aiLL 7/26/23 - The Transformative Power of AI

Description

Chapters

Transcript

0:01 you want to come inside come on Champion
0:04 come inside
0:05 no come on come inside
0:13 lay down
0:15 lay down you big dum dum
0:20 all right welcome welcome welcome
0:42 yo felonious you there
0:46 how you doing it's happening
0:50 ah wait hang on
0:53 how do I get
0:56 yeah it was shaking
0:59 hello so I got a an appointment for
1:02 Friday they're there Thursday and Friday
1:04 the last thing I sent you was a link
1:08 I think I can I still have it on my
1:10 clipboard
1:12 oh cool
1:14 for the uh for the appointment
1:16 yeah you can find out where they are
1:18 they're in different cities
1:24 I can look I can't look while I'm on
1:27 here I'll look when I go down and I'll
1:30 let you know yeah just let me know
1:32 because I'll be in here so I'll be able
1:33 to do it but if they come to Denver
1:35 that'd be nice
1:39 cool so what do you do
1:43 uh I am the director at the Salvation
1:47 Army of a family homeless shelter and
1:52 a drug rehab two different agencies on
1:55 the same property I get my steps in
1:57 every day
1:58 [Laughter]
2:00 uh so 30 families in the shelter
2:04 and a hundred well it's 40 now uh
2:09 addicts recovering from substance use
2:13 it's good work I used to work I used to
2:16 work in technology sales and a couple
2:19 years back decided to do something a
2:21 little more
2:23 uh fulfilling
2:25 it's nice also going to also in grad
2:29 school full-time for a therapist license
2:32 oh wow oh that's super cool yeah there's
2:35 a guy in the AI Salon here in Denver
2:37 that works for a non-profit as well and
2:40 is doing some cool stuff with AI so
2:42 that's pretty cool yeah it's really
2:44 helping in grad school it's helping with
2:47 licensing issues at work
2:50 um
2:51 and uh
2:53 hopefully I can show my mom how to use
2:56 it
2:56 yeah it's awesome man I don't know how
2:59 she's going to use it but
3:01 um I really appreciate your live I try
3:03 to make it every day
3:05 yeah I appreciate it man and I scuba
3:07 dive
3:08 oh nice nice I guess that's what the
3:11 picture is is that underwater somewhere
3:13 nice that's in the Maldives in November
3:15 on my site on my profile there's a video
3:18 I took in November of a whale shark the
3:22 one thing I wanted to always see
3:26 those are [ __ ] huge brilliant
3:29 brilliant so doctoral graceful and uh I
3:34 was in the water with it in the middle
3:35 of the night
3:37 that's so cool thanks for what you do
3:40 hey yeah I appreciate it good talking to
3:43 you
3:44 likewise all right cheers
3:50 all right welcome everybody welcome
3:53 welcome welcome this AI learning lab
3:55 great chatting with you there Thelonious
3:56 uh tell me tell me if they uh if they
4:00 zap your uh retina with the with the
4:02 laser when you get armed you can be the
4:05 guinea pig yeah apparently the uh the
4:08 world coin the line for getting the
4:09 world coin the world ID I think they
4:12 call it in New York City the lines are
4:15 incredibly long so if anyone's freaked
4:18 out
4:20 um
4:21 they're not in New York
4:23 at least at least the crypto Bros are
4:26 lining up for it
4:30 this day I learning lab uh I'll talk
4:32 about these URLs here in a bit
4:35 um my name is Kyle Shannon I'm a long
4:38 time entrepreneur started a bunch of
4:40 companies
4:41 um I was an entrepreneur in the early
4:42 days of the web
4:44 and what's going on with generative AI
4:46 right now is incredible feels incredibly
4:48 familiar to me it feels like ah we're
4:50 doing one of these again
4:52 um although it feels like this
4:54 generative AI stuff is coming a lot fast
4:56 a lot faster and a lot more profound
4:58 uh implications I think I think and uh
5:03 if you have uh questions comments about
5:05 AI pop them down below and then let me
5:07 uh flip the camera here
5:12 [Music]
5:12 um
5:15 all right so
5:19 so if you're new to generative AI if you
5:21 have not played with chat gbt that's a
5:24 large part of what this channel is about
5:25 is I just
5:27 curious
5:29 whether you want it or not
5:32 to give you some some context here
5:34 there's there's seven sites here that
5:36 are all different flavors of Chachi PT
5:40 um you know competitors you know
5:42 different sites where you can use chat
5:43 GPT
5:44 um this used to be three three sites
5:46 long and as of a week ago it was seven
5:49 sites long and
5:52 um
5:52 Elon musk's company's not on there but
5:55 that'll be there at some point Apple's
5:56 not on there that'll be there at some
5:58 point there's probably two or three
5:59 others that are not on there that should
6:01 be on there
6:03 um
6:05 chat.openai.com that's the official chat
6:06 gbt website if you haven't played with
6:09 it go there and start playing with it
6:10 bing is also chat GPT that's Microsoft's
6:14 uh search engine that has incorporated
6:17 gpt4 as their base chat model and it's
6:20 connected to the internet and you can do
6:22 image analysis and you can
6:24 make images with it then you've got
6:27 Claude AI which is uh a competitor to
6:31 chat GPT that's actually really good
6:33 it's got a hundred thousand token
6:35 context window you've got
6:37 bard.google.com that was laughably bad
6:40 just a week ago and then they swapped
6:42 out the model with this new multi-modal
6:46 beast called Gemini that seems like it's
6:49 actually pretty good
6:51 um and then
6:52 labs.perplexity.ai that's metas you can
6:55 go play with the meta llama 2 models
6:57 there and then Pi is super
6:59 conversational and Poe gives you lots of
7:02 these things you can play with all in
7:03 the same interface so that's all of
7:06 those things then
7:08 oh we don't stop there we we give value
7:12 all over
7:13 [Laughter]
7:16 um
7:16 so so the top two sites there are
7:20 getting started resources so
7:22 prompts.chat will teach you how to talk
7:23 with those large language models
7:26 openart.ai promptbook will teach you
7:29 about prompting image generation things
7:31 it's written for stable diffusion but
7:34 it's such good information you can
7:35 really it really applies to all of these
7:37 things
7:38 futurepedia.io is a big directory of AI
7:42 sites then you got mid Journey Adobe
7:45 Firefly and Leonardo slash stable
7:47 diffusion uh which are all uh
7:51 really good uh image generation tools
7:53 and then you got some things down below
7:55 all right so pop any questions you have
7:57 down in the comments below uh do me a
8:00 favor share the uh share the live if you
8:02 would if you're willing to
8:03 let people know you're a weirdo and you
8:05 like this AI [ __ ] and we'll we'll get
8:09 some more people in here
8:11 and we're just getting started what's
8:13 tonight Wednesday night middle of the
8:15 week middle of the week midweek midweek
8:19 um the other thing if you're interested
8:21 in joining
8:22 um a group of like-minded AI adventurers
8:26 um the AI Salon is something I started
8:28 back in December we meet bi-weekly we
8:31 meet in person here in Denver but then
8:34 have an online
8:35 uh component and our next meeting is is
8:39 next Tuesday
8:40 and we have a Discord server where you
8:42 can play with mid-journey and ask
8:43 questions and things like that so feel
8:45 free to check that out and
8:46 storybinding.com that's my company I'm
8:48 the founder and CEO of that so with that
8:51 let's get rolling let's get rolling
8:54 rolling rolling
8:56 rolling rolling rolling
9:01 all right got on my Fender t-shirt
9:05 any guitar players in the house
9:07 I had electric guitars for a while and
9:09 then I realized that the electric guitar
9:11 is actually a different instrument than
9:12 the acoustic guitar
9:14 and
9:16 I just wasn't as good at it but I just
9:18 didn't get it I didn't get it because I
9:20 guess I learned on an acoustic and just
9:21 never made the switch
9:23 all right let's get to some questions
9:26 questions
9:30 all right Miami New York is like not in
9:33 Denver now checking yeah I'll do that
9:35 we'll get there well we probably won't
9:39 so so do we want to verify any humans in
9:43 that Cow Town the base of the mountain
9:45 nah we're good
9:47 [Laughter]
9:50 uh yep the the world app has got all
9:53 that stuff I haven't tried Claude yet
9:54 but I heard it's great yeah Heather the
9:56 the Claude what's what's really nice
9:59 about Claude is
10:01 because it's got a hundred thousand
10:03 token context window you can either put
10:05 in really large documents or you can put
10:06 in multiple documents or you can just
10:09 have long conversations with it and it
10:11 doesn't lose its memory you know how you
10:12 know how in chat GPT you're talking and
10:15 all of a sudden it's like do you not
10:17 remember the beginning of this
10:18 conversation and it's like nope
10:21 I don't I'm I'm a little dumb dumb after
10:24 about three thousand words
10:26 so claude's much better for that
10:29 um a couple of places where it's it's
10:31 not
10:32 you know it's not up to Snuff with the
10:34 writing of gpt4 but for the most part
10:36 it's pretty good
10:38 uh what is image analysis how might we
10:41 use it oh let's go play with that I'll
10:43 go I'll show you that
10:44 it's not well it
10:47 well there's a difference between image
10:49 we can do image analysis using Code
10:51 interpreter
10:53 um but both Bard and Bing are multimodal
10:58 meaning they can read images
11:00 so you could upload an image and have it
11:03 OCR it you could upload an image and
11:05 have it tell you what's in it or write
11:07 an ad for that image so let's go let's
11:10 go try that
11:13 oh this is the new oh by the way there's
11:15 a new
11:16 um
11:17 stability AI released sdxl 1.0 so
11:22 they're they're new image model
11:27 [Music]
11:27 um
11:31 so that is an image generated
11:36 this model is pretty good this is my my
11:39 uh 70s muscle car in an abandoned
11:41 Factory
11:42 that's indeed a 70s muscle car
11:45 these things are always like these Weird
11:46 Hybrids between like all the different
11:48 manufacturers it's funny
11:51 that one's sort of a Camaro front end
11:53 with kind of a javelin in the middle and
11:55 then
11:57 sort of Monte Carlo lines it's weird
12:00 anyway at least it's 70s looks pretty
12:03 good
12:05 um you can go play with this at
12:07 clipdrop.co
12:09 there's also another site I forget what
12:11 it was you can go play with it at
12:14 but anyway let's actually let's grab one
12:16 of these
12:18 we'll grab one that uh
12:22 that one looks like a mess let's grab
12:24 this one
12:25 all right so we're gonna save this
12:30 as muscle
12:35 and then we're gonna hop over to Bing
12:39 and so so when you're Bing if you go to
12:41 bing.com it's a search engine and if you
12:44 click on the chat button chat button it
12:47 brings you to this which is this is
12:49 basically chat gbt but the big flavor of
12:51 it
12:52 um one nice thing if you haven't played
12:54 with Bing is you've got more creative
12:55 more balanced and more precise so this
12:58 is if you've ever played on the open AI
13:00 playground this is the equivalent of
13:02 temperature right so more creative would
13:04 be higher temperature more precisely
13:06 lower temperature I have no idea what
13:08 what actual settings are sitting beneath
13:11 those but we're not supposed to know
13:13 that because we're just lowly end users
13:17 uh which is probably a good thing but
13:21 down at the bottom here you know I could
13:23 I one thing you can do on Bing if you
13:25 didn't know you can make images so I can
13:27 do
13:29 um let's see
13:31 um
13:32 chubby
13:33 black and white
13:35 mutt
13:38 uh
13:41 dancing
13:45 we'll say cartoon
13:48 of chubby in uh let's see Ian Pixar
13:53 Style
13:56 all right and so it's gonna go off and
13:59 make us an image although because I said
14:00 Pixar style it might go we're not
14:02 allowed to do that oh no it looks like
14:04 it's gonna do it
14:06 one of the things about Bing's chat GPT
14:09 is the safety guard rails on it can be
14:12 really [ __ ] annoying there's
14:13 sometimes just a little too
14:16 too big but they're just generated that
14:18 image but now check this out so I can
14:20 say add an image
14:22 I can upload from the device and let's
14:24 go we'll go grab that green muscle car
14:27 image I just did this one
14:30 we're gonna pop that in there
14:33 and now we're going to say
14:35 please describe
14:39 the scene
15:03 uh let's see let me let me I'm going to
15:06 do a new topic it got confused because I
15:08 asked it to make an image and then I
15:10 asked it to describe the image but it
15:13 didn't know that I was asking it to
15:14 describe the uploaded image okay so I'll
15:16 say please uh
15:19 describe
15:22 what you see here
15:28 [Music]
15:39 images
15:41 vintage card an abandoned warehouse the
15:44 car appears to be a 70s era muscle car
15:46 with a black stripe on the hood and a
15:47 black vinyl roof the warehouse large
15:50 Windows broken panes peeling paint on
15:52 the walls the overall mood is desolate
15:54 and abandoned is there anything specific
15:55 you'd like to know about the image
15:57 please
15:59 right
