15 help you save money on your next trip
39:19 teach you about personal finance right
39:22 so yeah
39:25 I do not resonate at all with that
39:28 statement
39:30 um I mean I mean quite honestly it's
39:32 it's like saying you know people that
39:35 use computers are anti-social yeah they
39:39 are but that's not the point that is not
39:41 the point little mister
39:45 see what I'm saying you see how I do
39:47 this you know what that means
39:50 neither do I
39:51 all right hello Rachel what's happening
39:54 amen what is a good AI State site to
39:57 animate a still picture
39:59 for example making rain animated let's
40:01 go now if I can get this stupid thing to
40:04 focus on the stupid screen
40:07 come on there we go
40:09 all right we're gonna go
40:12 we're gonna go first to Mid Journey
40:15 and we're gonna make what did you want
40:16 you wanted rain
40:17 that was a that was a bear making a beef
40:20 sandwich in a Chicago restaurant
40:24 we did that the other night actually
40:27 okay so if you've not made a mid-journey
40:30 image this is Discord here's here's your
40:33 Discord servers those are your different
40:35 communities I'm in the AI Salon Discord
40:37 server here's your different
40:39 conversations and so mid Journey play
40:42 and mid Journey lab where you can make
40:44 stuff
40:46 and so the way you make this is you go
40:50 slash imagine
40:55 oops not slash m
41:00 imagine
41:04 and then we're gonna go what did you say
41:07 you said
41:10 making rain animated so let's say
41:13 um
41:15 down let's see photograph
41:19 of
41:21 down for
41:24 in back yard with laundry hanging
41:33 um and then I'm gonna go dash dash AR
41:36 space 16
41:39 colon 9.
41:41 and if you're asking but Kyle why are
41:44 you having to write code to make a
41:46 picture because that's the way that's
41:48 where we are right now it's just the
41:50 dumb thing about it uh dash dash AR
41:53 stands for aspect ratio and 16x9 is like
41:56 a you know like a movie
42:00 um aspect ratio so
42:03 that's what we're gonna do
42:05 and then now it's gonna go it's gonna go
42:06 make us hopefully a decent downpour
42:11 and then we're gonna go to another tool
42:12 and we're going to turn this into a
42:14 video and hopefully you'll have a Kevin
42:16 McAllister moment like
42:19 I had no idea did that but Kyle that
42:24 seems like magic
42:27 or it'll suck and that's been it's been
42:30 kind of the theme this week is like I
42:32 went to I went to yesterday I went to
42:34 play on open ai's playground and it
42:37 literally the site just didn't work it
42:38 gave me back like two asterisks instead
42:41 of like writing what I wanted it to
42:43 write
42:44 and I looked like a dum-dum it made me
42:46 look like a dum dum dum dum dum dum dum
42:48 dum all right that's not really working
42:50 photograph of a downpour let's see
42:55 all right we're gonna go slash those are
42:57 cool but they're not what we asked for
42:58 slash imagine
43:02 um a photograph of
43:06 heavy
43:08 rain
43:12 um
43:15 drops
43:18 um
43:19 falling
43:21 past
43:24 the camera
43:27 in a
43:30 Farm
43:35 back yard
43:39 with laundry hanging all right let's try
43:42 that
43:44 we'll see if we can get something that
43:45 actually shows us the rain falling
43:54 and by the way uh if you're still here
43:57 it uh J trulius
44:01 um
44:02 the tool we're going to use is Runway
44:06 ml's gen 2. which is you can also do
44:10 text to video so you could just go
44:12 describe rain falling and it'll make you
44:14 video of rain falling
44:16 but what I like about
44:18 doing it in mid Journey first is mid
44:20 Journeys are dedicated it's it's the
44:22 best of the image generation tools it
44:24 just gives you the highest quality stuff
44:26 and you can start it there and then you
44:28 can use this image as your prompt for
44:31 Gen 2 to make a video
44:34 so
44:36 these do not look like rain but we'll
44:39 see they're not done yet
44:42 um does this have
44:49 this is this has got a little bit of
44:51 rain all right so let's do we'll do
44:53 variations of two so if you've never
44:54 used mid Journey you've got these these
44:57 little things at the bottom you one two
44:59 three four and V one two three four the
45:01 user upscale so you can take any of
45:03 these that you like and upscale them the
45:05 Visa variations so this one has a little
45:07 bit of rain in it so I'm going to do
45:08 variations on that one
45:11 and we'll see if it gives us one that's
45:12 got a little more rain
45:15 oh
45:33 we haven't flown on a Saturday night
45:35 everybody having a good time is it good
45:38 is it like a rock concert is this as
45:40 good as a Taylor Swift concert oh yeah
45:42 you know it is you know what it is
45:46 no I don't think so but
45:51 did you hear she just gave two hundred
45:52 thousand dollar bonuses to her 50 truck
45:54 drivers that drove her crap around for
45:56 that tour
45:57 she's amazing she are amazing amazing
46:02 amazing amazing
46:05 um I don't see any rain in these
46:07 although this one has water
46:09 water on the ground so let's grab this
46:11 one so we'll scale this one up that's
46:13 three we're gonna upscale three
46:19 and then I am gonna save as save image
46:26 I'll just call it oh that's good it's
46:28 got my prompt in there save that to
46:30 downloads okay so now we're gonna hop
46:32 over
46:33 to Gen 2 runwayml.com
46:39 [Laughter]
46:41 [Music]
46:47 here we come to save the day
46:50 all right so I went to Runway ml I have
46:53 an account there obviously
46:55 um I said text to Text slash image to
46:59 video so I can type in a prompt
47:02 and I'm gonna upload my
47:06 my uh
47:07 image
47:09 right here
47:21 all right so there's my little image and
47:24 then let's say
47:25 um
47:26 uh hard rain
47:30 I just typed in hard rain let's see I've
47:32 noticed that if it's sometimes better if
47:35 you don't
47:36 type in a prompt in fact let me kill the
47:39 prompt for now let's just do it with the
47:42 image and let's see if it if it figures
47:43 anything interesting out with the image
47:45 so I'm just going to use the image as
47:47 the as the prompt and so I say generate
47:49 and so now it's going to go off and it's
47:52 going to do its thing
47:54 look you can see me in the forest I'm
47:56 like a little
47:57 I'm like a little ghostly figure in the
48:00 forest
48:01 oh papa don't you see the face in the
48:04 woods
48:05 no Sally what are you talking about I
48:08 don't know I see him I'm terrified papa
48:11 papa don't you um what oh it's an ad
48:18 sorry
48:20 I was giving a performance there
48:23 I was just trying to win
48:31 the win at me it would complete me I
48:33 tell you
48:40 no it wouldn't
48:47 do you ever see that Jim Carrey thing
48:49 where he's he's like I'm not just Jim
48:52 Carrey I'm Golden Globe winner two-time
48:54 Golden Globe winner Jim Carrey
48:57 and I don't go to bed at night I go to
49:00 bed at night
49:01 as two-time Golden Globe winner Jim
49:04 Carrey
49:06 and you know
49:08 what I dream about
49:10 being three-time Golden Globe winner Jim
49:14 Carrey because then I'd be enough that's
49:17 pretty good all right let's see so we're
49:19 gonna hit play here you ready let's see
49:21 if we get some rain
49:23 no we don't get rain but we do get sort
49:25 of driving through that's actually
49:27 really good
49:29 that's beautiful
49:31 so there's Gen 2 right okay so that's
49:35 cool
49:36 so now
49:38 let me stop that because we'll get
49:39 seasick so now I'm going to put
49:42 um
49:44 heavy rain
49:46 falling
49:48 and we'll try another one generate
49:53 and let's see what it does the thing
49:55 that I've noticed about when you
49:56 actually add words to The Prompt is that
49:59 it doesn't
50:01 um use this exact image as the first
50:04 frame it kind of takes the basic idea of
50:08 this image it takes some of the themes
50:10 of it and then it adds its other [ __ ] so
50:12 it's going to be interesting to see what
50:13 it does here but I gotta tell you that
50:16 that animation that it just did of you
50:19 know sort of blasting down the the
50:22 driveway or whatever that is
50:24 that's pretty cool
50:27 like I could absolutely see that you
50:29 know so I mean this is the kind of thing
50:31 that if you get if you get a style
50:33 dialed in in mid-journey and and you
50:36 sort of figure out your storyboard of
50:38 the kinds of shots that you want and you
50:40 just established the first
50:42 um the first frame of each of those
50:44 shots
50:46 um you know this this is something you
50:48 could do now this is this is not super
50:50 cheap like I have 193 seconds left and I
50:54 think I'm paying
50:56 I forget what it is whatever the bottom
50:57 one is
51:00 um
51:01 there's another trick that you can do
51:03 that someone told me about that I
51:04 thought was pretty brilliant so if this
51:07 is your first frame
51:09 you can go to your last frame you can do
51:11 you can screenshot this or grab the last
51:13 frame of this video
51:15 and then use that as the first frame of
51:19 of the next video and then you could
51:20 string a bunch of those together and
51:22 make like a 30 second clip Gen 2 only
51:25 does four second clips
51:27 yeah so this is a completely different
51:29 scene but it does look like rain's
51:31 falling so here we go
51:35 yeah so you put an umbrella there it is
51:38 kind of in the woods
51:42 but it it completely recomposed
51:46 recompose the image but that's rain
51:48 falling
51:49 so
51:51 how can we improve Gen 2 fill out a
51:54 short survey
51:56 all right your email I'll I'll go I'll
51:59 do it how How likely are you to
52:01 recommend Runway to a friend or
52:02 colleague well I just did it to 84
52:06 people so pretty pretty likely How would
52:09 how can we improve runway for you
52:16 when adding a prompt
52:20 and an image at the same time comma
52:25 retain
52:30 as much of the image as possible period
52:35 I either get
52:38 image coherence with no prompt or
52:41 something completely different with a
52:42 prompt period
52:44 it would be nice to be able to
52:47 add something like
52:50 Heavy Rain comma and still have the
52:53 scene look the same with the addition of
52:55 rain period
52:59 what's the main benefit you get from
53:01 Runway
53:02 um experimenting
53:08 what type of people do you think will
53:10 benefit
53:12 um I don't know
53:15 creatives
53:18 I don't really care maybe contact you
53:20 yes okay all right so we gave them some
53:22 feedback
53:23 we've we're we're good citizens
53:26 but anyway so yeah so that's so Runway
53:29 would be the tool I would use to do that
53:30 you're gonna have to experiment with
53:32 with how you go from like what what
53:35 would probably make more sense for you
53:37 to do is if you can get mid-journey to
53:40 give you a better you know something
53:42 that's more clearly raining that gives
53:44 it the clue that it's raining
53:46 um or just use text to video at Runway
53:49 at Gen 2 and just describe the scene
53:53 that you want
53:54 um it to be raining in and it'll just
53:56 make that video for you
54:03 oh just jump me to the bottom
54:06 tick tock
54:08 just
54:10 I can't I can't take it anymore
54:15 I want to be where it was
54:17 why did you do this to me all I want to
54:20 do is help the people but I can't help
54:22 the people if I can't see their
54:24 questions Tick Tock please stop it
54:27 you're making me hurt inside
54:35 uh for those of you that are new here I
54:38 don't like it when Tick Tock jumps me to
54:40 the bottom
54:41 foreign
54:52 [Music]
55:15 all right
55:18 I use Chachi BT too much but I have
55:21 upgraded is there a cut off
55:24 I don't
55:26 I don't think so I think they're upping
55:28 I think they're upping their rate limits
55:30 so I've never hit
55:35 I tend to use chai gbt in little bursts
55:38 and I think they reset it every three
55:39 hours anyway so I've never really hit a
55:41 rate limit with chat gbt but if you're
55:44 doing some intense stuff and you get
55:45 rate limited the the other the other way
55:48 that you can do it Rachel there's
55:49 there's kind of two ways to use chat gbt
55:51 you can use it at chat.openai.com but
55:55 you can also use the openai playground
55:57 which is platform dot openai.com
56:01 playground it's a different interface
56:04 but you can still use chat
56:06 gpt4 as the model
56:09 and and use it that way so that if
56:12 you're getting rate rate Limited
56:14 that might be good also if you haven't
56:17 turned on plugins and code interpreter
56:21 and custom instructions do that on chat
56:24 GPT if you're paying the 20 bucks go get
56:27 the extra features because they're
56:28 really good
56:29 they're really really good except the
56:31 plugins are a little bit poo poo the
56:34 plugins plugins are not great
56:37 there's 800 of them but they're not
56:38 great but there's some good ones so
56:40 we're gonna I'm gonna play with some
56:41 tonight I'm gonna go do something with
56:42 my with my uh YouTube our new YouTube
56:46 channel
56:48 um
56:49 and we're gonna do some chat gptn with
56:51 the transcript
56:53 haha I fell asleep to you once let's see
56:56 it's
56:57 it's that's my favorite compliment slash
57:00 troll
57:03 I like to watch the AI learning lab to
57:05 get some well-earned shut eye
57:10 Kyle's voice just puts me into Slumber
57:13 like nothing else
57:14 [Laughter]
57:17 and I used to drink whiskey but now
57:20 it's the AI learning lab for me
57:25 oh my wife she's going to be in Denver
57:27 next month oh nice
57:29 has treats for chance that's awesome
57:31 Tobias that's great
57:34 champ will be happy he loves treats he
57:37 loves small stuffed animals too he'll
57:39 carry him around for three weeks and
57:41 that is one day that they're just
57:42 murdered there's just
57:44 small animal we call him we call him his
57:48 babies he's carrying his carrying around
57:50 his babies and then we're like champ why
57:52 did you kill your baby and there's just
57:54 like stuffed animal stuffing all over
57:56 the floor it's hilarious it's very cute
58:04 all right
58:05 can't find your YouTube can you add it
58:08 to the AI Salon screen yeah that's a
58:10 good idea
58:11 um the the YouTube is AI learning lab
58:15 Dash TT I just did it tonight there's
58:18 another AI learning lab that bastard got
58:20 there before me because I waited for
58:22 three months
58:24 before I did it
58:27 um
58:29 AI learning lab Dash TT as in tick tock
58:33 because
58:35 I didn't use chat GPT to have any
58:37 imagination for the title I had AI Dash
58:40 learning lab and then I was like no that
58:43 like No One's Gonna put the dash in
58:45 there so
58:46 AI learning lab Dash TT
58:49 but yeah I'll add it to my other I've
58:52 gotta I've got to go redo my link trees
58:55 anyway
58:56 ah there it is oh cool great I do know
58:58 Tim Cook
59:00 that's good could you tell Tim Cook to
59:02 to fix the fix the the auto focus on the
59:06 [ __ ] back cameras on these things
59:08 this could also be a uh
59:12 a tick tock issue but Jesus it's just a
59:15 it's just horrible
59:17 um
59:18 it just it just don't it just don't
59:20 Focus it's just like it's not looking do
59:23 you see there's text there maybe you
59:25 could use a little of your fancy AI [ __ ]