16:04 me
16:05 five different
16:08 advertising
16:12 copy lines for this image
16:16 for a
16:20 Car Wash and detailing service
16:24 period
16:26 the ads
16:27 should be
16:29 clever humorous and memorable
16:36 I think oh did I not have this in
16:38 creative mode
16:40 I think I'm in
16:42 balanced mode but whatever we're not in
16:44 creative mode actually it'll be
16:45 interesting maybe we'll flip over to
16:47 creative mode and see if it's any better
16:49 don't let your car end up like this one
16:51 keep it clean with our Car Wash and
16:53 detailing services even this vintage
16:55 beauty deserves a good car wash
16:58 abandoned warehouse or not every car
17:01 deserves a good deserves to shine let us
17:04 help you with that
17:05 don't let the dirt and grime take over
17:08 so let's let's go new topic
17:10 yeah that was more precise so let's do
17:12 it we'll do it more creative now
17:15 so we'll pop that image back in there
17:21 [Music]
17:24 right me
17:28 a set of five
17:31 add copy lines
17:34 for this image
17:36 comma
17:37 that reference the elements in the image
17:42 but after a Car Wash and detailing
17:45 service period
17:47 make the ad clever funny and memorable
17:53 all right that's somewhere in the
17:55 neighbor of what we said before let's
17:57 see what it does in more creative mode
18:00 Let's see if Bing
18:02 disappoints us and writes like crappy
18:04 Microsoft copy or if it writes something
18:06 actually good
18:09 here are some possible headlines don't
18:10 let your car end up like this
18:12 don't let your card up like this
18:14 Warehouse bring it to a thorough service
18:16 your car deserves better than this
18:18 that's not bad this is what happens when
18:21 you neglect your car okay I'm gonna say
18:25 um
18:25 these are all too long make them three
18:29 to seven word ads
18:31 that are incredibly Punchy and memorable
18:40 watch detail repeat don't be this guy
18:43 that's pretty good save your car call US
18:45 Car Care We Care Warehouse bad wash good
18:48 that's good Warehouse bad watch good and
18:52 that one Warehouse badwash good I would
18:54 actually that's okay all right Microsoft
18:57 doesn't suck so you could do that with
18:59 images you could upload a PDF and have
19:01 an OCR it or or tell you what's in it
19:04 potentially even pull data out of a
19:06 chart or a graph potentially I don't
19:08 know
19:09 um
19:10 but yeah it can see images and and
19:13 Google bard does the same thing let's go
19:15 look at Bard let's do the same thing
19:16 with bard
19:17 so we're gonna go grab the old muscle
19:19 car there it is
19:21 I'm gonna pop that up in Jolly Old Bard
19:24 and let me go just I'll just go copy my
19:29 uh uh
19:31 um
19:32 [Music]
19:34 copy so the exact same
19:39 prompt
19:41 this is now in barred so the first set
19:44 that it did was those
19:47 scalebard does
19:49 I was pleasantly surprised with
19:50 something I did with Bard the other
19:51 night on a live
19:54 here are five ad copies our car wash is
19:56 so good it'll make your car look like it
19:58 just drove out in a of an abandoned
20:00 building we'll take your car from
20:01 looking like a relic
20:04 to looking like a 69. that's actually
20:07 really interesting
20:09 it kind of it kind of got the era of the
20:12 car
20:13 there's kind of the cultural reference
20:15 there
20:16 huh that's interesting our detailing
20:19 service is so thorough we'll even find
20:21 the keys that were lost in the trunk 10
20:23 years ago that's not bad
20:25 don't let your car become a permanent
20:27 resident of abandoned building our car
20:30 wash is so good it'll make you want to
20:31 start a band name and name it the
20:33 abandoned muscle cars that's actually
20:35 good let's and let's do the same thing
20:37 we did with with Bing where we told it
20:39 those are too long to give us short ones
20:44 same thing
20:50 [Music]
20:53 abandoned car wash it out dirt Will
20:55 Blast it off make your car look like new
20:58 again we'll find your keys your car
21:01 deserves a watch I you know that's not
21:03 fat your car deserves a wash
21:07 it's just sitting in a [ __ ] Factory
21:10 deserves to watch it's a little bit of
21:13 Shame that's that's that's good I like
21:15 it it's shame marketing
21:16 [Laughter]
21:20 ah welcome everybody this is the AI
21:23 learning lab we are learning about in
21:25 that case we were learning about both
21:27 Bing and Bard can read read images
21:32 um and see what's in them and that's
21:33 there's some interesting stuff going on
21:36 there all right let's see you probably
21:39 get asked this all the time but what's
21:41 the best one for code generation I don't
21:44 I'm not a CO I'm not an engineer so
21:47 um
21:49 I would say
21:51 probably the primary place to start is
21:54 GitHub
21:55 uh co-pilot right because it's for
21:58 coding it's it's it's basically AI pair
22:01 programming but if you
22:04 um haven't tried just you know pair
22:07 program pair programming with gpt4
22:11 it's pretty insane
22:14 um
22:16 I'll show you two things that are that
22:19 are kind of interesting
22:21 there we go
22:23 um
22:25 so so this is regular old chat GPT
22:30 um I'm just gonna go to default gpt4 and
22:32 I'll say
22:33 um
22:35 write me a python application
22:38 that
22:40 calculates fuel efficiency
22:43 of a car driving across country
22:47 and stopping for gas
22:53 every hour
22:59 and sure here's a simple Python program
23:02 that'll do that and it writes the python
23:05 code for you and then you can go off and
23:08 run the code and see if it works and and
23:10 you can just go back and forth I mean
23:11 the the other thing that that these
23:14 things are really good at they're good
23:15 at reading code they're good at
23:16 documenting code so maybe you have a
23:18 piece of code you can't find a bug in
23:20 you can just dump it in there it will
23:23 find the bug and fix it and probably
23:25 explain to you what's going on you can
23:28 take a chunk of code have it explained
23:30 to you what that code is you could do
23:33 that with trading Bots for example a lot
23:35 of people come in here and ask me about
23:36 trading which I know nothing about but
23:39 it can do that
23:41 um and then there's there's this mode in
23:44 chat GPT now called code interpreter and
23:47 this is even more insane so so what
23:49 we're going to do here
23:51 is we're going to go grab
23:54 this is a ZIP file I don't know what's
23:56 in it I do well I can't remember what it
23:59 was it was something we downloaded last
24:00 night from kaggle so if you want to play
24:03 with data so what code interpreter does
24:05 is you can upload a file to it I think
24:07 up to 100 megabytes and then you just
24:11 ask it [ __ ] about the file and it goes
24:14 and figures it out for you so we're just
24:16 going to say what's what is this
24:21 and why might
24:25 it be interesting
24:37 this appears to be a zip file
24:40 I would need to unzip it do you want me
24:42 to do that yes
24:45 you don't need to ask my permission to
24:46 unzip a zip file
24:48 so now it pops up that sort of plug-in
24:51 looking thing and if we open this what's
24:54 it doing
24:55 it actually wrote python code and
24:58 executed it it imported the zip file
25:00 Library
25:02 and wrote the code and unzipped the file
25:05 and so now it says the zip file contains
25:08 a single file named
25:11 backloggedgames.csv would you like me to
25:13 load and preview this so we'll go yes
25:19 and so again
25:21 it's opening pandas I think it's dumping
25:24 that into a what do they call it a data
25:27 frame or something within the virtual
25:29 machine so it can it now has all of this
25:33 data in its memory basically
25:38 so now it's analyzing what's in it
25:42 the data set might be interesting for a
25:44 variety of reasons could be used to
25:46 analyze Trends in the video game
25:48 industry to understand player behavior
25:50 and preferences what would you like to
25:51 do next with the data
25:54 please
25:56 visualize the data
26:00 in the most
26:02 uh
26:02 [Music]
26:04 the most beautiful way
26:09 possible
26:11 and that captures
26:14 what you think
26:18 are the most interesting trends
26:23 in the data
26:25 so this is the kind of stuff that you
26:27 would ask a data analyst right this is
26:29 the kind of stuff that if you were
26:30 tasked with finding out if your
26:32 marketing was working or if your sales
26:34 numbers why the the sales were not
26:37 working
26:38 um
26:39 you would either have to know how to do
26:41 this write your own python or you would
26:43 have to you know have someone at the
26:45 company that knows how to do this that
26:46 studied data analytics and now just as a
26:49 business person you can go
26:51 yeah just tell me what's going on here
26:53 and so it writes the code it writes it
26:57 creates charts those charts are not all
26:59 that interesting because the scale's all
27:01 [ __ ] up
27:05 um
27:08 more uh I'll say please
27:12 do a more in-depth analysis
27:18 so like I don't know [ __ ] about data
27:20 analysis so I'm like I'm like a bad one
27:22 to be demoing this but if you knew
27:24 anything at all about it I mean even
27:26 without knowing anything about it you
27:28 can see that this is pretty remarkable
27:29 what's going on here
27:31 and so
27:33 here so the
27:35 let's start with the first so now it did
27:37 that's an interesting table
27:39 a correlation Matrix of of game
27:42 attributes so so
27:45 you know
27:47 um comparing the different
27:50 game attributes let's see what else now
27:53 let's visualize the relationship between
27:55 rating and plays
27:58 and between reading and reviews let's so
28:01 it's going to create a scatter plot
28:06 so so there's your Scatter Plots between
28:09 ratings and reviews
28:11 like
28:12 I mean this is [ __ ] remarkable I mean
28:15 right this is insanity
28:18 it's insanity so what can you do with
28:21 this I don't know yet you know I asked
28:23 someone asked before when I said Bing
28:26 and and Bard can analyze images They
28:28 Don't Really analyze images but this
28:31 can anal analyze images okay there's a
28:34 Christopher walk-in that we made in
28:36 mid-journey the other night so let's
28:38 open this up
28:39 and let's say
28:41 what okay find
28:46 the top 15 average colors in this image
28:49 comma
28:50 create
28:54 a bar chart
28:57 of the percentage of each color
29:01 where the bar
29:04 is that color comma
29:06 and the names of the colors are creative
29:09 crayon colors
29:17 so let's see what this is doing here
29:20 so it pulled in
29:25 a couple of libraries
29:27 web colors numpy
29:30 it's still it's doing whatever it needs
29:32 to do to average the colors
29:34 using k-means k-means clusters clusters
29:38 equals 15. so I asked for 15 average
29:41 colors
29:44 let's see I apologize for the
29:46 inconvenience but appears that web
29:47 colors Library
29:50 is not available in this environment an
29:53 alternative
29:54 yes so do you want me to use an
29:57 alternative approach so check that out
30:00 it knew what it wanted to do it knew it
30:03 wanted to use a particular library and
30:05 then within code interpreter right now
30:07 that library is not available so it said
30:09 I can find another way to do it so this
30:14 says I apologize it seems like there was
30:16 an error colors hex wasn't defined
30:18 through the previous error let's correct
30:20 that by doing k-means clustering again
30:23 and then convert RGB to hex and finally
30:26 plot the dominant colors
30:31 um
30:31 still not happy with it so now it's
30:33 trying again so it's troubleshooting its
30:35 own work it's like
30:38 it's just [ __ ] crazy
30:42 see if we can get it to actually find
30:43 the colors
30:50 and right now code interpreter just uh
30:53 uh it looks like it might have hung code
30:55 interpreter just does python but you
30:58 know
30:59 when when you hear people talk about
31:00 autonomous agents or Auto GPT things
31:03 like that this is the kind of things
31:04 it's doing where you give it a goal and
31:06 then it goes off and does its thing
31:09 uh let me try continue
31:14 now
31:16 it's not it's still thinking it'll
31:18 probably time out here
31:20 in a minute but anyway isn't that wild
31:23 I mean
31:26 oh there we go
31:28 so there's the colors right so it it was
31:31 it was you know a face so so there there
31:34 was the those dark Moody colors
31:39 all right so so you get so so I'm gonna
31:41 say
31:43 um
31:45 no wait
31:47 that's great oh wait that's not what I
31:50 wanted
31:51 no
31:54 that's great you gave me the bar chart
31:57 comma but I also wanted you to rename
32:00 the colors in Creative crayon colors
32:04 that are appropriate
32:06 for each color period
32:09 and why don't you put it in a pie chart
32:11 instead
32:14 label each segment of the pie
32:18 with the new creative color name
32:27 so now it's got to read the the hex
32:30 values of the color and then somehow
32:32 reference that to figure out is it a
32:34 flesh tone is it a green is it a blue
32:36 or maybe it knows that just because of
32:38 where it is and let's go let's go look
32:40 at them so there's our chat pie chart
32:43 so Emperor's Emerald Sovereign Silver
32:49 let's see princesses Periwinkle Princess
32:51 Peach Queens quartz monarchs maroon
32:56 so this those seem Majestic mauve Regal
32:59 Rose some of these are too dark to see
33:01 Noble Navy Royal Red
33:05 so so you're taking some of the best of
33:07 what chat GPT does you're taking this
33:10 idea of autonomous
33:12 goal goal-oriented you know problem
33:15 solving
33:16 and you know
33:19 please note these names are creative and
33:21 do not necessarily reflect the exact
33:23 color they represent they're intended to
33:24 be a touch of royalty
33:27 so yeah so that's code interpreter
33:31 crazy it's nuts all right you wanna see
33:34 something else fun while we're here all
33:36 right
33:37 so there's a new feature so if you've
33:39 got in order to see all these these
33:41 fancy new features in chat gbt you have
33:43 to have chat GPT plus then even if you
33:45 have it you may not see them because you
33:47 have to explicitly turn them on so you
33:49 go to your settings
33:51 and then there's a beta feature section
33:53 so you turn on beta features and then
33:54 you actually have to enable them so turn
33:56 them on
33:57 but one of them here is called custom
33:59 instructions and if you've ever played
34:01 on the openai playground on the
34:03 playground
34:06 um you've probably seen system
34:08 instructions this is very similar to
34:10 that I think it's identical to that
34:13 so you've got about you and then what
34:16 you want the the chat bot to respond
34:19 with and so this is one I'll just turn
34:22 this back on from last night this one
34:23 was pretty funny
34:25 um let me read it to you
34:27 uh I'm a master thespian and pompous ass
34:30 I have a large vocabulary and I Relish
34:33 in sounding smarter and more dramatic
34:34 than people around me no matter what
34:36 anyone asks so this is um how would you
34:39 like chat GPT to respond no matter what
34:41 anyone asks asks start out with a
34:44 pretentious statement that draws from
34:45 the plot of a classic Broadway play
34:47 follow up that declarative statement
34:50 with an article formatted using markdown
34:52 complete with title subtitles and
34:53 bulleted lists no matter how trivial the
34:56 question was that was asked
34:59 um
35:00 compare
35:02 um
35:04 the writing of the article
35:09 to the
35:12 New Yorker
35:16 whenever possible
35:18 all right
35:21 all right so there's our thing so now no
35:22 matter what I do whether I do GPT 3.5 or
35:25 4 or code interpreter and sometimes this
35:27 is confusing if you forget to turn it
35:29 off I'll go to code interpreter to
35:31 analyze some stuff and it'll just be the
35:33 stupid pompous ass actor dude uh so I'll
35:36 just type in here
35:37 um
35:38 let's see uh we'll do something
35:41 non-theater related so we'll do
35:44 um why does fresh
35:49 asphalt smells so good
35:54 ah my inquisitive interlocutor
35:59 the Allure of a delightful Aroma that
36:01 emanates from fresh asphalt can be
36:04 likened to a poignant
36:07 actual production of life in the
36:10 melodious Overture of this olfactory
36:12 experience we find ourselves transported
36:14 to a scene reminiscent of the opening
36:16 acts of the streets of New York a
36:18 classic Broadway play
36:20 and then it uses
36:22 um markdown so notice it's got bold and
36:25 bullet points and things like that
36:27 and then it wrote the article The New
36:29 Yorker connection
36:31 so
36:32 while my example here is silly
36:36 um let's say you always wanted uh chat
36:39 GPT to respond with comparative tables
36:41 or you always wanted it to respond in
36:44 less wordy ways right right now Chachi
36:48 PT is a little wordy asking for stuff
36:50 it's like I'm so glad you asked Bob
36:52 let's dig into that
36:54 just want the answer you could actually
36:56 put that into these custom instructions
36:58 so
36:59 a solid interesting uh new feature
37:04 that came out
37:06 three days ago two days ago something
37:08 like that a lot of this stuff is new new
37:10 new new
37:11 welcome everyone if you're new my name
37:13 is Kyle Shannon this is the AI learning
37:15 lab these sites here are all different
37:17 flavors of Chachi BT if you have not
37:19 played with Chachi BT that's a
37:21 requirement of this channel you can hang
37:23 out here we hang out we cuss we have a
37:25 good time I drink iced coffee like why
37:28 wouldn't you want to watch that
37:30 old man on a tick tock old man on a tick
37:33 tock what'd you do Wednesday night well
37:35 I don't want to talk about it
37:41 um ask any questions you have below
37:43 here's the deal I'm 96 messages behind
37:46 right now so if I'm not answering your
37:48 question it's not because I'm not live
37:50 and it's not because I'm ignoring you
37:53 it's just because I'm going to answer
37:55 all the questions and it takes me a
37:57 while and that's the deal and if you
37:58 don't like it then you should go you
38:00 should go now you should leave what's
38:02 going on with the hair here's all right
38:03 tonight looks like
38:05 looks like speaking of Broadway it looks
38:08 like some uh
38:13 what am I in Greece
38:15 all right
38:20 Izzy Lux hola hello is Adobe Firefly a
38:24 separately buyable product I actually
38:27 don't know the answer to that that's a
38:28 really good question
38:34 can because it's only beta right now so
38:37 I don't know can you
38:40 subscribe
38:42 to Adobe
38:45 Firefly on its own
38:50 foreign
38:54 you need to have a Creative Cloud
38:56 subscription that includes Photoshop
39:01 nope
39:03 you cannot sorry
39:05 that's drag actually
39:07 hey Tobias shaken
39:09 that's a very animated wait what oh wait
39:11 I oh tick tock tick tock show me the
39:15 bottom I don't like it take that losing
39:19 cranky
39:22 someone said something about linktree
39:23 and don't ever trust anyone with link to
39:25 Troy Decker Kyle for president yeah
39:27 that'd be good
39:28 [Laughter]
39:34 um
39:34 oh man there's been a Cyber attack Kyle
39:36 what do we do I don't know [ __ ] deal
39:39 with it maybe
39:41 I drive my my Engineers crazy they're
39:44 they're like well you know what's what's
39:45 the product requirement here I'm like
39:47 product requirements actually quite
39:49 simple I wanted to [ __ ] work
39:56 laughs
39:59 ah it's comedy yeah we're gonna do
40:04 comedy all night maybe I'll just talk
40:06 like this all night no that even bugs me
40:09 all right sorry I'm
40:13 jumping all the way to the bottom I'm
40:14 trying to figure out where it was oh
40:15 here we go okay
40:17 Tobias hey oh you're you're very
40:20 animated and funny thank you
40:23 welcome to movie phone
40:25 all right where should newbies start
40:27 newbies should start
40:29 well actually there but let me go to one
40:32 other uh oh
40:33 did I lose my
40:36 yeah I did okay
40:37 newbie should start there so if if
40:40 you're just brand new
40:42 go to chat.openai.com
40:45 I'm trying to think if there's any
40:46 others I want to send you to
40:48 there or bing.com
40:50 you can go to any of them but but chat
40:54 chat.openingi.com that's the official
40:56 chat EBT website they're the ones that
40:58 started this whole [ __ ] mess
41:00 right ai's been around for decades
41:03 people have been slowly inserting it
41:05 into this we've been getting our cars to
41:06 drive with it
41:09 then November 30th 2022 Sam Altman and
41:12 team launched chat GPT and all this [ __ ]
41:15 hits the fan so that's what's going
41:16 that's why this is all insane right now
41:18 is because of that so go there that's
41:21 the original you know that's the
41:23 original burger
41:25 get the fries well done
41:28 and then when you're there since you're
41:31 brand new to this
41:32 go to this thing prompts.chat that's a
41:35 web address so just point your web
41:37 browser to prompts.chat and that's going
41:39 to take you to a thing called awesome
41:41 chat GPT prompts or something like that
41:45 and the first part of the document is
41:47 just a description of what these things
41:49 are and how they work what these large
41:51 language models are and how they work
41:52 and then the second half or the second
41:54 like
41:56 most of the document is all these
41:59 prompts that will
42:01 um tell chat jpt to act like a certain
42:04 Persona so act like a screenwriter act
42:06 like an accountant act like a therapist
42:08 act like an expert in history act like
42:10 whatever it is and what that document
42:13 will do is give you a real good sense of
42:18 how [ __ ] insanely deep these things
42:21 are so so when you go to them what
42:24 you're going to see is just a it looks
42:26 like a Google search bar but unlike
42:28 Google
42:30 that's just indexed the web these things
42:33 have been trained on all of the words on
42:35 all of the public internet
42:37 and it it has
42:40 mathematically distilled them into
42:44 some sort of friggin miraculous you know
42:48 what they call Vector space doesn't
42:49 really matter there's basically just
42:51 pairs of words and meanings and when you
42:53 type in your prompt it it answers you
42:56 directly and it can code it can write
42:59 poems it can write songs it can answer
43:02 questions it can analyze something
43:03 you've wrote and
43:05 pull the outline out of it it can tell
43:08 you where you've got grammatical errors
43:10 it's just it
43:13 so so here's the deal you owe it to
43:16 yourself if you're new to this stuff to
43:17 play with these tools long enough for
43:19 you to have what I call your Kevin
43:21 McAllister moment remember home alone
43:23 this moment
43:28 just play with them until you have that
43:32 and then the fun starts
43:35 and then the next thing after that is
43:37 like take something that you know really
43:38 well and see if it can do that
43:41 like maybe you're a copywriter or maybe
43:43 you're a project manager maybe you're an
43:45 account person or maybe you're a
43:46 salesperson
43:48 have it analyzed someone's LinkedIn
43:50 profile
43:51 just copy and paste their LinkedIn
43:53 profile pop it in there and say how
43:55 would you talk to this guy if I were if
43:56 you were a salesperson how would you
43:58 approach them thanks crypto Joe
44:00 I liked me some ice cream I like the
44:03 soft serve ice cream have you ever had
44:05 the soft serve ice cream sauce serves
44:07 really tasty the the harder ice cream I
44:10 like it but it's too hard and then all
44:12 of a sudden it's melting all over me the
44:14 soft serve I can just lick it
44:17 it's just delicious right from the
44:20 beginning that's why I like the soft
44:21 serve
44:23 okay
44:27 one of the things that gives me joy in
44:29 life is knowing that someone is flicking
44:31 through Tick Tock just sees that
44:37 the same person
44:39 ah
44:46 all right so hopefully that helps what's
44:48 your name uh Izzy Lux yeah that's where
44:51 you should start start there
44:53 just start with the words words words
44:55 words words Shakespeare once said words
44:57 words words words language is everything
45:02 language is everything
45:05 it's what gives us meaning
45:08 and these are called large language
45:09 models
45:12 and they are
45:18 hmm
45:23 I feel like Humanity reflectors so like
45:25 they've been trained on all the [ __ ]
45:26 we've put on the internet for the past
45:27 three and a half four decades
45:30 and their ability amplifiers so if
45:34 you're good at something these things
45:35 can make you really good at it if you're
45:37 not good at something these things can
45:38 make you good at it so instead of having
45:40 to put in your 10 000 hours the Malcolm
45:43 Gladwell ten thousand hours thing to get
45:45 good at a something
45:46 with this maybe you put in 100 hours and
45:49 you're [ __ ] brilliant at it
45:51 so if you have a hundred X Improvement
45:55 in what you can execute
46:00 versus what you could do without this
46:02 tool that's the kind of power these
46:04 things are
46:06 in in in San Francisco a lot of the uh
46:11 the tech shops that that have gone to
46:13 full you know AI assisted pair
46:16 programming for their Engineers they're
46:17 seeing 30 to 40 percent of productivity
46:22 30 to 40 percent
46:24 and I bet if they get better at it it
46:27 might go up and as these tools get
46:28 better it's going to go up
46:30 the other thing about these tools is
46:32 right now they're kind of off in the AI
46:35 ghetto right we've got our what we do
46:38 for work over here and then there's the
46:40 say I crap over here and we have ai
46:42 detectors we don't want anyone to use AI
46:44 we're going to put AI watermarks in here
46:46 and we're these things are all separate
46:48 but they're very very quickly being
46:50 integrated into all the tools that we
46:52 already use like for for example right
46:54 now I don't know if this is there for
46:55 everyone but it showed up for me three
46:57 days ago when I go to do a new post on
47:00 LinkedIn it now says would you like AI
47:02 to help with that
47:04 and so Microsoft is going to be rolling
47:07 chat GPT into all of office 365.