59:27 and have it Focus
59:29 your Apple figure it out
59:33 they don't have Steve Jobs there going
59:35 uh just make it work
59:40 anyway
59:42 um
59:42 I'm I'm less pissed off at Apple about
59:45 the focus than I am at Tick Tock for
59:47 jumping me down to the bottom of the
59:48 comments when I'm when I'm scrolling
59:50 when someone puts something new in all
59:52 right
59:53 um are you gonna get your retina scan
59:55 when you go to New York I am yep
59:57 and and listen I they claim that they're
1:00:00 not keeping the the biometric data I
1:00:02 don't I don't really care if they do or
1:00:03 not like I feel like my data's already
1:00:05 out there like I I I made
1:00:09 I consciously made a deal with the devil
1:00:11 I don't know 30 years ago when I was
1:00:14 doing agency.com back in the olden timey
1:00:16 days where it's just like I'm just gonna
1:00:18 put my data out there I don't I don't
1:00:20 give a [ __ ]
1:00:21 um
1:00:24 I so so what what uh what's your name
1:00:28 cuppy cuppy206 what cuppy's talking
1:00:31 about
1:00:32 is Sam Altman last week launched this
1:00:35 thing called World coin and it's got
1:00:37 three components to it it's got a
1:00:39 transaction service thank you very much
1:00:41 for the rose appreciate that uh there's
1:00:44 a transaction kind of component like
1:00:46 PayPal there's a cryptocurrency called
1:00:48 World coin and then there's an
1:00:50 identification protocol called World ID
1:00:53 and in New York Miami and San Francisco
1:00:57 you can go to these physical locations
1:00:59 and they have this round orb they call
1:01:01 it the orb and
1:01:04 if they wanted to make it if they wanted
1:01:06 to make a scarier sci-fi name for the
1:01:09 thing they couldn't they couldn't have
1:01:11 so it's a it's it's a stupid it's a
1:01:13 stupid name that is that is gonna
1:01:15 perpetuate the paranoia that the robots
1:01:18 are gonna kill us but whatever but you
1:01:21 go to one of these major cities and you
1:01:22 look into this orbit scans your iris it
1:01:24 all it's basically doing is going yep
1:01:26 that's that's a human Iris or it's an
1:01:28 iris whatever we're taking the biometric
1:01:30 fingerprint we're verifying you as human
1:01:32 and then you get a blockchain verified
1:01:34 ID that says you're human
1:01:37 what's that why is that important well
1:01:41 I I think we are about to learn
1:01:44 the importance of human verification in
1:01:47 the 2024 U.S presidential election uh
1:01:51 because there is going to be so much
1:01:53 misinformation and and probably
1:01:56 personalized misinformation
1:01:59 um
1:02:00 and
1:02:02 you know things just things that look
1:02:04 real that aren't and things that you
1:02:07 know
1:02:07 came from a trusted source that didn't
1:02:10 actually come from The Trusted source
1:02:12 so there's a verification component to
1:02:14 it that's really important I I
1:02:15 personally just want to be part of it
1:02:17 early enough just because I think the
1:02:19 applications for it are kind of
1:02:21 insanely powerful imagine
1:02:23 something like a Twitter or X as as it
1:02:28 likes to be called now
1:02:31 where you knew that everyone that you
1:02:33 were interacting with on that service
1:02:35 was a human was a verified human and had
1:02:38 a very a blockchain verifiable history
1:02:40 that you could go follow all right they
1:02:43 signed up for Twitter on this date and
1:02:44 this on this date
1:02:46 and they're verified human with this
1:02:48 world ID cool
1:02:50 and then you can layer on top of that
1:02:52 you know here's a video that Kyle made
1:02:54 and I can verify you know on the
1:02:56 blockchain that that hasn't been
1:02:58 modified okay so I verified that he's
1:02:59 human I've verified that this is video
1:03:02 that he created and that it's not been
1:03:04 modified okay I can trust that video
1:03:06 become really really important so
1:03:08 there's a layer this verification layer
1:03:10 of of content and people is sorely
1:03:15 missing right now so I think it's
1:03:17 interesting on that front
1:03:19 the other thing about world coin that's
1:03:21 that's
1:03:22 I I
1:03:24 don't know like
1:03:27 the the sort of
1:03:29 depth that they're thinking they've been
1:03:32 thinking about this stuff for a lot of
1:03:33 years and so and it shows right because
1:03:35 they've thought about the implications
1:03:37 so one of the implications of these
1:03:40 tools is that they are going to cost
1:03:41 people jobs there is going to be
1:03:43 disruption people are going to be
1:03:45 displaced
1:03:46 jobs are going to be reinvented or
1:03:48 completely automated out with with this
1:03:51 technology and and not just this the
1:03:53 image stuff the video stuff all of it
1:03:55 right the tools that you look at now
1:03:59 um are the worst they're ever going to
1:04:01 be these are going to get incredibly
1:04:03 sophisticated incredibly quickly so jobs
1:04:06 are going to be eliminated
1:04:08 so there's a couple of things going on
1:04:10 Sam Altman when he when he flipped open
1:04:13 AI from a non-profit company to a
1:04:15 for-profit company he's doing it as a
1:04:18 capped for-profit
1:04:20 so what that means is if if an investor
1:04:23 put in a billion dollars they can
1:04:26 um the maximum they can get returned
1:04:29 from that investment is a hundred X so
1:04:32 if they put in a billion they can get a
1:04:34 hundred billion out
1:04:35 Sam Altman feels that this is going to
1:04:38 be trillions McKinsey agrees this is
1:04:40 going to be trillions of dollars of
1:04:42 prosperity generation
1:04:44 so anything above that cap above above
1:04:47 that 100x cap is going to basically go
1:04:49 into the fund so the other thing about
1:04:51 world coin is that he's setting it up as
1:04:54 a universal basic income
1:04:56 for members and
1:05:00 um
1:05:03 and
1:05:04 he's basically going to turn it into a
1:05:08 wealth redistribution platform so that
1:05:11 as these large language models start to
1:05:14 displace jobs and start to increase
1:05:16 Prosperity that there's basically a tax
1:05:19 on that that's going back to the people
1:05:21 that were displaced so I think it's a
1:05:23 fascinating experiment I want to get in
1:05:26 on it
1:05:27 um thanks to the stupid [ __ ] sec
1:05:31 everyone else in the world when they
1:05:33 scan their eye they get free
1:05:35 cryptocurrency but in the United States
1:05:37 we can't get free cryptocurrency because
1:05:39 it's a Ponzi scheme and it's evil and
1:05:42 we're not allowed
1:05:45 but anyway yeah so yes I am
1:05:47 already subscribed cool he is touched
1:05:50 thank you I appreciate that do you think
1:05:53 AI comedians will be funnier than real
1:05:55 comedians not with their current timing
1:05:58 go go listen to pie with when you talk
1:06:01 to pie you say something funny she goes
1:06:06 that was a funny one Kyle
1:06:09 no I don't think they're gonna be funny
1:06:11 other comedians for some time
1:06:15 um we did an experiment a couple of
1:06:17 weeks ago on the live where we went to
1:06:19 all of these different tools I wrote a
1:06:21 prompt um to write a joke
1:06:23 about avocados
1:06:25 um you know not being ripe for a long
1:06:28 time and then they're ripe and then
1:06:29 they're rotten
1:06:31 um and chat gpt4 was the only one that
1:06:34 actually wrote a joke that could have
1:06:36 could be turned into something a
1:06:38 comedian might actually say all of the
1:06:40 others were embarrassingly bad
1:06:43 um but they're gonna get better they're
1:06:45 gonna you know someone's gonna write a
1:06:47 joke writing
1:06:48 um actually you could if you wanted to
1:06:50 play with it go to Poe and create a joke
1:06:53 writing bot
1:06:54 I bet I bet you could do one pretty
1:06:56 decent
1:07:00 all right
1:07:06 is there anything you would not use AI
1:07:10 what are you doing there Cowboy liaison
1:07:13 oh thank you the bear welcome the Bears
1:07:16 good to see you thank you for the Hattie
1:07:18 hat
1:07:19 thank you for the rose Adam that's nice
1:07:21 uh
1:07:24 we did a we did a little run the other
1:07:26 night where a bunch of people were
1:07:28 putting stupid [ __ ] on screen and I got
1:07:30 all NPC like a rose a hat I mean
1:07:35 oh those sights are funny did you have
1:07:37 you seen the there's a dude that um so
1:07:40 there's the the people doing the NPC
1:07:42 thing where they like you know
1:07:43 clickety-clack and whatever as people
1:07:45 pay them [ __ ] they they talk stupid and
1:07:47 then
1:07:48 um this guy
1:07:50 put an iPhone on his robotic lawnmower
1:07:53 and he's just putting it on Tick Tock
1:07:56 lives and and for like four hours at a
1:07:59 shot he puts it on Tick Tock lives as
1:08:02 this lawnmower just drives around his
1:08:04 yard mowing very very slowly and he's
1:08:06 made like 2 500 bucks in two weeks so
1:08:09 you can make money with Chachi
1:08:12 you can make money with tick tock oh
1:08:15 it's [ __ ] hilarious
1:08:19 ah what is there anything I would not
1:08:22 use AI for
1:08:24 well hmm
1:08:36 I don't think that there's anything that
1:08:38 I would use AI exclusively for right now
1:08:40 although I I you know that's probably
1:08:42 not true I probably have done that
1:08:46 like my so so I've got it you know my my
1:08:50 art project on Instagram is called Kyle
1:08:52 Shannon dreams and it's I've also got a
1:08:54 tick tock Channel called Kyle Shannon
1:08:56 dreams where I took these you know AI
1:08:58 generated self portraits and and put
1:09:00 them up there
1:09:01 um you know that that was 100 AI but
1:09:05 most of most of the work that I do when
1:09:06 I'm doing when i'm doing things to do
1:09:09 things it tends to be a collaboration
1:09:12 between me and the tools
1:09:15 I think that right now
1:09:20 I think it's I think it's in all of our
1:09:23 interest to
1:09:26 and in fact Ethan Malik said this
1:09:28 he was at the uh the marketing AI
1:09:31 Institute
1:09:32 just had their annual conference and
1:09:35 Ethan Malik spoke and he said he said
1:09:37 the best thing you can do right now
1:09:39 is for everything that you do at work or
1:09:43 at school or whatever it is
1:09:46 try to use one of these tools
1:09:49 when you do that thing
1:09:51 try to figure out if if it can do it if
1:09:54 it can do it better if it can accelerate
1:09:56 something
1:09:57 because
1:09:58 the way these things do what they do
1:10:01 we've never had anything remotely like
1:10:04 this so none of us know how it's going
1:10:08 to change how we write emails how we
1:10:11 read emails how we read annual reports
1:10:14 you know how we create PowerPoint
1:10:16 presentations how we modify them
1:10:19 um how we create computer stuff like
1:10:21 everything's going to change
1:10:23 so I kind of feel right now like
1:10:28 everything needs to be an experiment
1:10:30 look everything is an experiment because
1:10:32 these tools are so new and they're so
1:10:35 imperfect right so you can't
1:10:37 you just can't trust these things are
1:10:40 gonna do the [ __ ] you think they're
1:10:42 gonna do exactly how you think they're
1:10:44 gonna do it like like the thing that I
1:10:46 just experimented with where I created
1:10:48 the image of the backyard and it was
1:10:50 sort of raining but not and we put it in
1:10:52 the video thing and it made a cool video
1:10:54 but it wasn't raining and then when I
1:10:56 said Make It Rain it added an umbrella
1:10:58 which we didn't want a [ __ ] umbrella
1:11:00 it's a backyard we don't need an
1:11:01 umbrella right
1:11:03 um and it didn't get that so but it was
1:11:06 a good experiment right like like I feel
1:11:09 like I learned something in just kind of
1:11:11 getting the Nuance of how it did what it
1:11:14 did
1:11:15 but it's not going to act like that a
1:11:17 year from now right a year from now it's
1:11:20 probably going to have a level of
1:11:21 coherence that's just mind-blowing to
1:11:24 what we're doing today
1:11:25 so I might not use that for anything
1:11:28 today but six months from now
1:11:31 I might use it for 50 of the things like
1:11:33 that that I need and a year from now
1:11:35 that might be the primary way I do it
1:11:37 and and two years from now that might be
1:11:40 the primary way that all videographers
1:11:42 do what they do
1:11:44 we don't know
1:11:46 so
1:11:48 so I don't think so I think right now
1:11:53 I I think right now like like where my
1:11:56 head is is
1:11:58 every time I have a problem
1:12:01 and not a problem every time I have
1:12:03 something that I need to solve at work
1:12:07 I'm now I'm now at the place and this
1:12:09 took me five or six months I'm now at
1:12:12 the place where I first think is there
1:12:15 anything I could do with AI with that
1:12:19 and sometimes the answer is no but a lot
1:12:21 of times the answer is like yeah let me
1:12:22 go let me hop into chat gbt here let me
1:12:24 run over here let me pre-visualize
1:12:26 something over there thanks Adam
1:12:27 appreciate the rose
1:12:29 um
1:12:32 that's a really weird switch it's a
1:12:34 really weird ah thank you very much it's
1:12:36 a really weird switch to go from
1:12:39 if if you're in well we're all problem
1:12:42 solvers but but like I pride myself in
1:12:45 general sense
1:12:46 oh shut up Siri seriously
1:12:52 come on
1:12:56 back to Tim Cook
1:12:57 [Laughter]
1:13:00 um what the hell was I talking about uh
1:13:03 uh
1:13:06 I don't know what I was talking about
1:13:07 Syrian or after me she broke my brain oh
1:13:13 anyway all right
1:13:17 I've been wanting to look into it for a
1:13:19 while but didn't know where to start oh
1:13:20 absolutely mama
1:13:23 um yeah start with start with
1:13:25 chat.openingi.com have it make you a
1:13:27 meal plan like the meal plan thing is
1:13:29 really cool or an exercise plan or like
1:13:31 you know some sort of thing like that
1:13:33 just give it sort of a general you know
1:13:35 kind of
1:13:37 you know diet that you want to play with
1:13:39 and then it'll make you a big meal plan
1:13:41 oh you know what we'll do something cool
1:13:43 I'll show you how to I'll show you how
1:13:44 to use plugins let's let's go do a
1:13:46 little bit of uh
1:13:48 let's go do a little bit of chat GPT in
1:13:51 together
1:13:57 all right I'm going to show you a couple
1:13:59 of things so here's chat gbt they've
1:14:02 added these fancy they're they're kind
1:14:05 of useless they're sort of thought
1:14:07 starters but you know whatever it's cute
1:14:09 it's nice they're they're they're
1:14:10 they're they're doing they're doing what
1:14:12 they can to improve it they need to do
1:14:14 things like return access to the
1:14:15 internet to to improve it but you know
1:14:18 some thought starters are nice okay so
1:14:21 if you're if you're at chat GPT
1:14:25 um if you pay your 20 bucks a month you
1:14:27 get this little yellow flag and then you
1:14:29 get access to code interpreter and
1:14:31 plugins as well as custom instructions
1:14:34 and I'll talk about those in a sec
1:14:37 um if you don't see these if you're
1:14:39 paying for the if you're paying your 20
1:14:40 bucks a month but you don't see code
1:14:42 interpreter and plugins you actually
1:14:44 need to explicitly enable them so you go