47:11 um Google's going to be rolling it
47:12 across all of their Suite
47:16 um meta has open sourced their
47:20 um
47:21 large language model llama2 they open
47:23 sourced it with a commercial license
47:25 this week
47:27 so so the entire developer Community
47:29 like all the GitHub and hugging face
47:31 repositories all the trending things are
47:33 all llama 2 projects because it's now
47:36 commercially licensable and it wasn't
47:38 the Llama one was was only for research
47:41 so this stuff's coming and it's coming
47:44 fast and it's deeply profoundly powerful
47:47 so uh yeah so that's why this Channel's
47:50 here like I want people to get let's
47:52 listen I don't care if you try this
47:53 stuff and then you go I [ __ ] hate it
47:55 I'm never gonna use it again cool
47:58 I got total respect for that love that
48:01 but if you're sitting on the sidelines
48:03 and like yeah I heard it's not perfect
48:06 I heard it's stealing from people
48:11 you're gonna get blindsided by it and
48:13 and
48:14 the the more you can get your head
48:17 around it the better position you're
48:19 going to be in as these changes start to
48:21 come
48:24 um
48:25 Tyron truth nice name I'm gonna go 8.5
48:29 out of 10 because it took me a while to
48:30 pronounce it but I like it
48:33 um so hello Heather can any of these
48:36 llms read actual uploaded documents or
48:39 does it have to be text no they can well
48:42 hmm
48:44 I don't know what you mean by uploaded
48:46 documents like documents with images and
48:48 things in them I would try
48:55 I don't think so I think right now
48:57 they're all a little weird
49:00 I would say the one that you have the
49:02 best shot at with that is code
49:03 interpreter with chat GPT where it could
49:06 probably write a python script where it
49:09 does some sort of OCR or some sort of
49:12 image recognition on the images and then
49:14 pulls out the text
49:17 um but like Bing and Bard can both see
49:20 images and they and they can OCR images
49:23 but if you've got a mix of text and
49:25 images I don't think they can do that
49:28 and then
49:32 code interpreters mostly like data Cloud
49:34 can only see data
49:39 and Poe I think po can only see data
49:43 so I I don't know that's a really good
49:45 question that's that's worth an
49:47 experiment
49:48 and in fact Heather I mean
49:51 I I think for anyone on the channel
49:53 right now I think a really interesting
49:55 exercise even if even if you've been
49:57 using these things for a while
49:59 these things are evolving so quickly I
50:02 think a good exercise would be to take
50:06 two or three different kinds of quotes
50:08 like maybe maybe you've got a creative
50:09 writing quote and then you've got
50:11 something where you make social media
50:13 content and then there's another one
50:15 where you do I don't know project
50:16 management stuff whatever take three
50:18 different quotes and run the exact same
50:21 quote through all of these different
50:23 tools and it's beautiful
50:27 um
50:28 it's a it's an interesting experiment we
50:31 did it last week with uh writing jokes
50:35 someone asked can these large language
50:36 models write jokes and I'm like not
50:38 really
50:39 and then we we confirmed that but gpt4
50:42 actually wrote a joke that was a joke
50:44 you know would it have survived an open
50:48 mic night no
50:50 it was a joke the other ones were the
50:52 other ones were pretty bad they were
50:53 joke structures but they weren't funny
50:58 all right I asked Bing image to create a
51:02 picture of a certain president sitting
51:04 at a table
51:05 there you go
51:07 I'll tell you the 2024 election is going
51:10 to be
51:11 somewhere between comedic and uh
51:14 [ __ ] terrifying with all the
51:16 um
51:17 misinformation and image stuff that's
51:19 going to happen and video stuff
51:21 I think Bing is using the bong
51:25 yeah Bing
51:27 I want to like think more than I do
51:29 right it doesn't have a good image
51:32 generation tool behind it I I really
51:34 wish open AI would come out with Dolly
51:36 three like something that's in the
51:38 neighborhood of stable diffusion and
51:39 mid-journey because Dolly's just shitty
51:42 right now and then
51:44 Microsoft has Bing chat so
51:47 safety handcuffed that it's just it's
51:49 just bad
51:51 but it'll it'll get better I don't like
51:53 I'm not taking the other thing about all
51:55 of these tools don't feel like you have
51:57 to master one of them and you have to
51:59 dive deep into it they're all changing
52:02 so fast like
52:04 they're evolving so fast that whatever
52:07 is required for you to get good work out
52:09 of these today you're not gonna have to
52:11 do that six months from now but
52:14 getting AI literate right now I think
52:17 will pay dividends I think you I think
52:18 we all have about a year that if we can
52:20 be AI literate and on this stuff and and
52:23 getting our heads around it and talking
52:25 about it and communicating it
52:27 is
52:29 like and then
52:31 like an unimaginable about amount of
52:33 opportunity
52:35 uh if you ignore it there's going to be
52:37 an unimaginable
52:39 unimaginable wait
52:41 I'm like Mitch McConnell did you see him
52:44 he just stop talking today my brain just
52:46 shut down unimaginable yes that
52:50 um amount of pain if you just ignore
52:52 this stuff because I think it's going to
52:54 transform nearly every position
52:56 certainly every position that is
52:57 mediated with a computer so whether it's
53:00 creative or whether it's knowledge work
53:02 or whether it's consulting or whether
53:04 it's
53:05 just even basic simple [ __ ] like
53:08 Community organizing
53:09 um everything's going to be impacted by
53:11 this stuff copy and paste the
53:13 conversation so it remembers what you
53:15 wrote you can do that
53:18 Maga
53:20 um up to a certain point so with
53:23 um
53:24 chat GPT it's got about a 3000 word
53:27 context window so if what you copy and
53:29 paste is too long it will forget the
53:31 [ __ ] at the top right
53:34 Claude has a hundred thousand token
53:36 context windows so that's like 75 000
53:38 words so if you want a conversation with
53:42 one of these chat Bots where it doesn't
53:44 forget what you talked about at the
53:45 beginning of the conversation use Claude
53:48 or if you want to take the Great
53:50 American novel that you wrote when you
53:51 were 17 and actually rewrite it so that
53:55 it doesn't suck you could paste the
53:57 whole thing in the cloud and tell it to
53:59 help you rewrite it
54:01 so claude's pretty [ __ ] amazing and
54:03 it's free
54:05 and it's if you want to know who who's
54:06 behind it the people that made gpt3 the
54:10 the sort of thing that started it all
54:13 quit open Ai and went and started
54:15 anthropic
54:16 so it's a bunch of ex-open AI people
54:19 that created it so it's it's an
54:21 interesting uh competitive this whole
54:24 competitive this whole competitive scene
54:26 right now is [ __ ] fascinating
54:28 yesterday Apple announced yeah
54:31 coming it's gonna [ __ ] rock you know
54:35 it's gonna rock so Apple's gonna be
54:37 added to this list and then elon's got
54:39 once once he figures out how to get the
54:41 Twitter logo off the building in San
54:43 Francisco
54:44 and he renames that thing X and once he
54:47 gets through the uh trademark issues
54:49 with Microsoft Xbox and uh Mark
54:52 Zuckerberg with all of his trademarks
54:55 for this thing x uh once he gets through
54:57 all that
54:58 and he gets his AI company up and
55:00 running that'll probably be on this list
55:04 um I would not be at all surprised if
55:08 um
55:09 Amazon doesn't do something Snapchat's
55:11 got one which I haven't even played with
55:13 Snapchat
55:15 um
55:16 yeah there's gonna be a bunch of them so
55:18 anyway it's crazy times crazy times it's
55:21 so crazy
55:24 it's a private did you see did you see
55:29 is this image generator paid or free I
55:31 don't know which one
55:37 um just like a cocktail
55:40 [ __ ] dancing monkey you put a virtual
55:42 hat on me I'm like
55:48 hey we could we could turn this into one
55:50 of those channels where I just I just
55:51 act like an NPC heart heart heart
55:58 heart
56:02 no all right fine fine
56:05 if you don't want me to make 700 a day
56:08 reacting to cowboy hats that's fine
56:10 we'll just talk about AI then fine
56:13 I'll be broke I'll stay broke
56:18 um the image generator I was using
56:23 well let me go see there was
56:35 so if you go to clipdrop.co
56:37 that's stable diffusion
56:40 and it's their it's their um
56:45 it's like their showcase so they're
56:46 different things so I think clipdrop.co
56:49 is free and you can use stable diffusion
56:52 XL the other thing that I was using was
56:56 Bing's image generation tool that is
56:58 free
57:00 leonardo.ai is built on top of stable
57:03 diffusion and it's
57:05 not free but they give you like 150 50
57:08 credits a day so you can use it some
57:11 amount
57:11 of the day for free
57:16 sorry I don't remember which one I was
57:17 using it was with too many messages ago
57:20 I'm very far behind on the messages I
57:23 don't know why people are still
57:24 impatient all right what was the AI that
57:27 could generate PowerPoints tome tome.app
57:30 t-o-m-e
57:31 dot app but also go to
57:36 futurepedia.io
57:40 and that's just a big old nasty [ __ ]
57:44 list of like 4 000 AI tools there there
57:48 I would bet I haven't been there because
57:50 I can't it makes my head hurt
57:53 I would be surprised if there weren't a
57:56 hundred 150 PowerPoint generating tools
57:59 out there so there's probably a lot of
58:01 them here's the problem of the 150 you
58:04 may find 145 are gonna suck really bad
58:08 and there's not going to be a good way
58:10 for you to know what's the difference
58:11 until you actually go use them so what I
58:14 would do personally and I mean you know
58:16 maybe this is because I'm comfortable
58:17 with PowerPoint I would use gpt4 I would
58:21 go to chat gbt gpt4 and tell it here's
58:25 the presentation I want to write I just
58:27 put in my my key you know key Concepts
58:29 and then have it [ __ ] write it and
58:32 then just just copy and paste that into
58:34 uh
58:36 into a power pointer keynote outline I'm
58:39 going to Max I use keynote but we're
58:41 switching oh my God I can't believe
58:42 we're doing this we're switching my
58:43 company from Google to Microsoft
58:47 it's been [ __ ] 11 years we've been on
58:50 Google products we're switching it to
58:52 Microsoft which is I think it's stupid
58:57 but
59:00 they're rolling GPT across all of what
59:03 they do we do a lot of work in Pharma so
59:07 like a lot of our farmers are blocking
59:09 Google meet
59:11 and uh I think a lot of our
59:15 applications may need to run in on Azure
59:17 service so
59:19 that's going to be painful ask me how I
59:21 am a month from now because it's going
59:23 to take us 30 days to make the
59:25 transition oh
59:29 it's cool you're talking with each other
59:31 I love that fantastic
59:34 let me I'm going to check in on what
59:36 you're talking to each other about here
59:37 let's see what do you mean scan text
59:40 you told me to [ __ ] off it said if I
59:43 persisted I'd lose my access what was
59:45 that what what service told you to [ __ ]
59:47 off
59:48 that's interesting no way for real
59:51 okay good to know I'll need to know yes
59:53 I was shocked I should have
59:54 screenshotted it
59:55 yeah what what was the service that told
59:57 you to [ __ ] off that's pretty
59:58 interesting
1:00:01 was it it wasn't perplexity
1:00:03 labs.perplexity because that's that's
1:00:05 llama and it probably doesn't have any
1:00:07 safety guard rails on it
1:00:09 those things will tell you to [ __ ] off
1:00:12 all right
1:00:15 all right
1:00:17 I also use please when prompting the AI
1:00:19 might end up being your boss of the
1:00:21 future it's funny like I don't like I
1:00:24 don't do it I I don't think about it as
1:00:26 you know our robot overlords I just find
1:00:27 myself it's really polite to me and I
1:00:30 just find myself going oh thanks that
1:00:32 was great like like
1:00:33 my experience in working with these
1:00:35 feels very similar to working with the
1:00:37 remote intern who needs a lot of action
1:00:40 gets a lot of [ __ ] wrong but is [ __ ]
1:00:43 brilliant