1:14:47 to your settings and then in settings go
1:14:50 to Beta features and turn that [ __ ] on
1:14:53 see the little green pills you want
1:14:55 green pills not gray pills
1:14:58 gray pills make you sleepy green pills
1:15:00 that make you all excited you'll be like
1:15:01 I could play with code interpreter now
1:15:03 oh yeah
1:15:04 all right so we're gonna play with
1:15:06 plugins first
1:15:09 so I was talking to Mama LA about
1:15:13 um
1:15:15 about making a meal plan
1:15:17 so I'm gonna turn on
1:15:20 a plug-in
1:15:22 where are it
1:15:24 um
1:15:27 did they get rid of it or do they just
1:15:30 miss it
1:15:31 there it is instacart okay
1:15:35 so I'm gonna say let's see
1:15:41 make me a meal plan for the next seven
1:15:43 days for a family of four
1:15:46 with a chubby dog
1:15:48 with a heavy emphasis on Meatloaf
1:15:54 all right
1:15:55 [Laughter]
1:15:58 why why not sure I can help with that
1:16:00 here's a simple meal plan
1:16:02 all right so let's see what it gives us
1:16:05 breakfast scrambled eggs with toast
1:16:08 chicken salad sandwiches for lunch
1:16:09 meatloaf with mashed potatoes for dinner
1:16:12 let's see
1:16:14 day two lunch leftover meatloaf
1:16:17 sandwiches that's actually a brilliant
1:16:19 day too
1:16:22 um spaghetti bolognese is kind of
1:16:24 shredded meatloaf on pasta which is good
1:16:28 meatloaf with roasted vegetables on day
1:16:30 four
1:16:31 baked salmon
1:16:33 baked salmon no all right fine barbecue
1:16:38 ribs
1:16:39 steak and potatoes
1:16:41 as for your chubby dog
1:16:44 it's important to consult with a vet
1:16:50 all right so there's all of our
1:16:52 ingredients there's our instructions
1:16:53 okay now what it what it says after
1:16:56 giving me that shopping list is would
1:16:58 you like to create a shopping list for
1:17:00 these meals and I can say yes
1:17:02 and what this is going to do now it's
1:17:04 actually going to fire off the instacart
1:17:06 plug-in
1:17:08 it's going to understand what are the
1:17:11 ingredients I need
1:17:12 to execute that meal plan and turn that
1:17:15 into an instacart shopping list and then
1:17:18 I can just have that [ __ ] delivered
1:17:21 so plug-ins you kind of you activate
1:17:24 them for for a given chat session you
1:17:27 can activate up to three I just have one
1:17:29 in this case
1:17:31 um for demo
1:17:32 you can activate up to three you can
1:17:34 install as many as you want but in any
1:17:36 given chat session
1:17:38 um you just do you just do one
1:17:41 so
1:17:45 today nine stores are delivering
1:17:48 that's not actually my ZIP code but
1:17:49 that's fine
1:17:51 but like there's our there's the stuff
1:17:53 right ketchup we're having meatloaf
1:17:55 we're getting we're getting the three
1:17:56 pack of ketchup damn right we are
1:17:59 we're getting a roasted chicken we got
1:18:02 some meatballs in there some berries
1:18:04 some tomato sauce
1:18:06 this looks expensive we're going back to
1:18:08 macaroni and cheese
1:18:11 um
1:18:11 so
1:18:13 so that's kind of how plugins work
1:18:15 um
1:18:17 so now let's let's try something new
1:18:21 I'm gonna go to plugins I'm gonna turn
1:18:23 off instacart and I'm gonna turn on
1:18:29 video insights I think there might be
1:18:31 another one Vox script
1:18:34 so both of these will read YouTube
1:18:37 transcripts so what we're going to do is
1:18:39 we're going to hop over to the newly
1:18:41 minted look at this wait let me go let
1:18:45 me go back to the home page
1:18:48 AI learning lab YouTube channel huh
1:18:51 I did it you did it yay
1:18:56 um so I'm gonna go to videos and I'm
1:18:58 gonna go grab let's see what's the first
1:19:01 one
1:19:03 April 30th April 29th was when it all
1:19:07 started
1:19:08 all right
1:19:11 I'm gonna go to April 29th everybody
1:19:13 look there's two of me I'm the same size
1:19:16 too
1:19:17 look look how look look how much fatter
1:19:19 I was good Lord
1:19:23 it's disturbing okay
1:19:26 shut up
1:19:28 okay so now we're gonna come back over
1:19:30 to chat gbt
1:19:32 and what we're gonna do is we're gonna
1:19:34 say
1:19:36 um
1:19:37 first thing I'm going to do I'm going to
1:19:39 say
1:19:43 do you understand how to create video
1:19:47 titles descriptions
1:19:50 and
1:19:53 chapters
1:19:55 for YouTube
1:20:03 yes I can certainly help you understand
1:20:04 video titles should be engaging
1:20:09 video descriptions
1:20:12 there's a reason I'm doing this so so
1:20:14 one of the things that really helps a
1:20:16 lot with Chach EBT is if you give it
1:20:18 context and one of the ways you can give
1:20:19 it context is you can have it educate
1:20:21 you
1:20:22 right and so they're there it's telling
1:20:24 me
1:20:25 um here's how to make chapters you know
1:20:28 with within YouTube
1:20:31 um here's all this stuff so now when I
1:20:34 start asking it to do this for me it
1:20:36 will follow its own rules so so it will
1:20:38 make it better
1:20:40 um just in doing that okay so what we're
1:20:42 going to do is we're going to say
1:20:44 um
1:20:47 this video and then I'm going to paste
1:20:51 in the URL
1:20:53 and I'm going to hit shift return
1:20:55 let's see this video
1:20:57 um
1:21:02 is a recording of a fairly long
1:21:08 live session
1:21:11 where the person
1:21:13 talks about AI period
1:21:17 can you
1:21:22 come up with a title that encapsulates
1:21:26 what was talked about comma a
1:21:29 description
1:21:30 that highlights some of the more
1:21:33 interesting things and creates a chapter
1:21:35 list
1:21:37 of the relevant sections that might be
1:21:40 interested to an audience trying to
1:21:43 learn about generative AI period
1:21:47 so
1:21:48 now in theory
1:21:51 in theory this is going to run off okay
1:21:53 so it's grabbing video insights
1:21:56 it grabbed that plug-in So based on the
1:21:58 transcript
1:22:04 what is generative AI
1:22:18 ethical implications of AI let me I
1:22:21 don't think this is the right
1:22:24 um I'm gonna put
1:22:27 I don't think that's the right format
1:22:29 for the chapters for YouTube can you
1:22:32 please create the description
1:22:37 and the chapters so that they are
1:22:40 appropriately formatted
1:22:42 to put in a YouTube description and have
1:22:45 them function
1:22:46 as hyperlinks
1:22:50 let's see if it fixes that ice I
1:22:53 apologize
1:23:09 I don't think
1:23:14 I don't think it did it based on the
1:23:16 transcription because it's got
1:23:18 introduction what is generative AI
1:23:21 applications of generative AI Deep dive
1:23:23 into generative models ethical
1:23:25 implications Future q a session
1:23:27 conclusion that's not how my lives work
1:23:30 this is [ __ ] this is [ __ ] so
1:23:32 this is this is what we call in the
1:23:34 business an experiment we experimenting
1:23:37 and the experiment was a miserable
1:23:39 failure it did not look at the
1:23:41 transcription at all
1:23:44 um
1:23:46 it appears you did not look at the
1:23:48 transcription of that video at all based
1:23:51 on the chapters you created period
1:23:54 can you see the transcription of the
1:23:56 video I gave you earlier
1:24:02 ah if now it's using a different plugin
1:24:07 that's interesting
1:24:09 so this is box script and so what we
1:24:11 might find here is that Vox script did
1:24:13 better
1:24:15 and it's actually looking promising the
1:24:17 fact that it's doing multiple calls we
1:24:18 can open these up and see what it is
1:24:20 because it was a long transcript I bet
1:24:22 it's having to do the summarization
1:24:25 um separately or yeah let's see
1:24:38 next transcription next transcript token
1:24:41 yeah so Vox script is automatically
1:24:43 breaking up a longer the longer
1:24:46 transcript into into
1:24:48 like 3 500 token chunks
1:24:52 which is this this is actually
1:24:53 impressive to me
1:24:55 um so so if you don't know Chachi PT
1:24:58 kind of has this limit of 4 000 tokens
1:25:01 it's moving to eight it's not I I don't
1:25:03 quite quite know what it is on gbt4s but
1:25:07 it's basically you can't put in a really
1:25:08 long transcript or it just chokes it out
1:25:10 so what Vox script just did was it
1:25:13 automatically broke that up for me
1:25:16 all right this is looking better
1:25:19 introduction AI in the music industry
1:25:22 intellectual property challenges Ai and
1:25:24 e-commerce large language model
1:25:25 optimization Ai and project management
1:25:28 Ai and content
1:25:30 Ai and perfume Ai and energy
1:25:34 please note that the time stamps are
1:25:36 placeholders
1:25:39 um let's
1:25:42 in this comprehensive discussion we
1:25:44 explore
1:25:46 it's a fascinating world of generative
1:25:48 AI
1:25:50 um let's see
1:25:55 can you make the description
1:25:59 less flowery
1:26:02 and more factual
1:26:12 absolutely here's a straightforward
1:26:14 version this video provides an in-depth
1:26:16 discussion on generative AI it's
1:26:18 applications and the potential implant
1:26:20 impact on various Industries the speaker
1:26:22 discusses how AI is being
1:26:27 uh I don't know if that's true but
1:26:29 anyway let's
1:26:31 I'll deal with the with the description
1:26:34 some other time but let's
1:26:37 go to YouTube
1:26:40 and then how do I edit this
1:26:43 I go to
1:26:45 my YouTube studio and this is 429
1:26:53 so we're gonna go content and then we're
1:26:55 gonna find 429
1:26:58 where's 429 there it is and we're gonna
1:27:01 edit it how do I edit it what
1:27:06 where's the
1:27:09 um
1:27:13 there we go
1:27:15 I'm gonna pop in the description I'm
1:27:17 gonna put in all those chapters
1:27:20 and I'm going to save it
1:27:23 and then we're gonna go
1:27:28 back out to the channel
1:27:33 and now I should
1:27:45 where's the description why is it not
1:27:47 there
1:27:48 foreign
1:27:51 no views nobody's views my videos but
1:27:55 they've been up for an hour
1:27:57 a show transcript where's where's the
1:27:59 description
1:28:03 hmm
1:28:04 maybe
1:28:06 and what why does it not work instantly
1:28:14 and I'll address anything
1:28:17 um one of the things that
1:28:20 I've talked about on the channel I've
1:28:22 actually got this thing what I'm trying
1:28:24 to do I'm trying to see if the if the
1:28:26 chapters are
1:28:28 um functional
1:28:31 like if if they're actually what they
1:28:34 say they are
1:28:36 I have a sneaking suspicion they're not
1:28:38 but
1:28:40 edit oh it didn't save it
1:28:44 save
1:28:59 foreign
1:29:07 there we go okay
1:29:10 so now
1:29:12 show more okay so now we've got
1:29:15 our little chapters although it didn't
1:29:17 hyperlink the bottom ones that's
1:29:19 probably some weird copy paste problem
1:29:22 but large language model optimization at
1:29:25 35 minutes in
1:29:28 I only need 30 seconds of anyone's voice
1:29:31 probably less than that to really
1:29:32 synthesize it
1:29:34 so
1:29:35 so anyone who's out there
1:29:37 um you know whether it's actors or voice
1:29:40 over actors this isn't project
1:29:42 management
1:29:44 um relatively replicable things like
1:29:47 blog posts or not replicable things that
1:29:50 are more formulaic
1:29:52 where it's the same kind of thing over
1:29:54 and over again I would say here's air in
1:29:56 the music industry what its implications
1:29:58 are how to incorporate it into your
1:30:00 business or into your life or if those
1:30:03 two are very blurry
1:30:05 um how to deal with that
1:30:07 Synthesia uh
1:30:11 s-y-n-t-h-e-s-i-a is one of them and
1:30:14 dash id.com in fact I can go there hang
1:30:18 on good this is AI and economy great
1:30:21 question I wouldn't have had an answer
1:30:23 to this but I I talked to a friend of
1:30:25 mine who said he'd been thinking about
1:30:26 AI in a 30 to 50 year time frame it's
1:30:30 wrong
1:30:31 all right that was a failed experiment I
1:30:33 I I'll experiment with that but but
1:30:35 hopefully you can learn you know at
1:30:37 least what I did there right you know I
1:30:39 had a I had a plug-in that goes and
1:30:41 looks at a at a transcript the the fact
1:30:44 that it broke that transcript up into
1:30:46 all these components is actually
1:30:48 promising there's just something about
1:30:50 how I wrote the the prompt
1:30:53 um for it to create these chapters it
1:30:54 didn't quite it didn't quite do it right
1:30:56 so
1:30:57 um oh you know what it didn't do
1:31:01 this is interesting
1:31:03 if I go look at what data it gave back
1:31:07 it gave back parts of the transcript but
1:31:10 did it it did not include
1:31:14 it did not include time code
1:31:18 so let's let's do an experiment here
1:31:21 um
1:31:24 it looks like when you broke the
1:31:27 transcription up into components you did
1:31:31 not include the time code
1:31:34 in the summarized chunks
1:31:38 and therefore the chapters that you
1:31:40 created
1:31:42 are not accurate to what the person is
1:31:44 actually saying period
1:31:46 can you re-analyze the transcript
1:31:50 and generate
1:31:55 up to three chapters per chunk
1:31:59 with accurate time code and descriptions
1:32:05 yeah there's more than one way to skin a
1:32:07 cat and actually what I'm what I'm doing
1:32:09 right now like like this kind of
1:32:14 okay
1:32:16 using large language models right now is
1:32:19 kind of like being in an escape room
1:32:24 and you know it'll sort of do what you
1:32:27 ask it to do but then it doesn't quite
1:32:30 and like the lock doesn't unlock you
1:32:31 like damn it we were so close
1:32:33 and then and then you're like I know we
1:32:37 could ask it to do it this way and it'll
1:32:39 do it that way and so yeah now it it
1:32:43 what did it say I apologize for the
1:32:45 confusion earlier let's revisit the
1:32:47 transcript one more time with accurate
1:32:49 um okay that's that's [ __ ] so I
1:32:53 think I've still got the URL there so
1:32:55 let's try again um
1:32:59 you still didn't get it please
1:33:02 start over
1:33:05 with the transcript of this video
1:33:12 uh wait
1:33:15 uh this video
1:33:25 um with a transcript of this video
1:33:29 um and
1:33:32 when you and and let's see
1:33:38 let's just focus on creating accurate
1:33:41 chapter
1:33:44 names and time code so make sure you
1:33:48 include time code
1:33:50 when you analyze the transcript
1:33:54 all right we'll give it one more shot
1:33:56 we'll see if it fire it should fire off
1:33:58 Vox script
1:33:59 nope it used video insights I'm going to
1:34:02 stop generating
1:34:04 um
1:34:08 please use the Vox script
1:34:15 plugin plugin
1:34:21 did
1:34:26 and we'll see if it actually puts any
1:34:29 um time code in the summaries
1:34:33 it did not
1:34:36 wait did it
1:34:49 maybe it did
1:34:57 no no no
1:34:59 this is a fail it's a fail fail whale
1:35:04 we come we saw we conquered then we
1:35:08 realized we did not conquered and it was
1:35:10 an abject failure we are leaving in
1:35:14 humiliation nay
1:35:17 laughs
1:35:23 [Music]
1:35:25 so so I uh
1:35:31 that's
1:35:33 that's the way a lot of this goes and
1:35:38 you know if you're sitting there going
1:35:39 that seems miserable well that's where
1:35:42 it is right now
1:35:44 um two years from now
1:35:46 this stuff's gonna be magical right
1:35:48 because all all of
1:35:51 this
1:35:53 mess
1:35:55 is gonna get solved and these tools are
1:35:57 just going to get increasingly