1:00:44 that's what these things are like and
1:00:46 they they tend to work best when you
1:00:48 give them lots of context you give them
1:00:50 lots of Direction you explain why you're
1:00:53 doing things you say things like please
1:00:54 and thank you and oh that's great you
1:00:57 know try it again but make it suck less
1:00:59 I hadn't I had an acting teacher to do
1:01:02 that once I was in an acting class
1:01:06 oh my God this is called this this woman
1:01:09 gets up and she does her scene and the
1:01:12 scene ends and he goes ah
1:01:14 yes yes yes
1:01:18 um
1:01:20 why don't we try that again and this
1:01:22 time uh why don't you try it as if uh
1:01:24 you had Talent
1:01:26 it's like the whole point
1:01:31 it's a rough business the old acting
1:01:34 business you know because all this fun
1:01:35 and games hanging out at the Red Carpet
1:01:38 oh no
1:01:39 oh
1:01:43 let's try that again this time as if you
1:01:46 had Talent
1:01:48 what's the website address again for
1:01:51 what I don't know which one I was on but
1:01:53 I'll tell you
1:01:55 more Talent than Hollywood
1:01:58 come here first right now can't find a
1:02:01 sag actor I got a tick tocker for you
1:02:04 [Laughter]
1:02:07 I'm SAG eligible I still get I get I did
1:02:10 one film I did New Jack City in the 80s
1:02:12 I still get checks from that from sag
1:02:14 and after
1:02:16 Kyle for president I think that would be
1:02:18 a very bad idea be entertaining Maybe
1:02:21 I've been comparing Bard versus GPT and
1:02:24 GPT is a far better writer it is gpt4 is
1:02:27 amazing there's nothing close to it
1:02:30 um ask both to write a song and compare
1:02:32 yeah no but but listen Bard was
1:02:34 laughably bad for months and then the
1:02:37 new Gemini model that sits underneath it
1:02:40 is
1:02:43 um it's also interesting looking at the
1:02:45 the Llama models that that you can play
1:02:47 with at perplexity they're not bad but
1:02:51 they haven't been fine-tuned like one of
1:02:52 the reasons chat gbt is so good is they
1:02:54 fine-tuned the [ __ ] out of it so you got
1:02:56 your your base large language model down
1:02:58 here and then you got all of the sort of
1:03:00 you know power prompting they've got
1:03:02 sitting up in the application layer
1:03:05 um and then and then apparently gpt4 is
1:03:08 like they instead of it being one
1:03:10 massive model they broke it into like 12
1:03:12 experts and then there's 120 layers so
1:03:14 when you do a prompt it's doing all
1:03:16 sorts of
1:03:18 you know segmentation of which tool it
1:03:20 uses when and which expert it uses when
1:03:22 so there there's whatever they're doing
1:03:24 with gpt4s is just of of a different
1:03:28 Beast than what everyone else is doing
1:03:30 right now hello from Danville California
1:03:32 what's up Johnny Bean
1:03:34 let's shake it welcome welcome to the AI
1:03:37 learning lab this is the lab
1:03:39 you might go but Kyle kind of looks like
1:03:42 a shitty home office you'd be right
1:03:44 that's fantastic Johnny
1:03:46 [Laughter]
1:03:50 I need to get out
1:03:53 oh
1:03:55 all right
1:03:57 it's gonna be fine it's gonna be good
1:03:58 crypto Joe I like being the least for
1:04:01 code that doesn't surprise me although
1:04:05 um I hear that uh that uh
1:04:09 co-pilot for uh GitHub is pretty good
1:04:13 right because that
1:04:15 that was specifically tuned to do code
1:04:19 right so it's Chachi PPT but but tuned
1:04:21 for coding don't trust anyone from
1:04:23 linktree
1:04:25 don't trust anyone from linktree
1:04:29 is that because you went to my thing
1:04:31 there and there's a link tree or because
1:04:33 you went to my salon and there's a link
1:04:35 tree there
1:04:38 I must be evil
1:04:41 he's evil he doesn't answer the question
1:04:43 it's got a link tree
1:04:46 don't click on anything he'll drain your
1:04:49 wallet he'll drain your wallet don't do
1:04:51 it for the love of God don't click on
1:04:54 the link tree don't click on the link
1:04:56 tree
1:05:00 what was the last thing he said
1:05:02 I don't know he said something about I'm
1:05:04 going to click on the link tree and I
1:05:06 said no
1:05:11 no all right
1:05:13 glad we could help
1:05:15 this is exactly what I needed to know I
1:05:17 know link trees right exactly
1:05:20 I don't know what you're talking about
1:05:21 Heather but glad whatever I said was
1:05:23 what you needed to know oh my god wow
1:05:26 again not sure what that's in reference
1:05:28 to possibly my hair oh my it's gone
1:05:32 South the glue has broken it's it's it's
1:05:34 taken on a life of its own this and you
1:05:38 know what are we gonna do We're not
1:05:39 gonna do anything we're just gonna let
1:05:40 it entertain itself all right I've tried
1:05:43 to use this to code macros uh with VBA
1:05:47 and it hasn't really been able to
1:05:49 correctly any tip I don't know what this
1:05:51 is do you mean chat gbt I would say
1:05:55 um go to gpt4
1:05:59 um
1:06:00 for for
1:06:01 gbt4 is better than GPT 3.5 at coding
1:06:06 and don't go to Bing's gpt4 go to that
1:06:09 one
1:06:11 um I would also say try Bard because
1:06:13 Bard is better at code
1:06:15 than it is at words and it might be
1:06:18 better with macros
1:06:20 um I don't know enough about VBA and
1:06:22 macros in Excel to be able to help you
1:06:24 I'm relatively clueless on that other
1:06:27 than I know that you can do them I know
1:06:31 that if you know what you're doing you
1:06:32 can do some pretty remarkable [ __ ] and I
1:06:34 know that when I've been stuck like in
1:06:36 in Google Sheets trying to get it to do
1:06:38 something and I don't know what the
1:06:39 formula is and my ADD will not allow me
1:06:42 to look through a long list of formulas
1:06:45 yeah
1:06:46 since that from happening whatsoever
1:06:49 um it helps it's gotten me out of a
1:06:51 bunch of jams um but if you're doing
1:06:53 some more sophisticated stuff I would
1:06:55 say just
1:06:57 it there's so many variables here so so
1:07:02 if you're if you're new to prompting
1:07:03 Mika or Micah Micah Mika Micah um
1:07:08 uh Wyatt
1:07:11 um
1:07:15 if you haven't done things like tell it
1:07:17 to act like uh an Excel macro and VBA
1:07:22 specialist and and kind of you know sort
1:07:25 of focus the conversation down and have
1:07:27 it act like that kind of specialist that
1:07:29 will actually make a difference because
1:07:31 the way the large language models work
1:07:33 if you just give it a generic you know
1:07:35 give me a macro that does this it's kind
1:07:38 of looking at all of what it knows out
1:07:40 on the internet which includes the right
1:07:42 stuff where people have solved that
1:07:44 correctly and all of the people who have
1:07:46 not solved that correctly and put you
1:07:48 know bad information out there so if you
1:07:51 tell it to act like a VBA specialist it
1:07:53 will kind of narrow the the semantic
1:07:56 you know ooze that it's that it's
1:07:58 searching in and that may help
1:08:00 the other thing is
1:08:03 um do what's called prompt stacking
1:08:06 um where instead of trying to write one
1:08:08 perfect prompt and get it to give you
1:08:09 the perfect answer you could say
1:08:12 um
1:08:13 I'm trying to write some macros uh with
1:08:16 VBA
1:08:18 um is that something you understand yes
1:08:20 that's something I understand great
1:08:22 the area of vbas I want to do you know
1:08:26 centers around automating pivot table
1:08:29 filtering I don't [ __ ] know whatever
1:08:30 it is right and then and then you know
1:08:33 is that something you can help with and
1:08:34 just keep layering the prompts so what
1:08:36 you're doing is giving giving it
1:08:38 increasing thanks thanks Bear for the uh
1:08:40 the Roses I appreciate that
1:08:42 um it's giving an increase in context
1:08:44 where you're getting more and more
1:08:45 specific so it should hallucinate us
1:08:51 um the other thing to try if you've got
1:08:53 access to
1:08:54 um
1:08:55 to GPT plus is try code interpreter see
1:08:58 if see if grab some VBA out of excel pop
1:09:02 it into into a file throw that into code
1:09:05 interpreter and have it help you rewrite
1:09:07 it there because it's right that's
1:09:09 writing and executing python code so if
1:09:12 your VBA has to have any formulas in it
1:09:15 maybe the the python would would execute
1:09:18 that better than just chat gbt I don't
1:09:20 know
1:09:22 um
1:09:23 this is a all we're doing now is
1:09:25 mansplaining which is mansplaining we're
1:09:27 just saying words we don't know anything
1:09:28 we don't know any facts we don't know
1:09:30 any series we don't know nothing but
1:09:32 we're doing it confidently we're doing
1:09:34 you know what could help you Mika Micah
1:09:36 Micah Wyatt Mike Mika Michael Michael
1:09:38 Wyatt is you can do all this but I don't
1:09:40 know nothing so it's that's just
1:09:42 mansplaining
1:09:44 that's I'm a mansplainer that's a
1:09:46 mansplaining does the AI man's planing
1:09:48 lab
1:09:56 I sense that Bart is less emotionally
1:09:58 intelligent yeah Chris you know why
1:10:00 because it's all Engineers over there
1:10:05 everyone tries to figure out why is
1:10:07 Apple so good because they understand
1:10:09 the human side of things it ain't just
1:10:11 about the engineering
1:10:15 Google GPT Ai and go to the one opening
1:10:19 uh someone's helping someone else that's
1:10:21 cool data Is My Jam if data is your jam
1:10:23 man code interpreter you should be all
1:10:26 over code interpreter Heather all over
1:10:28 it all over there
1:10:32 me
1:10:37 excellent pedal crafts wow GPT writes
1:10:40 children's stories and you can enjoy the
1:10:42 stories there's actually a plug-in for a
1:10:45 chat GPT called stories that writes you
1:10:47 a children's book and illustrates it
1:10:51 the illustrations like the the prompt
1:10:53 work that they did on the illustrations
1:10:54 is pretty shitty hopefully they'll make
1:10:56 it better but there's a plug-in that
1:10:58 does that too
1:11:00 it's yeah the the children's book thing
1:11:03 is really cool insane I don't know what
1:11:05 I was doing but I was probably doing
1:11:06 code interpreter and people are saying
1:11:07 that's insane how fun to be able to
1:11:09 generate a new bedtime story yeah
1:11:11 personalized bedtime stories so this is
1:11:13 I just wrote a an article on LinkedIn
1:11:16 called the uh the dawn the dawn of hyper
1:11:19 personalization
1:11:20 um I think we're entering an era where
1:11:22 you know your your child wants a story
1:11:24 for the night you don't have to make it
1:11:26 up you just generate it and have it
1:11:29 starring your kid and your kid could be
1:11:31 a different animal every night you know
1:11:35 um and and very very soon we're going to
1:11:37 be able to say I want to watch Top Gun
1:11:39 but I want Jim Carrey you know
1:11:43 as as uh the Cable Guy
1:11:46 starring in Top Gun and it'll just start
1:11:48 playing like that that kind of [ __ ] is
1:11:50 coming like rather than having Spotify
1:11:52 playlists you're gonna have a Spotify
1:11:54 playlist where it's all original music
1:11:57 mixed with your favorite music but the
1:11:59 the original music is music that
1:12:03 is like knows what you like and every
1:12:07 single song is like a hit for you
1:12:10 foreign
1:12:13 is this chat gpt4 or a plug-in I think
1:12:17 what I was showing
1:12:20 was just regular chat gbd4 but then I
1:12:22 did code interpreter code interpreter is
1:12:25 a plug-in it's just a really [ __ ]
1:12:26 fancy one that's good
1:12:29 um you prompt your idea of what is the
1:12:32 story about it starts to write and makes
1:12:35 edits along the way
1:12:37 um another thing to play with if if you
1:12:39 want to if you want to do the the uh
1:12:42 Marvel at this thing thanks Groove
1:12:44 attack uh Marvel at this thing writing
1:12:47 narratives is go find a project called
1:12:50 uh
1:12:51 GPT author
1:12:54 that implemented on a Google collab
1:12:56 notebook so you have to be a little
1:12:58 geeky to to get it up and running but
1:13:00 once it's up and running
1:13:02 um it will write you a novel
1:13:05 you give it your open AI key and you
1:13:08 give it your stable diffusion key
1:13:12 and then you give it a topic you give it
1:13:14 a writing style and you give it a number
1:13:16 of chapters and then it you just watch
1:13:18 it it writes the outline