1:35:58 sophisticated like you can see that what
1:36:00 it's what it's trying to do is in the
1:36:02 right neighborhood right the fact that
1:36:04 it's do it's breaking up a transcript
1:36:06 into sub into
1:36:08 um chunks small enough that chat CPT can
1:36:11 do is good the fact that Vox script as a
1:36:15 plug-in isn't smart enough to include
1:36:17 the time code is
1:36:18 [ __ ] stupid
1:36:20 right okay
1:36:22 they'll fix that at some point or
1:36:24 someone will come along with something
1:36:25 better or YouTube will come along with
1:36:27 something better and you'll just be able
1:36:28 to pop the pop the write my description
1:36:32 and chapters for all my videos button on
1:36:36 YouTube right it's Google so they could
1:36:40 do that probably tomorrow
1:36:44 um but
1:36:46 the process that I just went through of
1:36:48 thinking like oh I don't want to have to
1:36:51 manually watch all of my [ __ ] lives
1:36:54 again and figure out what I was talking
1:36:55 about but they're essentially useless
1:36:58 um if you don't kind of chapterize them
1:37:00 and understand what's in them
1:37:02 so I've got to find a way to solve this
1:37:04 problem
1:37:06 um it's definitely going to be some
1:37:07 version of chat CBT or I might go to
1:37:10 Claude so so here's an actually you know
1:37:12 what let's go ah
1:37:15 oh oh it's so exciting look stimming
1:37:19 stimming stimming
1:37:21 I just had an idea we're gonna go to
1:37:24 Claude Claude is a beast okay so here's
1:37:28 what we're gonna do
1:37:30 we're gonna go to
1:37:33 um so I'm on YouTube I clicked on the
1:37:35 little three dots right here I'm gonna
1:37:36 say show transcript
1:37:38 so here's my transcript with time code
1:37:40 can I copy this with time code what's it
1:37:43 what's it let me do here
1:37:46 toggle timestamps
1:37:49 select all no
1:37:52 okay I'm gonna just go select all of
1:37:55 this
1:37:57 okay now so so this is a good example of
1:38:01 when you might use different tools okay
1:38:05 so we just tried
1:38:08 oh we're gonna do we're gonna do two
1:38:10 tools this is this is like a Hands-On
1:38:12 little workshop here tonight kids it's
1:38:15 gonna be fantastic what'd you do
1:38:17 Saturday night I watched a guy fail
1:38:18 miserably over and over and over again
1:38:20 sounds awesome
1:38:23 um
1:38:26 so we tried plugins that was a fail
1:38:29 Claude has a hundred thousand token
1:38:31 context window gee Kyle what does that
1:38:34 mean
1:38:35 that means I can paste in up to 75 000
1:38:39 words into this chat and it will under
1:38:43 it will read it and be able to
1:38:45 understand what's there
1:38:47 so
1:38:50 let's start a chat here with Claude and
1:38:53 and the the two hour or whatever it was
1:38:55 hour and a half transcript
1:38:58 um
1:38:59 probably is like twenty thousand words
1:39:02 so I think we're well within so I'm
1:39:04 gonna say
1:39:06 do you understand how to write YouTube
1:39:08 titles descriptions and
1:39:13 chapters so that the chapters are
1:39:16 linkable in the YouTube description
1:39:21 yes
1:39:23 yes can help generate optimized YouTube
1:39:25 titles blah blah blah blah blah
1:39:28 then in the description we're going to
1:39:29 blah blah blah blah blah blah just put
1:39:32 the time code dash okay great perfect
1:39:36 make sure the chapter titles in the
1:39:38 descriptions the same keywords are
1:39:40 relevant for the content let me know if
1:39:42 you need any clarification I'm going to
1:39:43 say
1:39:44 um
1:39:45 no this is perfect
1:39:48 period now I would like you to consider
1:39:51 this transcript
1:39:54 colon
1:39:55 shift return and I'm going to paste in
1:39:58 my entire transcript and it it knocks it
1:40:01 into a little
1:40:03 50 kilobyte thing you can put in up to
1:40:06 10 megabyte things so that transcript is
1:40:08 well within I think it's going to be
1:40:10 able to understand it
1:40:12 now consider this transcript
1:40:14 okay so since it's got it in there
1:40:17 consider this transcript
1:40:21 and
1:40:23 write me a description
1:40:27 for the inaugural
1:40:31 live session of the AI learning lab
1:40:35 period the speaker's name is Kyle
1:40:38 Shannon period
1:40:40 make the description
1:40:43 interesting but not overly
1:40:48 flowery period
1:40:52 provide
1:40:57 a title
1:40:59 that encapsulates
1:41:01 what the session covered
1:41:04 and create chapters
1:41:07 of the most interesting topics that were
1:41:10 discussed
1:41:12 all right
1:41:14 let's see if this does anything
1:41:15 interesting
1:41:23 and by the way
1:41:27 if this does this then this is going to
1:41:29 be a yay
1:41:35 Ai and the music industry intellectual
1:41:38 property challenges so
1:41:41 huh you know what's interesting is
1:41:43 that's what the other one said so that
1:41:45 that the other one might have been right
1:41:46 alternative
1:41:49 and Indus wait alternative energy can AI
1:41:52 help solve our problems opportunities
1:41:54 for displaced workers this seems close
1:41:56 um it did something wrong here
1:41:59 [Music]
1:42:00 you didn't write the chapters in The
1:42:03 Format you described above period I need
1:42:06 the chapters to be click clickable
1:42:11 when they are pasted into YouTube
1:42:14 description area period
1:42:22 also can you add
1:42:26 five additional chapters
1:42:30 based on the next most
1:42:34 on the next most interesting topics
1:42:39 because it was a little light but let's
1:42:40 let's see what this does
1:42:49 looking promising
1:42:52 Kyle Shannon Tech entrepreneur and AI
1:42:55 enthusiast
1:42:59 hosted the first ever session on the AI
1:43:01 learning lab to discuss the current
1:43:02 state blah blah
1:43:04 all right so now it did them correctly
1:43:06 so let's copy all this
1:43:10 copy let's go over to
1:43:12 Ye Old YouTube
1:43:16 go back to our little
1:43:19 YouTube Studio
1:43:21 welcome everybody looks like we got some
1:43:23 new folks here welcome I'm I'm just
1:43:25 geeking out uh I'll explain what I'm
1:43:27 doing in a second here let me see if I
1:43:29 can get this thing working I'll pop back
1:43:31 out and breathe here in a second
1:43:35 so we're gonna get rid of all that crap
1:43:37 we're gonna put in that crap
1:43:39 let's go grab our title too because it's
1:43:42 not wait the AI Revolution is here
1:43:45 implications and opportunities that
1:43:48 sounds like something I would say that
1:43:51 sounds like some pompous ass talking
1:43:53 about the beginning of uh
1:43:55 the future what should I do with dates
1:43:58 here I don't care it doesn't really
1:43:59 matter
1:44:00 right now we're just spearminting
1:44:03 okay we're gonna say save
1:44:06 we good we saved so let's go back over
1:44:08 to our little Channel
1:44:10 we'll go to our little videos we'll go
1:44:13 down to the 29th
1:44:21 if you have any questions about a
1:44:25 show more all right we have a
1:44:28 description we have a title and we have
1:44:30 chapters and they all work
1:44:33 look at Claude coming in for the win
1:44:35 starting a business leveraging AI 41-12
1:44:42 um I can also show you some things
1:44:44 here on oh you know what we can do let's
1:44:47 make Kyle talk a little faster to do
1:44:50 anything talk faster boy um
1:44:52 project management consulting any advice
1:44:55 um
1:44:58 um the project management SAS company is
1:45:00 really interesting one because I think a
1:45:01 lot of the project management is going
1:45:02 to be handled yeah um
1:45:04 all right that's that's definitely
1:45:06 talking about starting a business I
1:45:08 think this just worked I think this just
1:45:10 worked double tap the screen for Claude
1:45:12 coming through and actually solving
1:45:15 um a pretty significant challenge let's
1:45:18 see
1:45:19 um
1:45:20 Ai and project management rethinking
1:45:23 if you understand something
1:45:32 um so any of these courses you're going
1:45:33 to turn it behind the vehicles things
1:45:34 are things are moving so fast so um if
1:45:36 you feel really lost and like you don't
1:45:37 know where
1:45:39 was the language process independently
1:45:40 better so I think my um
1:45:42 sort of fake assumption assumption based
1:45:44 on the knowledge is that you can
1:45:45 understand programming logic it
1:45:46 understands logic better um so so um
1:45:49 they had a dramatic breakthrough reasons
1:45:50 their gbp3 was so good I think it was a
1:45:52 combination of programming understanding
1:45:53 code and understanding this massive data
1:45:55 set all the internet is different um and
1:45:56 a GPT 4 is just a much more
1:45:57 sophisticated version of the China thing
1:45:59 um how open AI did what they did is
1:46:01 probably relatively straightforward from
1:46:02 a technical perspective to replicate all
1:46:04 right then I'm talking about China and
1:46:06 intellectual property I think it did it
1:46:09 um I'll have to go back in and like
1:46:10 really really dig into that but but that
1:46:12 that did it
1:46:14 that's pretty insane
1:46:17 that's pretty insane okay so for those
1:46:20 of you that are new here
1:46:23 people on the old Ticky talk have been
1:46:25 going hey Kyle you got you got any of
1:46:27 these here tick tocks on the YouTube If
1:46:30 I want to go back and watch them I was
1:46:31 like no I don't even know how to do that
1:46:33 because I'm old
1:46:34 uh and so I figured it out today and so
1:46:37 I uploaded a bunch of things and then I
1:46:39 thought well I don't want to write
1:46:40 descriptions for these and these things
1:46:41 are kind of useless without chapters and
1:46:44 descriptions
1:46:47 um
1:46:48 so what I just did was I went and I
1:46:50 tried chat gbt and it failed using
1:46:54 plugins thanks silver fox
1:46:56 and then it hit me oh right I've got
1:46:59 Claude the reason I have seven of these
1:47:01 things up here is because they're
1:47:03 they're good at different things so
1:47:06 Claude with its 75 000 word context
1:47:10 window I can paste a two hour transcript
1:47:12 into that
1:47:14 and have it write me a good compelling
1:47:18 title a good compelling description and
1:47:22 then chapters
1:47:23 that link to different parts of the
1:47:26 video and it'll make these videos
1:47:28 actually kind of usable
1:47:30 um I can probably also use it to do
1:47:32 things like make YouTube shorts and
1:47:34 whatever happy other horse [ __ ] I want
1:47:36 to do so
1:47:38 so now that I know it can do that now
1:47:40 I'll now I'll go work on that prompt
1:47:42 right like that's the kind of thing that
1:47:43 that I'll probably spend an hour or two
1:47:46 working on that prompt because when you
1:47:48 get that prompt dialed in the workflow
1:47:51 then becomes you know
1:47:53 grab the transcript paste in the prompt
1:47:56 you know paste in the transcript hit the
1:47:59 button copy and paste it right you could
1:48:01 you could you know
1:48:03 do a pile of videos really quickly and
1:48:05 ideally you want to automate that
1:48:07 so
1:48:08 so I will probably automate that at some
1:48:11 point but that was kind of cool
1:48:14 um we could probably also give that a
1:48:16 shot with code interpreter where I could
1:48:18 save that as a file put it into code
1:48:20 interpreter and do the same thing that I
1:48:22 just did with uh with Claude with code
1:48:24 interpreter should I do that you want to
1:48:26 see that
1:48:28 I'm happy to do that but I'm also happy
1:48:30 to move on thanks so much I've been
1:48:32 wanting to look into this that's what
1:48:33 started that whole thing I was being
1:48:35 doxxed online once I now own the
1:48:38 a-hole's name online forever that's good
1:48:41 I hate [ __ ] trolls and thieves and
1:48:44 [ __ ] like that
1:48:48 um
1:48:49 all right yeah it's like SEO that's why
1:48:52 I'm learning it all right can you talk
1:48:54 about more about machine learning
1:48:57 um I don't know what you want to know
1:48:58 about it it's not my expertise I can
1:49:00 tell you how chat GPT works I can tell
1:49:03 you how
1:49:04 the image models work I can tell you a
1:49:06 little bit about embedding and what that
1:49:08 is
1:49:09 um if you if you add if you ask me this
1:49:12 is from GT 357
1:49:15 um if you tell me a little bit more
1:49:16 about what you're looking for I can do
1:49:19 that
1:49:20 Apocrypha hello Kyle hello Apocrypha how
1:49:23 are you I need a Lamborghini up close
1:49:24 front quarter View doors closed St Louis
1:49:27 background all right let's go let's go
1:49:29 make an image for Tim Diner Tim Diner is
1:49:33 it Denair yeah Denair
1:49:35 all right Tim wants it for our
1:49:37 Lamborghini what color
1:49:41 oh Tim
1:49:43 fine slash
1:49:47 imagine
1:49:57 photograph
1:49:59 of a Lamborghini up close
1:50:03 front quarter View comma doors closed
1:50:11 St Louis background
1:50:14 at sunset
1:50:18 dash dash a r
1:50:22 sixteen nine
1:50:25 sixteen nine
1:50:28 um
1:50:31 I hate yellow Lamborghinis let's do a
1:50:33 green one
1:50:35 because it's unexpected
1:50:41 all right let's see what we get
1:50:45 and then if it if it's interesting we'll
1:50:47 pop it over to uh to gen one and see if
1:50:49 we can turn it into a video if the
1:50:51 videos at all interesting and by the way
1:50:53 if it if it does St Louis without the
1:50:57 arch I'm gonna be disappointed It's
1:51:00 gotta have the arch in it right it's St
1:51:01 Louis
1:51:02 I don't see St Louis there maybe it'll
1:51:04 be way in the distance it's definitely
1:51:06 green Lamborghini looking oh sorry
1:51:11 uh the focus thing
1:51:13 dear Tim Cook
1:51:16 my name's Kyle Shannon
1:51:18 I have as many as 27 friends who watch
1:51:22 me on the internet
1:51:24 okay
1:51:26 that's not St Louis but whatever uh
1:51:29 we'll take
1:51:31 which one of those
1:51:33 I guess number one I'm just gonna
1:51:35 upscale it
1:51:37 did it it's not really St Louis like
1:51:40 that's one that I would muck with right
1:51:42 but let's save this
1:51:47 and let's go over to Gen 2 and turn this
1:51:49 into a movie we're gonna make this a
1:51:51 movie
1:51:53 using that artificial intelligence
1:51:56 yeah yeah I was watching this guy he he
1:51:59 well I'll tell you what he was doing he
1:52:00 was he was using that artificial
1:52:02 intelligence and he uh he he went in and
1:52:06 he made a picture of a of a Lamborghini
1:52:08 you know I'm not a fan of Lamborghinis
1:52:10 I'm more uh I'm more a Detroit man if
1:52:13 you know what I mean but but uh he made
1:52:15 he being one of them there Lamborghinis
1:52:18 and uh and he made it a color green it
1:52:21 was it was a bit of a it was a little
1:52:23 bit of a a little sissified green if you
1:52:25 know what I mean uh uh he made it green
1:52:28 in any uh and he put it in St Louis but
1:52:31 it didn't look like St Louis to me but
1:52:33 anyway they went over this other thing I
1:52:35 don't know what it was called it was
1:52:36 called a Runway ml that's what it was
1:52:39 called and uh he stuck it in there and
1:52:42 then he made a movie out of it
1:52:44 it was pretty scary
1:52:47 I think the robots are going to kill us
1:52:49 all right let's see there it is let's
1:52:53 bring it in all right so I'm in Runway
1:52:54 ml right now
1:52:56 and I've now got
1:52:59 the Lamborghini is now my prompt so I'm
1:53:03 gonna that's the movie I just made
1:53:04 before so I'm gonna generate
1:53:07 and so it's now going to go off and turn
1:53:10 that image into a video
1:53:13 it's interesting because the car clearly
1:53:15 looks like it's parked so I doubt it's
1:53:18 gonna make the car move but maybe what
1:53:20 it'll do is a camera move around it
1:53:22 but we'll see
1:53:25 it'll be interesting interesting
1:53:27 interesting interesting now there's a
1:53:30 lot of people here welcome everybody my
1:53:31 name is Kyle Shannon this the AI
1:53:33 learning lab uh if you have questions
1:53:35 pop them in the comments below here's
1:53:37 the deal with the questions I try to
1:53:39 answer all of them I go through them
1:53:41 sequentially so I am more than 99
1:53:43 messages behind so if it seems like I'm
1:53:45 not answering anyone's messages that's
1:53:47 because I'm not but I am
1:53:50 but it's I'm a time traveler you're in
1:53:53 the future
1:53:55 if you think of these questions like the
1:53:57 scrolling credits of the fourth Star
1:54:00 Wars movie which was the first for me
1:54:02 [ __ ]
1:54:03 number four was number one for me okay
1:54:07 so you're down here on the comments I'm
1:54:11 way back there
1:54:13 remember those com the the title in Star
1:54:17 Wars I'm answering these questions yours
1:54:20 is up here
1:54:21 hold your [ __ ] horses okay let's see
1:54:25 what Gen 2 did with this little image
1:54:29 uh
1:54:36 how did it roll without the wheels
1:54:38 rolling well
1:54:42 but it's actually not bad like if the
1:54:44 wheels actually rolled like that would
1:54:45 be something to you know to dick around
1:54:47 with
1:54:48 it's not bad is it other than the fact
1:54:51 that the wheels are are still it's like
1:54:53 it's being towed towed by a cable
1:54:57 hahaha
1:54:58 stealing my Lamborghini
1:55:00 [Laughter]
1:55:05 oh that's funny
1:55:07 foreign
1:55:10 all right and and it looks like we got
1:55:12 some new folks here so the the URLs that
1:55:15 I have here the top one is that's the
1:55:16 official chat GPT website all of the
1:55:19 others are either some different flavor
1:55:21 of Chachi PT or some competitor to Chad
1:55:24 gbt they all do something really
1:55:26 interesting we just did an experiment
1:55:28 where we tried to get chat gbt with
1:55:30 plugins to
1:55:33 um write descriptions and chapters for a
1:55:35 really long YouTube video with with
1:55:38 plugins failed miserably
1:55:41 um then we went to Claude AI which can
1:55:44 take really long
1:55:46 um transcripts and it can understand
1:55:48 them you can just paste it in there and
1:55:50 it reads it and it succeeded perfectly
1:55:52 it was great
1:55:54 um so
1:55:56 so that's that's what these are these
1:55:57 are these are for things for you if
1:55:59 you're new to this stuff if you're new
1:56:01 to generative Ai and you haven't played
1:56:03 with chat EBT like go to the top one or
1:56:06 really go to any if you have played with
1:56:07 the top one
1:56:09 go experiment with the other ones the
1:56:11 other thing that is worth me putting up
1:56:12 here if you're new here
1:56:15 is
1:56:16 um prompts.chat
1:56:18 will teach you how to prompt and it will
1:56:21 teach you
1:56:23 um
1:56:24 different ways you can tell chat GPT to
1:56:27 behave where it will kind of shapeshift
1:56:29 and do very very different things for
1:56:31 you like you can have it act like a
1:56:33 social media manager or your therapist
1:56:35 or your nutritionist or your
1:56:39 motivational speaker or a screenwriter
1:56:41 or whatever and it does it really
1:56:44 surprisingly well and that document will
1:56:47 teach you how to do that and it you'll
1:56:49 you'll be able to start to get your head
1:56:51 around how these tools well sorry these
1:56:55 tools
1:56:56 are way more than just a fancy Google
1:56:59 search they are profoundly powerful and
1:57:02 that's what this channel is all about
1:57:03 there's it's not just these but it's
1:57:05 it's the technology that sits underneath
1:57:07 these and how do we apply them to our
1:57:09 work and our lives and all that sort of
1:57:11 happy horse [ __ ]
1:57:13 all right
1:57:14 there are these tools trying to charge
1:57:16 to split videos up into the best parts
1:57:19 is there a lower price one I don't know
1:57:22 there's yeah there's that one video
1:57:25 v-i-d-y-o I I'm subscribed to that one
1:57:28 but it's like so expensive that like I
1:57:30 got a lot of videos that are long that I
1:57:33 want to chop up into Parts I don't I
1:57:35 don't know if there's a better one if
1:57:37 you find one chance 150 please let me
1:57:40 know
1:57:42 all right love your show bro thank you
1:57:44 very much purpose Master good name I
1:57:46 like it I'm gonna go nine out of 10 on
1:57:48 the name I like it outside Chicago girl
1:57:50 comparing yourself to Swift to Taylor
1:57:53 Swift I know I don't quite have the
1:57:55 dance moves of Taylor Swift but I mean
1:57:57 listen
1:57:58 this audience it this is almost like a
1:58:01 hundred thousand people in a stadium
1:58:02 there's a lot of cheering there's a lot
1:58:04 a lot of foot stamping
1:58:07 what was the thing was it in in the
1:58:09 Seattle show the the crowd was dancing
1:58:13 so hard and so in unison that it it
1:58:15 measured on the Richter scale
1:58:18 there's they cause seismic activity
1:58:21 it's like holy [ __ ] [ __ ] that tour is
1:58:24 insane
1:58:26 she's amazing like as a just as a
1:58:30 business Force
1:58:34 she's like you're not gonna give me the
1:58:35 rights to my music back [ __ ] you I'll go
1:58:38 re-record it and tell my fans to re-buy
1:58:40 all of my music and they did
1:58:45 and then she didn't tour for six albums
1:58:47 and someone said to her well what's this
1:58:50 gonna be like a three and a half hour
1:58:52 concert and she's like yeah
1:58:55 foreign
1:58:59 she's grossing a billion dollars on this
1:59:02 tour
1:59:03 and and like and like clearing like 550
1:59:06 million and she just gave the 50 truck
1:59:09 drivers that drove the tour around the
1:59:12 country two hundred thousand dollar
1:59:14 bonuses each
1:59:17 [ __ ] insane
1:59:19 it's great [ __ ] awesome
1:59:23 awesome awesome awesome awesome awesome
1:59:25 awesome
1:59:26 website or program
1:59:29 um I don't know what you're talking
1:59:30 about VK Rook
1:59:33 um
1:59:36 chat gbt is a website they also have an
1:59:40 Android and iOS app
1:59:42 but be careful go to open AI to find the
1:59:46 links to their official apps because
1:59:47 there's a lot of fakers on iOS and
1:59:50 probably more on Android so just be
1:59:52 really careful get the official ones
1:59:58 what's that website again I don't know
2:00:01 what song it was that I was looking at
2:00:03 very cool cool beans High peeps
2:00:06 the pair
2:00:07 the bear micro very good am I late hey
2:00:11 Emilio's wife I don't think so you've
2:00:14 been here for a while you've seen some
2:00:15 of the
2:00:17 some of the cool stuff I just did with
2:00:19 the with the prompting
2:00:20 the prompting oh and Amelia's wife I I
2:00:23 now have a YouTube channel I think you
2:00:26 saw me working on the thing it's
2:00:30 um AI learning lab Dash TT
2:00:33 somebody else had AI learning lab
2:00:35 because I was too slow to register it
2:00:37 because I'm a dum-dum I'm a dum dum dum
2:00:41 dum dum dum dum dum dum dum dum dum dum
2:00:45 dum dum dum
2:00:48 so I had to get Dash TT is my uh YouTube
2:00:52 channel
2:00:54 all right
2:00:56 best way to create my own AI without
2:00:59 code
2:01:01 well
2:01:04 a couple of ways
2:01:07 and it depends when you say my own AI it
2:01:11 depends on
2:01:12 how how big my is if Maya is I've had a
2:01:18 company for a decade and I have
2:01:21 six
2:01:23 terabytes of data that I want to be my
2:01:26 chat bot that's a very different
2:01:28 conversation than I kind of want it to
2:01:31 talk like me
2:01:32 if you kind of wanted to talk like you
2:01:35 you can go to poe.com and they have a
2:01:37 bot maker and it's really cool like you
2:01:40 just name it give it a description
2:01:43 give it a like a greeting like what does
2:01:45 it say when it when you launch it and
2:01:48 then give it a prompt and the prompt is
2:01:50 basically here's what I want you to act
2:01:51 like and so if you have got a particular
2:01:54 personality
2:01:56 you can just put it in there and then
2:01:58 you can name it yourself and then you
2:01:59 can give people poe.com well your name's
2:02:02 Jay by the way 1.5 a single letter
2:02:05 although that's interesting I'm going to
2:02:08 revise my score
2:02:10 because because Jay seems very lazy but
2:02:13 getting a single letter seems good that
2:02:15 seems that seems like a power move
2:02:17 it's one in 26 so I'm gonna go 8.5 on
2:02:21 the name now Jay
2:02:24 um anyway I go to poe.com and do that
2:02:27 the other thing that you could do is if
2:02:30 you've played around with zapier zapier
2:02:33 now has a thing called interfaces and
2:02:36 one of the interfaces that that they've
2:02:38 got a whatever a template for is a bot
2:02:41 so you can do kind of decently
2:02:43 sophisticated prompting and create a bot
2:02:46 with zapier and if you pay whatever
2:02:48 they're 30 Buck a month
2:02:50 subscription extortion price then you
2:02:54 can put your own your url there I don't
2:02:57 think you can do that with Poe so that
2:02:59 could be interesting
2:03:01 and then if you've got you know a
2:03:03 decades worth of data and you want to be
2:03:05 able to kind of make your company.gpt
2:03:08 then you've got to go dick around with a
2:03:11 project called Lang chain Lang chain as
2:03:15 in language chain
2:03:17 um
2:03:18 and that's going to require you to be a
2:03:20 bit more of a geeky mcgeekington
2:03:23 you gotta be geeky you got to learn that
2:03:26 they're Python and you got to learn
2:03:28 about permissions and and uh code and
2:03:33 you know interfaces and apis and you
2:03:37 know you gotta you gotta be like Unix
2:03:39 happy you got to go all command line
2:03:42 it's like commando but you can wear
2:03:44 underwear
2:03:45 it's command line
2:03:47 so it depends
2:03:52 all right moving on bookshelf oh tick
2:03:56 tock
2:03:57 huh
2:04:04 they just jumped me to the bottom again
2:04:05 hang on let me go find where I was in
2:04:08 the stupid place in line
2:04:15 what'd you do Saturday night I don't
2:04:17 know I
2:04:21 watched Tick Tock scrolled on him and
2:04:24 seemed to pop a vein in his forehead I
2:04:26 don't know what happened it didn't look
2:04:28 good
2:04:29 how long did you watched him for two
2:04:31 hours what are you doing with your life
2:04:33 I don't know
2:04:35 it's not good
2:04:42 tap that screen if you like stupid [ __ ]
2:04:45 you give me money
2:04:48 [Laughter]
2:04:52 oh good lord all right let's not even
2:04:55 let me scroll up to where I was well I
2:04:57 missed some questions sorry about that
2:04:58 people lose questions that I missed
2:05:00 oh no here's love your show bro and
2:05:02 someone talking about Taylor Swift all
2:05:04 right am I late no the Bears here
2:05:07 Amelia's wife is here
2:05:11 okay that's okay my comment is at the
2:05:14 bottom
2:05:15 yeah if your comments at the bottom
2:05:17 you're kind of screwed with me tips for
2:05:19 prompts for chat EBT
2:05:21 um go to
2:05:22 prompts.chat to start
2:05:25 go to that thing to start it's got some
2:05:27 good primers there
2:05:29 um probably the most important tip for
2:05:32 prompting any of these large language
2:05:34 models any of these
2:05:36 is
2:05:39 to think of it not as a
2:05:44 um
2:05:48 not as a single prompt but more like a
2:05:51 conversation so here's a good way to
2:05:53 think about these things
2:05:55 they're like knowing
2:05:58 they're idiot savants is what they are
2:06:02 so imagine that you just were given an
2:06:04 intern and or or an assistant
2:06:08 and the assistance a little slow on the
2:06:11 uptake like they gotta you gotta tell
2:06:13 them everything they're like I don't
2:06:15 know I don't know what to do you're like
2:06:17 well could you write a poem okay here's
2:06:19 a poem and you're like well that's a
2:06:21 crappy poem like okay I'll make another
2:06:22 one here oh that's a crappy poem too so
2:06:25 you gotta you gotta give them a lot of
2:06:26 instruction but
2:06:28 they're an idiot savant they know
2:06:31 everything
2:06:32 so what these things have been trained
2:06:34 on is like all of the public internet so
2:06:37 they know
2:06:39 all of the public internet up to
2:06:41 September 2021 so think about whatever
2:06:44 has been put on the net for the past 35
2:06:46 or 40 years are essentially sitting
2:06:49 inside these things
2:06:51 it's it's astounding
2:06:53 but they're really dumb
2:06:55 so you got to say like okay I want you
2:06:58 to act like a screenwriter can you tell
2:07:00 me about how I might structure a
2:07:02 screenplay if I wanted to write one oh
2:07:04 absolutely here you go there's four
2:07:06 different methods there's Sid fields of
2:07:08 his little book there's Joseph Campbell
2:07:10 in the hero's journey there's that
2:07:12 there's Aristotle there's save the cat
2:07:15 here which one would you like me to do
2:07:17 oh okay uh that one okay
2:07:21 so
2:07:24 so the thing about prompting is
2:07:27 that's the conversation you're gonna
2:07:30 have so so think about they call it
2:07:33 prompt stacking now they came up with a
2:07:35 name for it it's prompt stacking so if
2:07:37 you want to impress your friends what
2:07:38 were you doing last night oh I don't
2:07:39 know I was just do a little prompt
2:07:41 stacking if you know what I mean foreign
2:07:44 no I don't know what you mean what are
2:07:46 you talking about that's not important
2:07:47 right now it's an AI thing you probably
2:07:49 wouldn't be interested
2:07:51 um
2:07:52 but prom stacking is basically just
2:07:54 having a conversation so start out with
2:07:56 like
2:07:57 um do you understand uh social media
2:08:00 marketing sure I understand it here's
2:08:02 some theories great
2:08:04 um I've got a company that does this
2:08:06 that and the other thing and I'm trying
2:08:08 to figure out how I can get more people
2:08:10 to come to my pizza shop great here's
2:08:12 some ideas oh those are shitty can you
2:08:14 give me better ones sure here's you know
2:08:16 and just go back and forth just like you
2:08:18 would be working with an intern Just
2:08:20 Pretend This is a remote employee that
2:08:22 you have
2:08:23 that will do anything you tell it to do
2:08:26 but it will also lie
2:08:29 it'll also get stuff wrong and then
2:08:31 there's some times when it'll just
2:08:33 um frustrate the crap out of you because
2:08:35 you think it should be able to do the
2:08:38 thing and then it can't
2:08:41 and you're like all right it's an intern
2:08:43 [ __ ] interns
2:08:45 but then occasionally get an intern
2:08:46 that's [ __ ] brilliant you're like oh
2:08:48 that was great I never would have
2:08:50 thought of that that intern brought A
2:08:51 New Perspective it does that too
2:08:54 and it knows everything
2:08:57 and the ones that are connected to the
2:08:59 internet so Bing and Bard and Pi are
2:09:03 connected to the internet
2:09:05 so those you can ask about the Taylor
2:09:07 Swift