1:13:20 and it says okay outline complete
1:13:22 writing chapter one
1:13:23 writing chapter two writing chapter
1:13:26 three and just goes through writes all
1:13:27 your chapters then it goes to stable
1:13:30 diffusion and creates cover art for your
1:13:33 book based on what it wrote
1:13:35 and then it saves that as an ebook and
1:13:37 you can open it in
1:13:39 any ebook reader
1:13:42 um there's also a thing called
1:13:44 showrunner AI that you give it a prompt
1:13:47 and it will write you or it will create
1:13:49 for you an episode of South Park
1:13:52 with voices and animations and music and
1:13:55 story
1:13:57 um and both of those are kind of crappy
1:13:59 right now like the the book that it
1:14:01 writes is kind of crappy but it kind of
1:14:04 just wrote a book for you like in five
1:14:07 minutes
1:14:08 and you know the South Park thing is a
1:14:10 similar sort of thing so don't look at
1:14:12 those projects for what they are look at
1:14:14 them for what they represent
1:14:16 right
1:14:18 this idea of a truly hyper personalized
1:14:21 future
1:14:23 let's go
1:14:25 let me show you another one that I I
1:14:27 think this one is just
1:14:30 stupid fun
1:14:33 so this is
1:14:35 blockade labs this is called
1:14:39 skybox.blockaidlabs.com so here's a
1:14:43 cool little 3D scape
1:14:46 right
1:14:48 but we can do
1:14:50 let's see
1:14:52 we can draw a door
1:14:57 Brown
1:14:59 [Music]
1:15:04 and we'll do a little window
1:15:09 and we'll do a little window
1:15:13 and then we'll do
1:15:16 um
1:15:22 door leading out
1:15:25 to The Stairway to Heaven
1:15:28 in a rock and roll universe
1:15:32 and then I'm gonna go
1:15:38 dreamlike
1:15:40 and I'm gonna go generate
1:15:44 and now it's going to do that so it's
1:15:46 going to make me
1:15:49 some Doorway to a Stairway to Heaven
1:15:53 we'll see if it got any of the reference
1:15:56 to the rock and roll universe is there
1:15:58 gonna be a guitar here there better be a
1:16:00 [ __ ] guitar
1:16:09 here we go
1:16:12 there's the door there is no Stairway to
1:16:14 Heaven this is a massive disappointment
1:16:16 there's a stairway not to heaven where's
1:16:18 the rock and roll here oh what's this
1:16:21 oh that's cool oh
1:16:24 there's a little portal
1:16:26 that's pretty cool
1:16:28 but anyway so and we could you know we
1:16:31 could do like I could do Neo
1:16:35 Tokyo
1:16:38 at midnight
1:16:41 in the rain and do I think there's a
1:16:45 cyberpunk style
1:16:47 uh
1:16:50 generate
1:16:54 and so now it's going to go do that so
1:16:56 now what this is doing this is actually
1:16:57 kind of a simple thing you can't really
1:17:00 enter this world but someone has now
1:17:02 figured out how you can hack this so you
1:17:05 kind of combine it with Unreal Engine
1:17:08 and there's a way you can take images
1:17:10 from this tool put them in Unreal Engine
1:17:13 drop in 3D objects and then actually
1:17:17 start interacting in the world so we are
1:17:20 not that far from the ability to
1:17:24 that's pretty cool
1:17:25 we're not that far from the ability to
1:17:28 say hey I want to play this kind of game
1:17:30 in this kind of style with this kind of
1:17:32 game mechanics and these kind of objects
1:17:34 and these kind of weapons and these kind
1:17:36 of characters with these three friends
1:17:38 and I want to play it right now
1:17:41 and we'll just start playing it and you
1:17:44 know the initial versions of these games
1:17:46 are going to be like you know the real
1:17:48 game developers are going to be like
1:17:50 yeah but it's not Triple A it's not it's
1:17:53 not a real game it's that's janky
1:17:55 gameplay and then a year after that the
1:17:58 janky gameplay will be gone and then
1:18:00 there's Tripoli game developers are
1:18:03 either going to be you know
1:18:05 dealing with this reality or not and if
1:18:07 not they're just gonna watch the demand
1:18:10 for bespoke handcrafted games go down as
1:18:14 the demand for hyper personalized games
1:18:15 goes up
1:18:17 right because if I can play that kind of
1:18:19 game that I like I love
1:18:22 um trigger Twitchy games and I like
1:18:24 racing games I can't stand strategy like
1:18:27 my 80d I can't do it so I got particular
1:18:30 things I like but I like certain Styles
1:18:32 but they always have the gameplay that I
1:18:34 don't like well what if I could do a
1:18:36 style that I like that's normally
1:18:38 reserved for you know MMOs and [ __ ] like
1:18:42 that
1:18:43 and get all the adrenaline [ __ ] that I
1:18:47 need
1:18:49 it you know instead of the uh the
1:18:51 strategy stuff well I can I I will be
1:18:53 able to do that right and that's
1:18:56 that's crazy that's just [ __ ] nuts
1:18:59 it's nuts
1:19:02 just cuckoo it's just cuckoo
1:19:06 I don't know if anybody really
1:19:07 understands they're all like oh this is
1:19:09 over hyped this is over hyped
1:19:11 I'm a question for you
1:19:14 what if it's under hyped
1:19:17 hmm
1:19:22 what was the original picture oh I don't
1:19:26 know what I was doing but if you remind
1:19:27 me what I was doing I'll grab the
1:19:29 original picture Wild
1:19:33 ah Magnum CI yes Thelonious Thelonius is
1:19:37 the one that coined the term Magnum CI
1:19:39 what we talked about last week was that
1:19:41 the
1:19:42 product name code interpreter is [ __ ]
1:19:45 horrible
1:19:47 because it's way more magical than code
1:19:50 interpreter
1:19:52 um so Thelonious came up with the Gen X
1:19:54 friendly Magnum CI which I like it's
1:19:57 solid
1:20:01 but I don't think they're going to go
1:20:03 for that so we gotta we gotta rename it
1:20:04 at some point we're going to do that
1:20:05 dang it it would have been funny to name
1:20:08 the colors with puns of his name
1:20:11 cold please
1:20:28 please all right
1:20:31 [Laughter]
1:20:34 is there a way to share gpt4 with my
1:20:37 partner like sharing Netflix
1:20:39 um well
1:20:40 so so you have a couple of options if if
1:20:43 it's like a cost thing like you don't
1:20:44 want to pay the 20 bucks
1:20:46 I I assume you could
1:20:48 I don't know if you can both be logged
1:20:50 in at the same time but
1:20:52 claude.ai is free Bing is free Bing is
1:20:56 gpt4 it's chat ebt4 but it's it's kind
1:21:00 of Microsoft's version of it so it's a
1:21:01 little bit hobbled but but it's it's
1:21:04 you can do image generation there
1:21:09 um Bart is free
1:21:11 perplexities free Pi is free pose got a
1:21:15 bunch of free stuff in it
1:21:17 so there's a bunch of free options here
1:21:21 um I know that you can share prompts
1:21:24 from chat gbt with someone else but they
1:21:27 have to have their own
1:21:28 uh kind of thing but
1:21:32 yeah I don't know I
1:21:34 I don't quite know what you're trying to
1:21:35 do river but but there's enough free
1:21:38 options here that that I think you have
1:21:40 some options without having to have two
1:21:42 subscriptions
1:21:44 and with the ability to share your
1:21:46 prompt
1:21:47 um I think
1:21:48 I think if you share a prompt your
1:21:50 partner could bring a GPT 3.5 which is
1:21:53 free and just use your prompt within GPT
1:21:55 3.5 so that might help what filter do
1:21:58 you use to have your face
1:22:01 in a square while on a computer it's
1:22:03 called dual camera
1:22:07 so when you go live there's just a
1:22:10 button there for dual camera if you
1:22:11 don't see that button that's because
1:22:13 Tick Tock took it away from you because
1:22:15 you're a bad human being like I had dual
1:22:18 camera for a while and then they took it
1:22:20 away for like a month because I was a
1:22:22 bad boy it was bad it was very bad they
1:22:25 took it away
1:22:27 and then for no reason then give it back
1:22:31 why because they're [ __ ] Tick Tock I
1:22:33 don't know they make their own rules
1:22:34 they do what they do
1:22:35 they banned me twice for seven days and
1:22:37 I have no idea what I said
1:22:40 one time I think I was talking about
1:22:41 bias and I think they just went oh if
1:22:43 it's biased it's bad get out
1:22:48 I was trying to
1:22:50 explain what was going on with bias and
1:22:52 they didn't want to know they didn't
1:22:53 want to know they didn't want to know
1:22:55 you know what they wanted to know
1:22:56 nothing they were like get off stop
1:22:58 you don't do lives for seven days get
1:23:01 out get out now
1:23:02 poof gone
1:23:05 this is fantastic thank you thank you
1:23:09 the website to make images of cars yeah
1:23:12 that was that was
1:23:16 uh
1:23:17 um
1:23:18 I always forget the name of a
1:23:19 clipdrop.co
1:23:24 clipdrop.co let me go to the beginning
1:23:27 of it
1:23:28 so if you go here
1:23:32 there's like all sorts of Cool Tools
1:23:33 here so this is stability ai's little
1:23:37 playground
1:23:38 so like you can
1:23:40 they got stable diffusion XL that's
1:23:42 where I made the cars they've got
1:23:44 uncropped they've got clean up they've
1:23:47 got stable doodle is really cool have
1:23:49 you seen stable doodle yet look at this
1:23:55 um
1:24:00 [Music]
1:24:09 foreign
1:24:24 what the hell is he doing I don't know
1:24:26 he does this every night why does he do
1:24:29 this we don't know
1:24:31 why do you watch him we don't know
1:24:34 all right so there's that and then I can
1:24:36 put the prompt uh let's see
1:24:39 um
1:24:41 blonde blonde haired
1:24:48 funny old man
1:24:51 generate
1:24:56 so this site is clipdrop.com that's
1:25:00 actually kind of cool
1:25:02 those are good
1:25:03 I like that one
1:25:06 I mean
1:25:08 is it
1:25:10 made or is
1:25:12 it is is no one else like just [ __ ]
1:25:14 like
1:25:15 just [ __ ] gobsmacked daily that we we
1:25:20 get to play with these tools
1:25:23 I mean this [ __ ] insanity I drew a
1:25:26 ridiculous
1:25:28 stupid wait how do I get out of here
1:25:31 how do I get it how do I get back oh
1:25:33 back to grid I just I draw this stupid
1:25:36 thing
1:25:40 and it turns into that
1:25:43 without trying
1:25:49 I mean
1:25:52 I don't know
1:25:55 when I hear people like well it's not
1:25:57 perfect it's got seven fingers
1:26:01 I'm like
1:26:02 have you ever tried to draw
1:26:05 you [ __ ] idiot you ever tried to
1:26:08 write a screenplay you know how [ __ ]
1:26:10 hard it is
1:26:12 have you ever tried to outline a
1:26:14 screenplay you know how [ __ ] hard
1:26:16 that is
1:26:24 right
1:26:30 and don't forget
1:26:35 all right it's an important lesson we
1:26:38 like to learn here over and over and
1:26:40 over again
1:26:41 uh I'm also in my journey with
1:26:44 generative AI I love the conviction in
1:26:45 your Tick Tock lives thank you very much
1:26:47 what's your name
1:26:49 wait a beer uh AB here up
1:26:56 banerhi
1:26:58 abhira banerhi cool name
1:27:01 very hard for my American Mouth to get
1:27:04 my tongue around but it's very cool
1:27:07 uh
1:27:08 um
1:27:13 paste all the comments into chat gbt and
1:27:17 ask it to highlight the best questions
1:27:19 can I do that how do I get to all the
1:27:21 comments
1:27:25 anybody now go left go right oh wait
1:27:32 how do I copy them all
1:27:35 no
1:27:38 hmm
1:27:40 oh this is wild I didn't know this was
1:27:43 here how do I copy all the comments
1:27:45 anybody swipe left to close
1:27:47 these are a hell of a lot easier to read
1:27:50 if I if I swipe left look look
1:27:55 you know it's like
1:27:57 oh man learn to use tick tock
1:28:04 he was so cute when he discovered copy
1:28:07 and paste
1:28:12 oh my god oh what am I doing with my
1:28:16 life what am I doing people what am I
1:28:18 doing with my life I don't know we're
1:28:20 having fun we having fun if we're having
1:28:22 fun then glad jingus
1:28:25 let's use terms like by jingus
1:28:28 I don't know how to copy and paste the
1:28:30 comments but but that's a really good
1:28:32 idea actually so if I can figure out how
1:28:34 to copy well I'll tell you what dear
1:28:36 text
1:28:40 does tick tock not realize that some
1:28:43 people might actually have fun with the
1:28:45 hat
1:28:47 they give it to you for a second and a
1:28:49 half come on
1:28:51 how do I get GPT bus you go to
1:28:53 gptbus.com
1:28:59 McKenzie's a good dude I would go I I
1:29:02 support you supporting him
1:29:04 GPT boss is a cool project and by the
1:29:07 way
1:29:08 um uh open ai's custom instructions kind
1:29:12 of [ __ ] him because custom
1:29:13 instructions are kind of gbt because uh
1:29:17 but now he created custom instructions
1:29:18 oh I lost my place sorry I lost my place
1:29:21 in the comments when I tried to figure
1:29:22 out how to copy questions I completely
1:29:25 lost my place in the question so we'll
1:29:26 get back to it we'll get there
1:29:28 eventually it's gonna be so good
1:29:31 it's gonna be so good it's gonna be so
1:29:33 worth the wait
1:29:34 what'd you do Wednesday night I don't
1:29:36 know
1:29:37 I think I watched some old guy try to
1:29:38 learn how to use tick tock that sounds
1:29:40 awful yeah
1:29:43 what are you doing tonight I think I'm
1:29:45 gonna watch the same guy
1:29:46 what's going on with you I don't know
1:29:49 all right
1:29:50 I've still lost in my questions I'm just
1:29:52 making fun of myself at this point
1:29:55 all right Magnum CI we're back to that
1:29:57 the website make the images and cars
1:30:00 we're back to that oh
1:30:03 you're so funny oh my God yeah but it
1:30:07 looks aren't everything
1:30:08 [Laughter]
1:30:10 is the sound randomly static for anyone
1:30:13 else Dr J I think that's probably me
1:30:16 every time I switch cameras I think it
1:30:17 [ __ ] up the Stream
1:30:20 and I switched the cameras a lot because
1:30:22 you know I'm touched
1:30:25 I'm a survival crafting gamer okay
1:30:27 awesome 10 out of 10. text to
1:30:30 personalized game would be a dream come
1:30:32 true I'm telling you man I used to I
1:30:34 used to joke with my kids because I I
1:30:37 just could not for the life of me figure
1:30:38 out why they played these strategy games
1:30:41 and so I said hey I got an idea I came
1:30:43 up with an idea it's called an RT Triple
1:30:45 G it's a real-time grass growing game
1:30:48 and you plant seeds and then you watch
1:30:51 it grow in real time
1:30:53 it's like
1:30:56 they play like that that's the level of
1:30:58 excitement they have for me
1:31:01 RT Triple G baby
1:31:04 gonna be a new genre
1:31:07 I I think watching this channel is
1:31:10 probably like playing an RT Triple G
1:31:12 what's VBA uh Visual Basic something
1:31:16 it's basically
1:31:18 programmed into Excel to make it do
1:31:21 fancy [ __ ]
1:31:22 it might not be Visual Basic it's
1:31:24 probably not Visual Basic virtual basic
1:31:26 I don't know it's something to do with
1:31:29 Excel what you have ADD I never would
1:31:31 guess that okay solid troll solid troll
1:31:35 it's kind of funny kind of mean
1:31:38 kind of a compliment not sure if it's a
1:31:42 real put down or like that's my favorite
1:31:44 kind of troll the one where you can't
1:31:46 quite figure out if you just got slapped
1:31:48 in the face or not but that's solid it's
1:31:51 solid it's good
1:31:52 stressful week
1:31:54 um it's actually been a good week it's
1:31:56 actually been a good week
1:31:57 um
1:32:00 a lot of this uh
1:32:02 one of the things I talk about in here
1:32:05 a lot is go learn to use this [ __ ] it's
1:32:09 what it's why I'm a [ __ ] walking
1:32:11 billboard for open AI
1:32:14 um and I'm not getting paid for them
1:32:16 from them by the way
1:32:18 but one of the reasons I talk about it
1:32:20 is I actually think it's really
1:32:21 important I think good things will come
1:32:22 of it I've had a very good week this
1:32:24 week because I've been talking about
1:32:26 this [ __ ]
1:32:27 um not just on this channel but just in
1:32:29 general I've been talking about it on
1:32:30 LinkedIn I've been putting myself out
1:32:31 there I got the salon and people are
1:32:34 starting to reach out to me and it's
1:32:36 leading to real opportunities for my
1:32:38 business so it's been a good week
1:32:40 because it's the business environment
1:32:42 right now is just [ __ ] weird anyone
1:32:46 else experiencing this like like I just
1:32:48 call it bad corporate Behavior no one's
1:32:51 returning [ __ ] emails they don't pick
1:32:52 up their phones they don't show up to
1:32:54 meetings they don't show up to meetings
1:32:55 they don't really apologize for it
1:32:57 they're like oh yeah let's do something
1:32:59 and then three months later you're still
1:33:00 trying to chase them down like it's I've
1:33:02 never seen it like this we've in in 11
1:33:04 years of my business we've never had
1:33:06 collections issues until this last year
1:33:09 and it's not just small companies right
1:33:11 oh they're struggling no no this big
1:33:13 Pharma companies just not paying their
1:33:15 [ __ ] bills like what is going on
1:33:25 that's some Gen X sass
1:33:27 [Laughter]
1:33:30 he started out kind of calm and dim and
1:33:33 then he just got raging for a while I
1:33:35 don't know what was going on
1:33:38 I'm by the way I'm trying to find my
1:33:40 place in the questions again this is
1:33:42 this is this is what I do now
1:33:46 oh
1:33:49 I might be thinking Kyle why don't you
1:33:50 just skip some questions like I don't
1:33:53 know have you ever been in my head no
1:33:58 I'm codependent okay I want to answer
1:34:02 everyone's questions I want you all to
1:34:04 be happy
1:34:06 mmm
1:34:14 welcome to Moviefone is that iced coffee
1:34:16 why yes it is
1:34:20 remember Juan Valdez ah don't we all
1:34:23 those are fantastic days in the 70s
1:34:28 when your coffee was tasty tasty and
1:34:31 laced with arsenic oh yeah
1:34:33 bring back the super fund
1:34:36 hello Emilio's wife welcome welcome
1:34:38 welcome I know you've probably been here
1:34:40 for half an hour I'm so far behind on
1:34:42 questions but you know I'm trying to
1:34:44 figure out where I was in my questions I
1:34:46 just saw you joined
1:34:47 river is there a way to share ah here we
1:34:49 go this was River's question about
1:34:51 sharing Chach EBT with her partner this
1:34:54 is fantastic the website that makes
1:34:55 images for cars now I'm too far back the
1:34:58 beta was released
1:34:59 with my Creative Cloud screen
1:35:01 subscription in the US
1:35:04 best AI for iOS okay
1:35:08 so for me so there there is an official
1:35:10 chat GPT app for iOS
1:35:16 there is an official Claude
1:35:19 is there an official Cloud app I can't
1:35:21 remember
1:35:24 I think Bing has an app
1:35:28 or doesn't
1:35:30 um
1:35:35 I use I just use Google Chrome or Safari
1:35:39 with the chat GPT website I don't use
1:35:42 the chat GPT app because
1:35:45 it's hobbled it's not as good and the
1:35:47 site is so simple that it works as a
1:35:49 website so I I would say the best one is
1:35:51 that but
1:35:52 the one that I think is most interesting
1:35:54 right now I wouldn't say it's the best
1:35:55 but the one that's most interesting is
1:35:58 pi the PI app Pi dot AI or just the pi
1:36:02 app on iOS
1:36:04 it's just like chat gbt but it's way
1:36:07 more conversational and it's got a mode
1:36:09 where you just sort of flip it into
1:36:10 phone mode and then you just talk to it
1:36:12 and it talks back there's six voices you
1:36:14 can choose from
1:36:16 and like what I've been doing is I I pop
1:36:18 it in my car I put it on Bluetooth and I
1:36:20 just start having a conversation with
1:36:22 all of the history of the internet you
1:36:25 know and it's connected to the internet
1:36:27 so you could ask it like you could be
1:36:29 driving to a place and say you'd be
1:36:31 driving to the mall and say hey what
1:36:33 movies are playing at the mall and it'll
1:36:34 tell you and then you say what are the
1:36:36 reviews like for that one oh here's
1:36:38 what's going on with the reviews for
1:36:39 that one it's really interesting and it
1:36:41 like for me pi kind of gives me a
1:36:44 glimpse into what's coming so if you
1:36:46 imagine pie without the delay
1:36:50 with a less annoying robotic voice and
1:36:53 with an animated
1:36:55 you know either version of you or
1:36:57 stylized you or your dog or whatever
1:37:02 where you just have this persistent
1:37:05 assistant oh someone trademarked that
1:37:07 the persistent assistant huh
1:37:10 we're doing nothing but printing money
1:37:12 in this channel like you take persistent
1:37:15 assistant you turn that into an app no
1:37:18 one can pronounce it and you'll probably
1:37:20 lose some money but anyway but seriously
1:37:21 someone will remember it and then
1:37:24 as much about
1:37:27 per month this is
1:37:30 mrr recurring Revenue up to ten dollars
1:37:34 per month with persistent assistance
1:37:36 you're welcome
1:37:38 you heard it here first thank you thank
1:37:39 you
1:37:40 thank you all
1:37:43 Bob good to see you
1:37:47 all right but my colleagues in India
1:37:49 didn't get the beta on theirs yet oh
1:37:52 that's interesting
1:37:53 oh oh for Photoshop yeah well so Adobe
1:37:57 right now
1:37:59 I could potentially get you an answer on
1:38:01 that I know a guy at Adobe
1:38:02 um
1:38:04 they're really still in beta like
1:38:06 they're really still trying to figure
1:38:08 out like how this stuff
1:38:11 they're in a weird position as is Google
1:38:13 right how do you how do you how do you
1:38:16 take something that
1:38:18 in a lot of ways is a direct competitor
1:38:21 to your existing product in Google's
1:38:23 case this stuff is a direct competitor
1:38:25 in adobe's case it's tangential
1:38:28 um but I I would assume that they're
1:38:30 slow rolling it internationally
1:38:32 um is just my guess and that's probably
1:38:35 something you could you could Google
1:38:37 uh so yeah I don't have a clear answer
1:38:40 for you that on on that one Heather
1:38:42 sorry about that um hey it's Mr Shannon
1:38:45 Jeff will AI replace software developers
1:38:48 yes because I'm a CS student I'm worried
1:38:50 about my future okay so Jeff here's the
1:38:53 deal
1:38:55 yes they will and
1:38:59 they're still going to be the need for
1:39:01 engineers and programmers and computer
1:39:03 scientists for years and years to come
1:39:05 but what you do and how you do it is
1:39:08 likely going to shift dramatically
1:39:10 so if what you're spending your time
1:39:12 doing right now is like learning python
1:39:14 or learning C sharp or learning Swift or
1:39:17 whatever it is whatever framework react
1:39:19 node.js whatever the [ __ ] you're
1:39:20 learning
1:39:22 and you're thinking like okay I'm gonna
1:39:24 learn to code and I'm going to they're
1:39:26 not ready
1:39:29 that shit's going away
1:39:32 now
1:39:34 what stays well critical thinking
1:39:37 product design
1:39:39 um understanding code logic how things
1:39:42 go together understanding things like
1:39:45 um
1:39:46 like like the the operation side of
1:39:49 things like like
1:39:51 all of these tools are cool but if I
1:39:53 want to do this Within
1:39:55 a large organization how do we do stuff
1:39:58 that's this cool but is also safe and
1:40:01 private and has correct permissions
1:40:03 right so so the operations of AI is
1:40:06 going to be critical
1:40:08 um learning automation
1:40:10 um
1:40:11 you know whether you do it with code or
1:40:14 not learning things like Lang chain
1:40:16 where you can
1:40:17 um uh train these AI on on your own data
1:40:21 or on company data learning things like
1:40:23 Auto GPT and baby AGI autonomous agents
1:40:27 um playing around with code interpreter
1:40:29 if you're not right now pair programming
1:40:33 with github's copilot you should be
1:40:35 right because if you're a competent
1:40:38 programmer getting out of school
1:40:41 you're going to have some level of value
1:40:42 if you're a competent programmer that's
1:40:44 AI literate getting out of school like
1:40:47 your value goes like this
1:40:49 so keep learning what you're learning
1:40:51 but later on this stuff
1:40:54 and and if I were you like if I had the
1:40:56 chops and the and the the brain to be
1:40:58 able to do more engineering stuff like I
1:41:00 would be joining
1:41:02 open source projects I'd be I'd be
1:41:05 participating in Lang chain I'd be
1:41:07 participating in the autonomous agent
1:41:09 things
1:41:10 um I'd be go check out Mark Shapiro is
1:41:14 it Mark Shapiro yeah I think it's Mark
1:41:15 Shapiro and YouTube who talks about AGI
1:41:18 he's written five books on it he's got
1:41:20 three or four open source projects for
1:41:23 that
1:41:24 so I would just [ __ ] go all in learn
1:41:27 everything you can be the most
1:41:31 and then learn whatever the [ __ ] they're
1:41:34 teaching you at college
1:41:35 and quite frankly the deeper you dig
1:41:38 into this you may end up disrupting
1:41:41 yourself so rather than saying you know