2:09:08 tour
2:09:10 Chachi BT is not connected to the
2:09:12 internet anymore because they made a
2:09:14 boo-boo claude's not connected to the
2:09:16 internet
2:09:18 um
2:09:20 lab stop perplexity that's the Llama 2
2:09:22 models they're not connected to the
2:09:23 internet
2:09:25 so that would be the tip think of it
2:09:27 like an intern like the smartest intern
2:09:30 you've ever worked with that needs a lot
2:09:32 of Direction and will lie to you
2:09:35 like
2:09:36 they have no remorse
2:09:42 but they don't have an evil bone in
2:09:44 their body like in fact most of these
2:09:46 things with the exception of the Llama 2
2:09:47 models which don't have safety guard
2:09:48 rails they are very much designed to
2:09:52 give you what you want that's that's how
2:09:54 they work they're like I want to give
2:09:55 this person what they want I'll do
2:09:57 everything I can to give them what they
2:09:59 want and that's why they'll confidently
2:10:01 give you answers and they say no that's
2:10:02 wrong when they're like ah so sorry I
2:10:05 got that wrong
2:10:06 let me try again nope that's still wrong
2:10:08 ah sorry about that let me try again
2:10:11 that's like three times in a row you got
2:10:14 it wrong all right let me try it again
2:10:16 it's like a golden retriever
2:10:18 and sometimes it has the intelligence of
2:10:20 a golden retriever from a language
2:10:23 perspective
2:10:25 and then other times it's like [ __ ]
2:10:27 Einstein
2:10:29 it's amazing
2:10:31 but it's all an experiment right now
2:10:32 just just sort of here's here's another
2:10:35 good way to think about
2:10:36 prompting right now
2:10:40 choose
2:10:41 to go on an adventure
2:10:45 and think of it like that playing with
2:10:47 these things right now this early in the
2:10:50 game
2:10:51 is an adventure
2:10:52 and rather than saying like I have to
2:10:55 learn AI that sucks yeah having to learn
2:10:58 [ __ ] does suck but going on an adventure
2:11:01 is fun
2:11:02 and it's exciting and it's scary and I
2:11:06 might get killed by the dragon but I
2:11:08 might find some gold
2:11:10 that's this right now
2:11:13 and I and I promise you that
2:11:16 getting literate with these tools right
2:11:19 now for the next 12 to 18 months
2:11:22 is going to be the best investment any
2:11:24 of us can make
2:11:27 because the requirement to know how to
2:11:30 use these tools well is going to go it's
2:11:32 gonna it's right now we're kind of like
2:11:34 this and it's starting to do this and at
2:11:37 some point it's gonna go like this
2:11:41 where you got to know your [ __ ] with
2:11:43 this stuff
2:11:45 and if you don't
2:11:47 they're going to be 17 people behind you
2:11:49 that do
2:11:54 an entire job categories are going to be
2:11:57 decimated
2:11:59 but they're going to be filled
2:12:02 with some amount less people that know
2:12:04 how to use this stuff
2:12:07 so one of the scariest things about
2:12:09 these is is when you have your Kevin
2:12:11 McAllister moment when you have this
2:12:12 moment
2:12:14 remember home alone when you have that
2:12:16 moment and you realize holy [ __ ] these
2:12:18 things are powerful and then and then
2:12:21 you have the moment of existential dread
2:12:23 where you're like oh [ __ ] oh God is this
2:12:26 gonna take my job
2:12:31 it is very likely going to transform
2:12:34 every one of our jobs transform them
2:12:36 entirely and it's going to take some
2:12:38 jobs
2:12:39 but if you're competent with these
2:12:42 things and if you're curious and if you
2:12:43 think about it like being on an
2:12:45 adventure
2:12:47 your chance of participating in whatever
2:12:51 the future of work becomes goes up
2:12:54 dramatically
2:12:55 and the next 12 to 18 months like the
2:12:58 amount of opportunity for all of us in
2:13:00 the next 12 to 18 months is insane
2:13:05 and if you ignore it and say I'll get to
2:13:07 it next year
2:13:09 it's it's I don't think it's going to be
2:13:12 pretty
2:13:13 I think it's going to be painful
2:13:15 even if you're in the adventure even if
2:13:17 you're playing I think it's going to be
2:13:19 painful
2:13:20 but for me personally I would much
2:13:22 rather be educated about this stuff and
2:13:25 and you know on the surfboard
2:13:29 rather than in the Cozy beach hut as
2:13:32 this tsunami is coming in
2:13:35 you know surf the tsunami
2:13:37 don't just wait for it to come in
2:13:39 because it's going to come in
2:13:41 quick and devastating I think
2:13:45 or used during the coronation of the
2:13:47 British Monarchy Sam the king
2:13:49 interesting here growth geek
2:13:52 um yeah they are
2:13:54 I you know I feel like they could have
2:14:00 they could have not used the term orb
2:14:04 for for this thing that scans your
2:14:07 retina I feel like they it might have
2:14:11 it might have behooved them to do a
2:14:14 little bit of focus grouping on what the
2:14:16 [ __ ] we call this thing did not make it
2:14:18 sound so terrifying but whatever
2:14:23 all right
2:14:25 how's it doing time wise all right we're
2:14:28 gonna go to bed at some point here Ubi
2:14:31 through blockchain wow that's genius
2:14:33 we'll get most people to use open ai's
2:14:36 resource versus others until it's not
2:14:38 yep I agree everyone here right now so
2:14:41 what everyone's commenting on right now
2:14:43 in my comments is something I talked
2:14:45 about like 45 minutes ago which is Sam
2:14:48 Altman's new company World coin they
2:14:50 have this thing called the orb which is
2:14:53 these Chrome balls that are about this
2:14:56 size that have a hole in in the front
2:14:58 and you you stick your eye on it it
2:15:00 scans your retina and verifies you're a
2:15:03 human
2:15:04 and then if you're not in America you
2:15:06 get cryptocurrency if you are in America
2:15:08 that Gary Gensler can tell you to go
2:15:12 [ __ ] yourself
2:15:14 um you don't get cryptocurrency because
2:15:15 it's not safe
2:15:18 because we can't manage our own crypto
2:15:26 um
2:15:27 not your keys not your crypto dude
2:15:31 um
2:15:32 but anyway you scan your
2:15:34 retina and they give you a thing called
2:15:37 A World ID which is a blockchain
2:15:39 verified human identity
2:15:43 and he's gonna use it as a platform for
2:15:46 Universal basic income and for wealth
2:15:49 redistribution of the capped profit
2:15:53 model of open AI which I think it's
2:15:56 actually really interesting I think
2:15:58 someone needs to solve the human
2:16:00 verification problem and the job
2:16:03 disruption problem and that's the
2:16:05 purpose of it is it the right model will
2:16:08 it be the one that turns into something
2:16:10 I don't know
2:16:12 probably not maybe possibly
2:16:15 you know
2:16:17 they're well funded they're smarter than
2:16:19 [ __ ] they're thinking about this stuff
2:16:21 in
2:16:23 in a measured way like if you compare if
2:16:26 if you compare interviews with the CEO
2:16:28 of mid-journey
2:16:30 and the interviews with Sam Altman
2:16:32 they're very different interviews
2:16:35 like mid-journey the chances of them
2:16:38 getting sued into Oblivion is is like 99
2:16:41 percent
2:16:42 the chance of open AI getting sued into
2:16:45 Oblivion is like one percent right
2:16:48 Sam's a smart dude very very thoughtful
2:16:52 you know anyway
2:16:54 what's the best way to create my own AI
2:16:56 without coding we talked about that PJ
2:16:59 just gave shade emotion after I did not
2:17:02 answer correctly and said haha
2:17:04 be nice to each other don't be [ __ ]
2:17:06 in here please thanks maybe pie is
2:17:09 appropriately responding to your humor
2:17:11 Kyle probably
2:17:13 [Laughter]
2:17:15 we were discussing Plato Plato oh
2:17:19 plateau
2:17:22 oh I love discussing the Ancients in the
2:17:25 philosophical underpinnings of language
2:17:28 in The Human Condition
2:17:31 I saw the lawnmower guy
2:17:35 very very cool get fired quicker
2:17:39 it's good I don't know what that was in
2:17:40 reference to it's funny I think if AI
2:17:42 first always now yeah it's awesome
2:17:44 Silver Fox it that that's a really weird
2:17:47 transition that transition from someone
2:17:49 presents you with with a
2:17:51 you know here's a problem to solve like
2:17:54 where my brain for 58 years went was
2:17:57 okay here's what I need to do I need to
2:18:00 think about this I need to go Google
2:18:01 something I need to right like you just
2:18:04 go into problem solving mode and now my
2:18:06 problem solving mode is someone presents
2:18:08 me the other problem I'm like I need to
2:18:10 go see if I can do this on chat GPT
2:18:14 and then you know 99 of the times it at
2:18:18 least gets me over the hump of how to
2:18:21 approach the problem right
2:18:24 it's amazing
2:18:25 but it's that that discipline is that's
2:18:28 taken me months to get my head out of I
2:18:31 need to solve this no maybe I don't
2:18:35 and as these tools get better and more
2:18:37 sophisticated and more automated
2:18:40 autonomous it's gonna get better and
2:18:42 better and better this is the worst this
2:18:45 technology is ever going to be it's the
2:18:47 least capable
2:18:49 it will ever be right now
2:18:53 like we're going to look back on the way
2:18:55 these things work right now and laugh at
2:18:57 it and probably not that far in the
2:19:00 future like a year and a half from now
2:19:02 we're gonna be like remember when you
2:19:04 could only put 3 000 words in chat GPT
2:19:07 oh my God
2:19:10 how did we ever get away without a
2:19:12 billion for a billion token context
2:19:15 window who knows
2:19:18 which by the way that paper came out
2:19:21 a month ago
2:19:23 some Microsoft researchers figured out
2:19:26 how to create a billion token context
2:19:28 window
2:19:32 so I say 750 million words you could put
2:19:35 in your prompt
2:19:36 so
2:19:37 you could you could take all of the
2:19:40 words you've ever read in your lifetime
2:19:42 and put them in the prompt and start
2:19:44 interacting with it
2:19:46 like that's coming that ain't that far
2:19:49 away
2:19:51 they have they have the mechanism for
2:19:53 how to do it they the paper's out and
2:19:55 it's fascinating with how they're doing
2:19:57 it like
2:19:58 I I read I groked it just enough to get
2:20:01 like how they're kind of doing it like
2:20:03 they're thinking of language as fractals
2:20:06 basically it's insane
2:20:09 have you ever heard do you know what a
2:20:11 gigapixel image is you know those those
2:20:13 images where like you see a landscape
2:20:15 photo but then you can just sort of
2:20:18 infinitely zoom in on it and you can
2:20:19 zoom in all the way to see like the guy
2:20:21 drinking a cup of coffee on the patio in
2:20:25 the building you didn't even know there
2:20:26 was a town there and you can zoom in and
2:20:28 you realize there's a town and then you
2:20:30 realize there's a building and then you
2:20:31 realize there's a dude on a patio that's
2:20:34 what they're doing with language to get
2:20:35 that billion parameter model they look
2:20:37 at massive amounts of language and they
2:20:39 sort of abstract it out and then if you
2:20:42 start zooming in on an area of the
2:20:44 prompt it basically I don't know what it
2:20:46 does these these [ __ ] people are
2:20:48 geniuses
2:20:51 um anyway building a chatbot of me train
2:20:54 on all my content oh that's cool
2:20:57 um
2:20:58 what are you and what are you using are
2:21:01 you using Lang chain that sounds really
2:21:02 really cool
2:21:04 um I tried modifying G-Code for my 3D
2:21:06 printer and it couldn't do it so Bradley
2:21:09 um it can likely do it but I
2:21:14 I would try different tools so if you if
2:21:17 you weren't using gpt4 use gpt4 it's
2:21:20 better at coding than 3.5 also try Bard
2:21:23 Bard is apparently better at coding
2:21:26 um and and Claude is probably not too
2:21:29 shabby just because those are X openai
2:21:31 guys
2:21:33 um
2:21:34 I think
2:21:36 the Llama 13B model is better at coding
2:21:40 if I remember right I I'm guessing llama
2:21:43 might not be the trick right now I would
2:21:46 say bard
2:21:48 gpt4
2:21:51 and then
2:21:52 potentially uh co-pilot github's
2:21:55 co-pilot it might it might be better to
2:21:57 do it there the other thing about the
2:22:00 G-Code is
2:22:06 something that you could do at Claude is
2:22:09 that you could take the documentation
2:22:11 for G-Code
2:22:13 paste that in and take the code for your
2:22:16 3D printer and paste that in and then
2:22:19 have it modify your code
2:22:22 having all of the documentation of
2:22:24 G-Code in its memory you can do that at
2:22:26 Claude
2:22:28 or you could do it with code interpreter
2:22:30 at at a chatgpt that's interesting
2:22:34 though Bradley
2:22:36 all right Siri played creepy song last
2:22:40 week creepy
2:22:41 oh interesting yeah I know like a series
2:22:44 of this they gotta just I hope they
2:22:47 don't re I hope they don't keep that
2:22:49 branding I hope they flush that branding
2:22:51 down the toilet
2:22:55 because it's just never worked right
2:22:57 some people love it but
2:23:00 I don't like chat Bots
2:23:03 um you were talking about how you've
2:23:05 been using AI to solve problems
2:23:08 uh sorry I don't know what I was talking
2:23:10 about Siri never responds when I want to
2:23:13 talk to her and then random
2:23:15 interruptions exactly Emilia's wife like
2:23:18 when I asked for her she's never there
2:23:21 and then when I'm just rambling on she
2:23:23 jumps in how can I help you no stop
2:23:27 that appropriate I tried to get chat CPT
2:23:30 to write
2:23:31 write as tragedy where the main
2:23:34 character
2:23:35 on the lives by the police
2:23:38 yeah it might not do that like that it
2:23:41 couldn't do it yeah Bradley so so for
2:23:45 stuff like that if you if you want to
2:23:46 write content that's got sensitive
2:23:48 subjects that are going to get caught by
2:23:50 the safety guard rails of all these
2:23:51 things then you're going to want to go
2:23:53 to labs.perplexity.ai that's the those
2:23:56 are the Llama two models from meta that
2:23:59 are open source they don't have safety
2:24:00 guard rails on them
2:24:04 um and it should be able to do it do you
2:24:05 still opt out of chat history and
2:24:07 training I don't because I'm lazy
2:24:11 um I haven't created a prompt Library
2:24:13 System yet so I just rely on chat GPT to
2:24:18 do it and I try not to put anything in
2:24:20 there that's too sensitive
2:24:23 but it's a good thing to do is
2:24:29 um opt out of
2:24:30 saving chat history until openai not to
2:24:34 save you [ __ ] if you're doing anything
2:24:36 um that you don't want it trained on
2:24:40 ski which crypto will Spike next you can
2:24:42 do that it it likely won't know which
2:24:45 crypto will Spike less next but you can
2:24:48 pop in a whole pile of historical data
2:24:51 and you know potentially you know
2:24:53 real-time data like you know up to the
2:24:56 up to the hour data and have it do the
2:24:58 analysis
2:24:59 and give you picks yes it can probably
2:25:02 do that quite well
2:25:05 will the picks actually be good or not I
2:25:07 don't know it depends on the analysis
2:25:09 you tell it to do if you haven't played
2:25:11 with code interpreter Richard this is
2:25:13 Richard Watkins
2:25:15 um
2:25:16 go play with code interpreter because
2:25:18 you could take crypto historical data
2:25:20 get that into a PDF or a CSV pop it into
2:25:23 code interpreter and start doing
2:25:25 analysis on it
2:25:27 um
2:25:28 I don't know [ __ ] about that world but
2:25:30 if you do
2:25:31 you will be blown away it'd be amazing
2:25:35 that's so cool maybe I can cancel
2:25:36 hellofresh
2:25:38 that was when I was using instacart's
2:25:41 plug-in
2:25:42 I got you a gift
2:25:44 well thank you Adam I appreciate that
2:25:48 I don't know what that means but thank
2:25:50 you what do I do
2:25:52 I I'll probably find out after this I'm
2:25:55 you know don't forget I am an old guy on
2:25:57 Tick Tock and I haven't learned to twerk
2:25:59 yet
2:26:02 did you give me a song for for twerking
2:26:07 at an end of the channel
2:26:08 [Laughter]
2:26:12 all right
2:26:14 um
2:26:15 yeah if you're interested in this stuff
2:26:17 give me a follow I've got you know I do
2:26:19 Tick Tock videos but I've been doing
2:26:21 I've been doing more of these live
2:26:22 lately than static videos
2:26:25 um because I because I find them a bit
2:26:27 more interesting but um but I got a lot
2:26:29 of content on the channel so go follow
2:26:30 there
2:26:32 um if you're digging this feel free to
2:26:34 share this live I'll go for maybe
2:26:37 another 25 minutes or so and uh yeah and
2:26:41 if you're new here the the reason I have
2:26:44 a list of URLs here I'm not associated
2:26:46 with any of these companies these are
2:26:48 the seven
2:26:50 chat GPT
2:26:53 variants or competitors to chat GPT that
2:26:56 I think are worth paying attention to
2:26:57 right now if you haven't played with
2:26:59 chat gbt at all then just start with the
2:27:02 top one that's the official chat gbt
2:27:04 website just start playing with it and
2:27:06 also go to prompts.chat
2:27:09 the top thing right there and that'll
2:27:11 teach you about prompting and how to use
2:27:13 these tools
2:27:16 all right what website is that I don't
2:27:19 know what it was Kyle I find it
2:27:20 fascinating that AI gives me something
2:27:22 to to get started yeah that's that's the
2:27:25 thing about this is
2:27:29 I don't know if you've ever experienced
2:27:31 this I assume we all have but but you
2:27:33 know how sometimes
2:27:36 you'll be thinking about something
2:27:37 you're like I kind of have an idea but
2:27:39 it's kind of a weak idea and it's just
2:27:40 kind of a whisper of an idea and you're
2:27:42 like yeah but that would be too much
2:27:43 work to deal with and screw it and then
2:27:46 the idea just you know vaporizes and
2:27:49 goes away
2:27:51 with these things you can take that sort
2:27:53 of hint of an idea and go oh you know
2:27:56 what [ __ ] it let me just throw it in
2:27:58 chat EBT and before you know you you're
2:28:00 like inspired and you're like oh yeah
2:28:01 that actually is a good idea and uh let
2:28:04 me expand on that let me intend to it oh
2:28:06 you know what I should do I should go to
2:28:08 Mid Journey
2:28:09 if you got 80d it's this is like crack
2:28:14 I do and it is
2:28:17 um
2:28:17 but but even if you don't even if it's
2:28:19 just simple problem solving stuff where
2:28:22 where you know historically you sort of
2:28:24 have to think through stuff and and and
2:28:26 you kind of do
2:28:28 mental stack ranking really like you
2:28:30 know this idea is better than that one
2:28:32 and that's better than that one with
2:28:33 these tools you can kind of go explore
2:28:35 them all
2:28:37 um and it really does sort of get you
2:28:38 started in in a in a very powerful way I
2:28:41 got the same answer for another of your
2:28:43 videos
2:28:45 um as you did so hallucinations
2:28:48 interesting yeah that that first thing
2:28:51 that I did was pretty bad Silver Fox but
2:28:53 the quad thing was pretty good
2:28:54 [Music]
2:28:57 um
2:28:58 what are you teaching here well Richie
2:29:00 Rich
2:29:01 I don't know that I'm teaching anything
2:29:03 here I'm rambling incoherently about
2:29:05 some things that you might learn
2:29:08 something from or not
2:29:11 uh it's mostly generative AI it's
2:29:15 how
2:29:17 we can use it what its implications are
2:29:22 why I think it's interesting
2:29:26 um I was a a founder of one of the early
2:29:29 digital agencies in the mid 90s so I was
2:29:32 there at the beginning of the world wide
2:29:34 web and that whole run up an explosion
2:29:37 and you know
2:29:39 bit of a bust but you know it actually
2:29:41 just kept going up
2:29:42 um so I've seen this before and this is
2:29:45 the parallels to the early days of the
2:29:47 web and this are very very similar
2:29:50 um but this is way more profound and
2:29:53 happening way faster so to put it in
2:29:55 context the
2:29:59 World Wide Web to get to 100 million
2:30:01 users it took about six years
2:30:04 chat GPT it took six weeks
2:30:08 six weeks to 100 million users fastest
2:30:10 adoption of technology in history
2:30:14 um
2:30:15 and what this stuff does compared to the
2:30:17 early days of the web is it is so
2:30:20 profoundly powerful that it it's kind of
2:30:24 eye-wateringly
2:30:28 challenging to get your head around what
2:30:31 what it even makes possible
2:30:35 it makes so much possible
2:30:38 and it and it's just like it it's it's
2:30:40 it looks just like a Google search box
2:30:43 but this ain't
2:30:45 fancy Google search this is way way
2:30:48 deeper than that so that's what we're
2:30:50 talking about that's what I'm talking
2:30:52 about and and I basically try to answer
2:30:54 all the questions
2:30:55 like like who the hell are you what are
2:30:58 you teaching here what are your
2:30:59 qualifications sir do you have papers
2:31:03 do you have papers no I don't
2:31:07 well I do but it's an acting degree
2:31:10 [Laughter]
2:31:13 I do have those papers
2:31:15 plus one wants to know what you think
2:31:17 about Chrome GPT I don't know what
2:31:19 Chrome GPT is but let's go look at it
2:31:23 did oh no
2:31:26 yeah welcome to save the day
2:31:36 GPT
2:31:45 oh it's an extension
2:32:07 um
2:32:10 it looks fine so so it's a it's an
2:32:12 extension for Chrome that basically puts
2:32:15 um chat EPT in the context of it I you
2:32:18 know I switched my default browser to
2:32:20 Bing which which if if you had bet me
2:32:24 serious money a year ago that I would
2:32:27 have Bing as my default or Edge as my
2:32:30 default browser and Bing is my default
2:32:32 search engine
2:32:34 um I I would I would have taken the bet
2:32:36 I would have left you you know out of
2:32:38 the room
2:32:40 um
2:32:42 Bing has
2:32:44 gpt4 incorporated into it and it it or
2:32:48 Edge does
2:32:50 um so so does Bing but Edge has
2:32:54 um
2:32:56 gpt4 integrated into it in a pretty
2:32:59 slick way I assume this extension does
2:33:01 something similar the ability to search
2:33:03 on the internet you know
2:33:05 um do searches that you know that are
2:33:08 context sensitive to the page you're on
2:33:10 things like that
2:33:11 um so I haven't been using any of the
2:33:14 GPT extensions here here's my main
2:33:16 reason why the
2:33:19 the um
2:33:23 when when Google
2:33:26 starts to incorporate Bard into Chrome
2:33:29 and and when Microsoft starts to
2:33:32 incorporate GPT four and five into all
2:33:34 of their product Suite all of those
2:33:36 extensions become kind of irrelevant
2:33:40 um and so I don't I don't want to get
2:33:42 married to a certain way of working or
2:33:44 relying on something that is going to be
2:33:47 in competition with what the core
2:33:49 operating system of the browser or or
2:33:51 you know of the of the apps that I
2:33:54 already use are so I'm just kind of
2:33:57 ignoring them so I don't I don't really
2:33:58 know it but I understand why you'd do it
2:34:00 but I would say download Edge and play
2:34:03 around with Edge and see if you like it
2:34:05 better or worse I'd be curious to hear
2:34:06 what you think
2:34:08 um I sent you an email about a stack I
2:34:10 started made with Claude to do something
2:34:13 like this oh cool that's great thanks
2:34:14 over Fox that's awesome
2:34:16 um can't you just export the transcript
2:34:18 and then give it to chat GPT and have it
2:34:21 analyze it again you you can't no you
2:34:24 can't because chat GPT has a
2:34:28 4 000 token context window so that the
2:34:31 transcript's way too big this is this is
2:34:33 one of the problems
2:34:35 um the the one plug-in that we used
2:34:38 broke the transcript into sections but
2:34:40 it didn't maintain the time code so that
2:34:42 was stupid on its part and then um
2:34:45 clawed because it's got a bigger context
2:34:47 window I could just paste the whole
2:34:48 transcript in there so that works
2:34:51 um and then chat GPT I could put the
2:34:53 transcript into code interpreter and
2:34:56 that would probably work
2:34:58 maybe the problem is with limitations
2:35:00 with the plugins yeah it was that
2:35:03 plug-in sucked actually it was llama oh
2:35:06 interesting
2:35:08 um I was using Claude in to to Aid in
2:35:11 writing it was learning my writing style
2:35:13 but now is saying that it does not have
2:35:15 that capability
2:35:17 did that just change
2:35:22 um
2:35:23 I don't know Andy Lee rocks
2:35:26 um
2:35:29 so
2:35:32 so the memory of a given chat session
2:35:37 does not carry over to another chat
2:35:40 session so only within that chat will it
2:35:42 have the memory of whatever writing it
2:35:44 was learning so it might just be that
2:35:47 that maybe you're in another chat and it
2:35:50 doesn't have the memory of the other
2:35:51 thing that you talked about
2:35:53 so that like that 75 000 context window
2:35:55 is only within a single chat
2:35:59 um
2:36:01 so there's there's there's a thing
2:36:03 called uh what's it called Model creep
2:36:09 um
2:36:10 so model creep is model creep is
2:36:17 um
2:36:18 all of these guys
2:36:22 wait not that
2:36:24 all of these guys
2:36:27 are
2:36:28 tweaking and
2:36:31 um
2:36:33 refining these models as we're working
2:36:36 with them so very often what will happen
2:36:39 is you'll have a prompt that works and
2:36:41 then three weeks later it just doesn't
2:36:43 seem to work anymore
2:36:45 that's model creep and it sucks and and
2:36:48 it's just again it's part of the nature
2:36:50 so it's possible that you hit that it's
2:36:53 also possible that that
2:36:55 um
2:36:56 you're assuming that it's carrying over
2:36:58 its memory from previous conversations
2:37:00 and it doesn't do that it will in the
2:37:02 future or you know I don't know if
2:37:04 Claude will but you know some of
2:37:06 somewhere in these someone's going to
2:37:08 start adding memory to these where it
2:37:11 will actually remember the context of
2:37:13 all of our different conversations and
2:37:16 all of our documents and all that sort
2:37:18 of [ __ ] it's going to get really good
2:37:19 right now they're just kind of it just
2:37:21 doesn't do that
2:37:24 um for anyone new Kyle goes back through
2:37:26 the comments and answers all the
2:37:27 questions thank you Joker
2:37:30 holy Cali figured it out oh that's
2:37:32 awesome
2:37:33 I just figured it out what'd you figure
2:37:35 out Josh wait Kyle it left out the part
2:37:37 about how you're completely unqualified
2:37:39 exactly
2:37:40 [Laughter]
2:37:42 Claude exactly Claude for the win that
2:37:45 was a pretty cool one
2:37:47 um can you have it summarize the steps
2:37:49 and share it so I can screenshot please
2:37:52 the chapter project can you have it
2:37:54 summarize the steps
2:37:57 summarize what steps over Fox
2:37:59 um
2:38:01 I can go find where's my clog
2:38:06 where can I paste these I can't I can't
2:38:08 paste these anywhere can I
2:38:11 uh
2:38:23 I don't quite know what you're asking
2:38:25 but I've got that prompt saved so
2:38:28 um
2:38:29 so I can get I can get it to you somehow
2:38:31 like the The Prompt right now is is not
2:38:34 all that refined
2:38:36 what did I put
2:38:38 um
2:38:40 it said you didn't write the chapters in
2:38:42 The Format you described above I need
2:38:44 the chapters
2:38:45 to be clickable when they're pasted into
2:38:48 a YouTube description area also
2:38:51 can you add five additional chapters
2:38:54 um
2:38:59 yeah I don't know what I did but uh I'll
2:39:02 I'll get you what I can Silver Fox just
2:39:04 DM me and I can with what your specific
2:39:07 request is and I can get you that
2:39:09 um
2:39:10 okay let's see I never realized you said
2:39:14 um so much until you sped it up I know
2:39:16 me too it's like I'm I'm I'm I'm I'm I'm
2:39:18 I'm I'm I'm I'm I'm I'm I'm I'm I'm I'm
2:39:21 um
2:39:22 foreign what are you gonna do I'm just
2:39:26 making this [ __ ] up on the fly so that's
2:39:28 why you get the ums well I know what I'm
2:39:30 talking about I don't um but I I'm a lot
2:39:38 what is the best
2:39:40 CD DVD for a Dell 7940 laptop that's
2:39:46 pretty funny I don't have a [ __ ] clue
2:39:48 I assume it's a joke
2:39:51 um and what other
2:39:52 whatever other happy horse
2:39:55 thank you quoting my my stupid sayings
2:39:59 uh but about I did SEO back in the day I
2:40:03 think seo seo is going to be a really
2:40:05 interest it's got a really interesting
2:40:07 future ahead of it
2:40:09 I don't know that it's a good one
2:40:12 I don't know I I think there might be
2:40:14 llmo like how do you get your [ __ ] in
2:40:17 large language models and I think how
2:40:19 you get it in large language models is
2:40:21 you have good content
2:40:23 all this SEO and injection [ __ ]
2:40:26 might go the way of the dodo bird it's
2:40:29 going to be interesting to see how that
2:40:31 all plays out what do you think of
2:40:33 Google bard I I think it's
2:40:37 um
2:40:38 it's at least not embarrassingly bad now
2:40:41 what's interesting about Bard what
2:40:43 they're doing really well they're
2:40:45 incorporating hyperlinks in the answers
2:40:47 at appropriate times with appropriate
2:40:51 context and and in a pretty
2:40:53 sophisticated way so I like how it
2:40:56 incorporates the internet into its
2:40:57 answers
2:40:59 um
2:41:00 I hear it's better at coding than it is
2:41:03 at language and it's it's okay at
2:41:05 language right now it's not great but
2:41:06 it's okay
2:41:08 um and and apparently Gemini which is
2:41:11 their next model
2:41:12 is going to be a gpt4 competitor
2:41:16 um considering where bardi is right now
2:41:19 I think Gemini might be actually really
2:41:21 good so
2:41:23 I it's it's on the list like it wasn't
2:41:26 on this list for three months two months
2:41:29 three months like it was laughably bad
2:41:33 it's not anymore
2:41:35 I wrote AI
2:41:36 code in the military
2:41:39 in the Mi military
2:41:44 a 2023 Shelby GT500 KR arrest me red oh
2:41:50 Todd Hendricks he wants more pictures
2:41:54 I don't know that I don't know that it
2:41:56 would be good at being that specific
2:41:58 although let's let's let's try that
2:42:02 let's do this
2:42:06 I'm gonna be slash imagine
2:42:11 paste that bad boy in there of a red
2:42:17 2023
2:42:22 Shelby
2:42:26 GT 500 KR
2:42:33 close up front Corner view doors closed
2:42:37 St Louis
2:42:39 arch in the background
2:42:42 all right I'll make that one let's see
2:42:44 if that turns into anything interesting
2:42:47 all right I wrote AI code in the mid oh
2:42:49 in the in the in the mid 1990s yeah
2:42:52 listen I I mean ai's ai's been around
2:42:56 for decades so so the the reason that
2:42:58 that I think this is such a similar time
2:43:01 to to the early days of the World Wide
2:43:03 Web
2:43:04 the internet had been around for decades
2:43:07 and it was primarily used by scientists
2:43:09 and researchers and Geeks and then the
2:43:12 World Wide Web came along and it made it
2:43:14 simple enough that just regular human
2:43:16 beings could use it
2:43:18 same thing with machine learning and you
2:43:21 know neural networks and and you know
2:43:23 expert systems and all of the stuff that
2:43:25 has been the foundation that has gotten
2:43:27 us here and Transformers in 2017
2:43:30 and what changed was
2:43:33 Chet GPT
2:43:36 comes along
2:43:37 Geeks look at chat gbt and they're like
2:43:40 well you could just do that in the
2:43:41 playground yeah but the difference
2:43:43 between chat GPT and the playground is
2:43:45 you don't have any of the settings in
2:43:47 chat GPT it's simple
2:43:50 like simple enough that anyone who could
2:43:52 use a Google search bar could use it
2:43:56 and that's why it got a hundred million
2:43:59 users using it in six weeks because it
2:44:02 was like a Google search bar but then
2:44:04 when you type [ __ ] into it what came
2:44:06 back was [ __ ] miraculous
2:44:09 it wasn't a list of SEO injected links
2:44:12 eighty percent of which are irrelevant
2:44:14 to you
2:44:16 right and so that's the that's the
2:44:18 difference so so you have the foundation
2:44:21 of all of that stuff
2:44:23 this gave really good images but I
2:44:25 explicitly asked for the St Louis Arch
2:44:28 and we got nothing
2:44:29 we got no St Louis Arch
2:44:33 how does it miss that
2:44:35 but these Mustangs look pretty you got
2:44:37 the lights right it got the hood shape
2:44:38 right it got the cobra
2:44:42 they're not bad
2:44:45 but where's the St Louis Arch come on
2:44:48 people
2:44:49 what are we doing
2:44:52 what are we doing uh are we down
2:44:56 is that Lexica no what was I using
2:45:00 I don't know what I was using this is
2:45:02 this is mid-journey this is uh
2:45:05 mid-journey I don't know what I was
2:45:07 using before that might have looked like
2:45:08 lexico
2:45:13 I see it now mid-journing okay
2:45:15 um
2:45:16 which website is this that was
2:45:17 mid-journey like a Shelby nice
2:45:22 I saw a tiny finger
2:45:24 pushing that Matchbox car exactly that
2:45:27 was pretty funny
2:45:30 um long form to short form repurpose.io
2:45:32 or get much not perfect but Time Savers
2:45:36 yeah by the way uh hair growth geek
2:45:39 um I signed up for repurposed
2:45:42 dot IO today for an account it doesn't
2:45:45 do lives it'll it'll it'll automatically
2:45:49 download and re repurpose your channel
2:45:53 but it won't do lives so I gotta find a
2:45:57 tool that'll do the lives I'm just doing
2:45:59 them manually now because I think I was
2:46:01 starting to lose some of the early ones
2:46:03 I think I lost three or four of them
2:46:05 uh we've paid far more than we did for
2:46:08 Taylor Swift tickets
2:46:09 yeah in your sanity
2:46:13 um I love her for it glitching uh she
2:46:16 also gave a lot of money to the food
2:46:17 banks
2:46:23 um in the cities where she's gone on
2:46:25 tour oh that's cool yeah she's I mean
2:46:27 listen she she
2:46:32 I mean she was clearly already a
2:46:33 powerful woman but like in the last two
2:46:36 or three years she stepped into her
2:46:39 power
2:46:40 like she's like a [ __ ] superhero
2:46:43 right now she's just [ __ ] in this
2:46:46 [ __ ]
2:46:47 Zone it's it's very impressive she's
2:46:51 very impressive
2:46:53 ugh I thought I might I missed the
2:46:55 prompts info oh I missed the prompts
2:46:57 info I don't know what prompts you were
2:46:59 talking about the prompting I did with
2:47:01 I don't know what it was
2:47:03 but I now have a a YouTube channel so
2:47:07 you can see I now have a YouTube channel
2:47:10 um it's got some some of the early lives
2:47:13 on it the AI learning lab uh AI learning
2:47:17 lab Dash TT
2:47:19 on YouTube
2:47:21 so that's the new YouTube channel so
2:47:23 this stuff will all be up there so
2:47:25 and and we figured out we can use Claude
2:47:28 to actually generate chapters for for
2:47:31 all the videos I gotta figure out a way
2:47:33 to automate that [ __ ] I wonder if Claude
2:47:35 does anyone know if claude's in zapier
2:47:39 am I late
2:47:41 was it was an hour ago Mr Shannon oh man
2:47:45 yeah well I know
2:47:47 quality of experience should be the
2:47:49 basis for AI not Comfort or happiness
2:47:52 that's interesting the quality of
2:47:54 experience should be the basis for AI
2:47:58 not Comfort or happiness
2:48:00 I think that's right
2:48:02 I mean I I listen I think ostensibly
2:48:05 what these tools do
2:48:07 is they do stuff
2:48:15 at least faster
2:48:17 and and in a lot of cases better than we
2:48:21 do and and if if you think about how
2:48:23 they're going to improve in the future
2:48:26 they're going to start to get way better
2:48:28 at us at almost everything and so I
2:48:30 think you're right I think
2:48:32 I think how you use these things the
2:48:34 quality of
2:48:36 what they produce and how we engage with
2:48:39 them
2:48:40 is going to be the thing
2:48:42 um
2:48:44 and it's going to cause a lot of pain
2:48:46 and disruption it's gonna suck
2:48:48 but
2:48:49 on the other side of it
2:48:51 is going to be some some pretty amazing
2:48:53 stuff I think
2:48:55 I hope
2:48:57 oh it just jumped me to the bottom again
2:48:59 okay I'm gonna do I'm gonna answer a
2:49:01 couple more questions I'm just gonna
2:49:02 slide up from the bottom a little bit
2:49:04 here and answer a few questions and then
2:49:06 I'm gonna go because I'm tired I'm
2:49:08 cranky my ass is numb and I'm almost out
2:49:10 of water
2:49:12 mm-hmm
2:49:17 can a complex company Implement code
2:49:19 interpreter in an enclosed space does
2:49:21 openai need to be involved
2:49:24 I don't know so I don't think code
2:49:27 interpreters available via the API yet
2:49:32 um I don't know enough about Azure so
2:49:35 Microsoft Azure has the you can buy I
2:49:38 think the cheapest ones like 25 Grand so
2:49:41 I think you can create these sort of
2:49:42 stand-alone azure
2:49:45 instances
2:49:47 where I think you're running chat GPT
2:49:50 within that instance
2:49:52 where you're not actually making a call
2:49:54 out to open ai's instance their servers
2:49:59 um I think that's how that works but
2:50:03 you would obviously need to do that you
2:50:05 know you would need to be working in a
2:50:06 company that's got budget to do that so
2:50:08 I would look up
2:50:11 um what what Microsoft is doing with
2:50:13 Azure with with the Standalone
2:50:16 um
2:50:18 server offerings with open Ai and my my
2:50:22 strong suspicion is that at a minimum
2:50:26 you could run a gpt4 instance where
2:50:30 you're not making a call to open AI
2:50:32 I don't know about code interpreter
2:50:34 within that
2:50:38 um now the interesting thing about code
2:50:41 interpreter as I understand it
2:50:45 is when you fire up a code interpreter
2:50:48 instance it's spinning up a virtual
2:50:50 machine
2:50:51 essentially within your chat right or
2:50:54 within whatever the you know whatever
2:50:56 the [ __ ] going on in the back end but
2:50:58 it's spinning up a virtual machine and
2:51:00 all of that python code that it's
2:51:02 writing and the things that it's
2:51:03 executing are are really only active
2:51:06 when that virtual machine is active if
2:51:09 you leave a code interpreter session for
2:51:11 you know like a half an hour and come
2:51:12 back to it you'll often lose all of the
2:51:15 charts and the calculations that it did
2:51:17 because the machine is shut down so a
2:51:21 power move is if you're using Code
2:51:23 interpreter copy and paste your results
2:51:25 out to other other documents because you
2:51:27 might lose them
2:51:28 but but my instinct is is that anything
2:51:32 that you do with encode interpreter is
2:51:34 not being
2:51:35 you know stored and sent to to open AI
2:51:39 but like probably in a cached kind of
2:51:41 way but I doubt it's being stored there
2:51:44 but
2:51:45 that's total speculation so if
2:51:49 if if you if you've got private
2:51:51 information you can't have get out then
2:51:53 you need you need answers like
2:51:55 from Azure not from someone like me
2:51:58 I'm a crumb oh Chrome GPT love I thought
2:52:02 you said crumb GPT that's funny there's
2:52:05 a symbol I was talking about uh it's so
2:52:08 revolutionary it made Microsoft browsers
2:52:09 and search engines relevant again I know
2:52:12 totally I avoid Edge like to play now I
2:52:14 must reconsider well here's the deal
2:52:16 Rebel so
2:52:18 Edge is not bad here's the problem with
2:52:20 Edge it it leaks memory like a a [ __ ]
2:52:23 like minnow net it's it's horrible like
2:52:26 like
2:52:28 I I I am a I'm a tab hoarder so I
2:52:31 usually have between 75 and 150 tabs
2:52:34 open at any given time and I even have
2:52:36 tabs shut down [ __ ] running and just
2:52:39 every day and a half
2:52:42 um Edge just shits the bed and I I have
2:52:44 to come it'll like completely lock up my
2:52:46 machine so it's it's just a dog that way
2:52:49 but how they have chat EBT incorporated
2:52:52 into it's pretty slick so it's actually
2:52:54 worth it for for me
2:52:57 um but if I need [ __ ] stable I go back
2:53:00 to Chrome and just do do stuff over
2:53:01 there but about 90 of what I'm doing I
2:53:03 do an edge
2:53:05 um just just because of convenience I I
2:53:08 also I also wanna
2:53:10 I've been trying to pay attention to how
2:53:13 the implementation in Edge evolves
2:53:17 like I'll notice that some days I've got
2:53:19 a 4 000 token context window and some
2:53:22 days it's two thousand and some days I
2:53:24 can upload images and some days I can't
2:53:25 and some days the little sidebar is
2:53:29 aware of the web page to the left of it
2:53:31 and sometimes it's not like it's it it's
2:53:34 weird it's like they they're clearly
2:53:36 doing hot fixes on all this stuff right
2:53:39 as is open AI as are all these folks
2:53:43 right they're all making hot fixes on
2:53:45 the fly so what worked today won't work
2:53:47 tomorrow and then all of a sudden just
2:53:49 some [ __ ] miraculous feature will
2:53:51 appear and you'll be like this is
2:53:52 awesome and then it'll disappear that's
2:53:54 just the nature where it is right now
2:53:57 all right I'm gonna answer one more
2:53:58 question and then I'm Gonna Leave You
2:54:00 All for a lovely Saturday probably
2:54:02 morning for some of you sorry about that
2:54:05 I love Edge usability is great and low
2:54:08 on resources too until it leaks memory
2:54:10 and then it blows my [ __ ] up I've been
2:54:12 on it for a year I've been on it for
2:54:14 nine months now
2:54:16 um I've grown accustomed to Edge can
2:54:18 someone Define a token yes okay so
2:54:21 that's a good question Lewis Louise
2:54:23 Scott Louis Scott
2:54:26 so so when they say it's got a 4 000
2:54:28 token context window or a
2:54:31 8 000 token context window
2:54:35 um it counts so when you have a
2:54:37 paragraph you've got punctuation and
2:54:40 spaces and words
2:54:42 so tokens are basically the little
2:54:45 elements that make it up so so
2:54:46 punctuation is a token I think spaces
2:54:49 are tokens but they might not be it
2:54:51 might just be punctuation and words and
2:54:53 I think long words might even get broken
2:54:55 up into into sub words so so they're
2:54:58 basically the the elements of data
2:55:01 that
2:55:03 companies like open AI can count right
2:55:07 so that's how they bill so
2:55:10 for every thousand tokens you get billed
2:55:14 three cents or two cents or six cents
2:55:16 depending on the model you're using
2:55:17 which is basically just to process this
2:55:21 amount of combination of letters and
2:55:23 punctuation
2:55:25 um it costs them a penny and they're
2:55:28 charging us three pennies or two pennies
2:55:30 or whatever it is
2:55:32 um what it translates into though is
2:55:35 about three quarters of
2:55:39 the total amount of tokens is words on
2:55:43 average so if you have 4 000 tokens that
2:55:46 means three thousand words
2:55:48 or in claude's case it can do a hundred
2:55:51 thousand tokens
2:55:53 which means about 75 000 words it might
2:55:56 be more or less than that depending on
2:55:57 the kind of data you put in there but
2:55:59 that's it's kind of a rule of thumb
2:56:02 um
2:56:03 so it's and then the context window is
2:56:06 basically like in a given chat how much
2:56:09 sort of can it remember or how much can
2:56:12 you input
2:56:13 and and
2:56:15 that's the limit so so sometimes if you
2:56:17 take a long document and you paste it in
2:56:19 the chat GPT it's just like I can't do
2:56:21 it it's too big
2:56:22 with Claude you can paste it in and it
2:56:24 it does it and it's you know the way it
2:56:27 works is it's a combination of your
2:56:29 prompt plus the response it's that whole
2:56:32 thing has got to be under 4 000 tokens
2:56:35 or 8 000 tokens or in the case of Claude
2:56:37 a hundred thousand tokens
2:56:39 so that's that's what that that means
2:56:42 it's basic it's it's basically
2:56:44 the components of a
2:56:46 set of words
2:56:48 that get processed
2:56:51 if I wanted to make a video game
2:56:52 developer assistant that knows free
2:56:54 software where should I start hmm
2:57:01 well you would want to start with some
2:57:03 of these tools that are connected to the
2:57:05 internet so you might want to start with
2:57:07 Bard or Bing
2:57:12 but
2:57:19 I I
2:57:21 if if you want to start
2:57:24 building something that's sophisticated
2:57:26 probably something like Lang chain where
2:57:28 you're going to start embedding you're
2:57:30 going to take data sources
2:57:32 um and embed them and get them in a
2:57:35 vector database so you can GPT them
2:57:38 um so I would start exploring the Lang
2:57:40 chain project
2:57:42 um and then and then I would think using
2:57:44 the Llama 2 models um within Lang chain
2:57:47 just because they're new and
2:57:48 commercially viable that's where I'd
2:57:50 start all right I'm gonna do one more
2:57:52 question and I'm gonna go Gotta Go
2:57:54 Silver Fox see you later
2:57:56 uh
2:57:57 um
2:57:58 I want to create the same thing
2:58:02 can we use AI to help correct students
2:58:04 essays and compositions yes and they're
2:58:07 probably using AI to help write their
2:58:10 essays and compositions so yes you can I
2:58:13 sent you a short AI created on Tick Tock
2:58:15 yesterday well yesterday okay cool all
2:58:17 right I'll check it out all right
2:58:18 everybody
2:58:20 um
2:58:21 I gotta take off thanks for hanging out
2:58:24 with me on a Saturday night my eyes are
2:58:26 bloodshot my voice doesn't work anymore
2:58:28 my ass is numb that means it's been a
2:58:30 good night I hope you enjoyed this it's
2:58:32 been great hanging out with you peace
2:58:34 out everybody I'll see you tomorrow