1:41:43 hey I just learned all this [ __ ] and you
1:41:46 know they don't need it now you might
1:41:48 figure out oh
1:41:49 I might start a company doing something
1:41:51 because I'm really excited about this
1:41:53 new [ __ ] I would lean in heavy with this
1:41:55 stuff if I were you because I think
1:41:57 there's going to be massive demand but
1:41:59 not a few ignore that stuff
1:42:03 what's going on champion
1:42:05 you gotta pee what are you doing
1:42:08 oh he does
1:42:32 [Music]
1:42:36 come on computer come back
1:42:39 come back come back come back is that
1:42:42 iced coffee wow yes it is movie phone
1:42:46 yes cool stuff
1:42:52 which does everyone prefer stable
1:42:54 diffusion or mid-journey okay so here's
1:42:57 my thoughts on this let me flip over to
1:42:59 the slide
1:43:01 to that slide
1:43:03 so I've got mid-journey Leonardo and
1:43:05 Adobe fireflies as the three
1:43:09 uh image generation things I put here
1:43:12 the only reason I put Leonardo there
1:43:14 instead of stable diffusion is it's just
1:43:16 a web front end in front of stable
1:43:17 diffusion
1:43:19 um
1:43:20 for Pure image
1:43:24 amazingness mid-journey is the winner in
1:43:28 my opinion
1:43:30 but
1:43:32 I think it's really good together and
1:43:35 they just released Excel 1.0 today so
1:43:39 that's probably my guess is it's in the
1:43:41 neighborhood of a mid-journey competitor
1:43:44 um the thing about mid-journey is
1:43:46 because it's open source there is so
1:43:49 much activity from the development
1:43:51 community that if you're geeky and have
1:43:54 the the tolerance to keep up with all
1:43:56 that stuff I I would go mid-journey
1:43:58 because there's just so much going on in
1:44:00 the development Community right there's
1:44:01 all sorts of Cool Tools there's control
1:44:03 net there's automatic 1111 which is a
1:44:06 cool front end for it there's all these
1:44:08 extensions there's Laura's which are all
1:44:10 these different styles that you you know
1:44:13 you can sort of fine tune
1:44:15 um stable diffusion to do pretty much
1:44:17 anything you want it to do but it it
1:44:20 requires
1:44:21 some
1:44:22 some
1:44:24 rather robust tolerance for the
1:44:27 geekatube
1:44:29 a lot of geeky Tunes with stable
1:44:30 diffusion
1:44:31 so and mid-journey you know mid Journey
1:44:35 so if you haven't played with
1:44:36 mid-journey you you can only generate
1:44:39 images right now in Discord and if
1:44:42 you're wondering well what the [ __ ] is
1:44:43 Discord it's basically slack for gamers
1:44:46 you know if you know slack for your
1:44:49 business it's kind of like slack but
1:44:50 more annoying and
1:44:54 um
1:44:56 and so you have to kind of do command
1:44:58 line things to generate your images but
1:45:00 the images are [ __ ] miraculous like
1:45:03 they're just mid journey is amazing
1:45:06 um
1:45:08 so I think that's I think right now it's
1:45:10 just like Different Strokes for
1:45:11 different different folks and and you
1:45:14 know if you're if you're in a
1:45:15 corporation if if you're in a commercial
1:45:17 environment the Adobe products are the
1:45:19 way to go because they've been trained
1:45:21 on public domain and Adobe stock images
1:45:24 they have the license for they're going
1:45:26 to be more artist friendly where they're
1:45:28 going to allow artists to opt out of
1:45:29 being trained opt into being trained
1:45:31 attribution models
1:45:34 um compensation
1:45:36 to do it right so if you're in a
1:45:38 corporate environment I would say the
1:45:40 Adobe products
1:45:41 and then if you're
1:45:43 if you're just out in the world and want
1:45:45 to make part of pictures
1:45:47 then you know for me it's mid-journey
1:45:50 because mid journey is just prettier and
1:45:51 I like the pretty you know I like the
1:45:53 Purdy
1:45:54 I like Purdy I don't mind the geeky but
1:45:56 I like to Purdy
1:45:58 and and you know they're close enough
1:46:00 now that
1:46:03 I would definitely
1:46:06 I'm I am definitely paying attention to
1:46:08 stable diffusion
1:46:10 because I think it's a really
1:46:11 interesting company they're open
1:46:12 sourcing all of their Tech
1:46:16 their CEO is an interesting dude
1:46:19 I forget his name but he's super
1:46:21 optimistic like he makes me look like a
1:46:23 pessimist
1:46:24 he's just like she's like in a year
1:46:26 you'll be you know describing the movie
1:46:28 and then watching it you know it's like
1:46:30 yeah it's probably not coming that quick
1:46:32 dude but but what they're doing is
1:46:34 really interesting he entire ass thing
1:46:36 if I have thousands of hours of live
1:46:39 stream video is there a way to feed an
1:46:41 air model that data
1:46:43 maybe I don't
1:46:45 um
1:46:46 what do you want to do with it do you
1:46:48 want to
1:46:49 take the transcriptions and find
1:46:51 interesting bits of it like I would
1:46:53 think that would be expensive but if
1:46:55 you've got it on a platform is it on
1:46:57 Twitch or is it on
1:46:59 like where where did you live stream it
1:47:01 and is it on a platform that's that's
1:47:04 already done the transcription for you
1:47:06 because if you've got transcriptions
1:47:08 from all that then you can just run that
1:47:09 through
1:47:11 um through some of these tools you could
1:47:13 maybe even make an automation on zapier
1:47:15 or something like that
1:47:17 to just you know point to you pop a URL
1:47:19 in there have it take that transcript
1:47:22 and generate a whole bunch of things
1:47:23 from it
1:47:25 um I built one of those today for my
1:47:27 business
1:47:30 um
1:47:31 foreign
1:47:42 like Auto analysis of video but it ain't
1:47:45 cheap and it's like if you've got you
1:47:47 know if your live stream is two hours or
1:47:50 an hour like these I've been doing and
1:47:53 they're um
1:47:55 they're I've been doing these like on
1:47:56 average they're probably two and a half
1:47:58 hours long
1:47:59 and so
1:48:00 there's there's some [ __ ] in here that's
1:48:03 worth replaying but it's like I don't I
1:48:05 don't necessarily want to spend a ton of
1:48:07 money going to find it all so I don't
1:48:09 know
1:48:10 Tick Tock does a thing where they try to
1:48:12 identify some clips but they're they're
1:48:14 all over the place and they don't let
1:48:15 you it's there's not an easy way to go
1:48:17 in there and mess with them if someone
1:48:19 knows how to do that let me know because
1:48:22 I don't I'm just like old man with Tick
1:48:24 Tock I'm like here's three Clips can I
1:48:26 do anything with these doesn't look like
1:48:27 it but maybe there's some magical
1:48:30 [ __ ] pixel I gotta click that goes
1:48:32 into the actual editor I don't know uh
1:48:35 can you have it perfect oh predict the
1:48:37 lottery numbers well you could certainly
1:48:39 have a pick them
1:48:42 uh that's the Kyle Shannon Comedy Hour
1:48:45 my favorite time of the day oh Lauren
1:48:47 I'm so sorry
1:48:50 foreign
1:48:53 here at the yeah learning lab
1:48:55 welcome to yuck yucks
1:48:58 try the iced coffee tip your waitress
1:49:03 uh
1:49:06 AI is in Snapchat now yep I know it's
1:49:09 it's in that and it's in some other
1:49:11 thing
1:49:13 it's gonna be everywhere I'm telling you
1:49:15 this stuff is
1:49:17 I just I I still can't get my head
1:49:19 around the fact that there's so many
1:49:20 major players right now and there's and
1:49:22 there's some that are not on that list
1:49:25 there are AI skills wait there are AI
1:49:29 skills you can enable in Alexa yeah oh
1:49:33 in Alexa oh that's interesting
1:49:36 yeah I'm also curious if apple is going
1:49:39 to
1:49:40 um
1:49:42 now Siri sucks less or are they gonna be
1:49:46 like all right that was fun while it
1:49:48 lasted here's a new thing
1:49:50 I think they have to bury Siri I think
1:49:52 it's just I think it's got
1:49:54 I think it's just not
1:49:56 I think it's got bad branding but I
1:49:58 don't know
1:49:59 Apple always surprises me
1:50:02 uh elect the ability to become useful to
1:50:05 come to useful conclusions or I had to
1:50:07 do all the thinking and feed it I don't
1:50:09 know which one you're talking about
1:50:11 yeah the the I I mean the the art of
1:50:14 prompting is
1:50:16 you know in a lot of ways is more like a
1:50:19 psychological exercise than an
1:50:21 engineering exercise
1:50:24 um because you really do have to think
1:50:25 in language and if you're not getting
1:50:27 the answer you want
1:50:29 it's it's kind of like I actually I've
1:50:32 never thought about this for as a
1:50:33 metaphor
1:50:34 prompting in a lot of ways is like an
1:50:36 escape room
1:50:38 right or if you ever played the game
1:50:39 Mist any if if I got got any uh gen xers
1:50:42 any if I got any gen xers on the line
1:50:45 welcome welcome who played Mist it was
1:50:47 on a shiny thing called A CD-ROM
1:50:51 um
1:50:53 prompting prompting these large language
1:50:55 models is is
1:50:58 it's it's a similar cognitive
1:51:01 exercise to escape rooms where it's just
1:51:04 like like it gives you a shitty answer
1:51:06 and you're like I know the answer is in
1:51:08 there somewhere
1:51:09 and then
1:51:12 and then you kind of go around this
1:51:14 corner and you go around that corner
1:51:15 you're like I've looked at these drawers
1:51:17 seven times is there anything different
1:51:19 in any of these drawers oh
1:51:22 wait why is this one upside down click
1:51:24 oh
1:51:25 oh you know and then you figure out that
1:51:28 layer of the escape room that that's
1:51:30 kind of what it feels like to do these
1:51:31 so I wouldn't give up on it just yet I
1:51:34 like I dig deeper and sort of peer
1:51:36 around the corners of whatever you're
1:51:38 doing linguistically add more context
1:51:40 sort of narrow the conversation before
1:51:46 um
1:51:48 you know ask it things like do you
1:51:51 understand you know what I'm asking you
1:51:53 or if it gives you a shitty answer
1:51:57 ask it
1:52:00 um to review the answer and tell you
1:52:03 tell you how it would improve its own
1:52:05 answer and that will sometimes give you
1:52:08 dramatically better answers
1:52:12 that's the nerdiest thing I may have
1:52:13 ever typed I don't know what you typed
1:52:16 oh wait if you have this chat to wait
1:52:20 if you have this chat text to speech
1:52:22 baby oh oh I see that's funny uh
1:52:26 Fender T-shirt oh yeah and the baby and
1:52:30 uh dogs barking
1:52:32 champ
1:52:35 all right are there any safeguards to
1:52:37 keep AI
1:52:39 from being used for for weird stick
1:52:45 as in comedy
1:52:51 uh
1:52:53 there there are safety guard rails on
1:52:56 all of them except the that one
1:53:01 labs.perplexity.ai
1:53:04 sitting underneath that well so
1:53:06 perplexity.ai that's the whole thing but
1:53:09 labs.perplexity.ai they're letting you
1:53:11 play with
1:53:12 meta's llama 2 models
1:53:16 and those are open source and they don't
1:53:18 have any guard rails on them so people
1:53:20 can do weird stick
1:53:22 with them all they want
1:53:24 um all the other ones have safety guard
1:53:26 rails and in fact they just made a big
1:53:28 announcement Google open AI anthropic
1:53:31 inflection
1:53:35 and one or two others just agreed
1:53:39 that when they all
1:53:42 make the next generation of these things
1:53:44 here's the thing
1:53:48 [ __ ] mind-blowing as gpt4 is like
1:53:52 what's coming
1:53:54 is just gonna be like a whole other
1:53:56 level
1:53:57 so they've all agreed that they're gonna
1:54:00 they're gonna collaborate on the safety
1:54:02 guard rails that they
1:54:04 um
1:54:06 that they put in place to keep these
1:54:08 things from
1:54:10 weird shtick is an interesting way to
1:54:12 put it you know this ain't exactly a
1:54:14 Vaudeville act gone bad the AI learning
1:54:17 lab is a Vaudeville act con pad
1:54:19 hahaha
1:54:20 but but these things gone bad are going
1:54:22 to be way more significant than that
1:54:26 dogs back all right Kyle would you give
1:54:29 us daily homework assignments
1:54:31 um
1:54:34 would you create five project teams
1:54:36 maybe ten possible
1:54:38 um
1:54:39 I don't know I mean
1:54:42 I did a I did a thing on my
1:54:45 on The Tick Tock Channel called
1:54:48 um the 2023 smart challenge
1:54:51 and the 2023 smart challenge was this
1:54:54 shit's coming
1:54:56 whether you want it to or not
1:54:58 and so the challenge was this there were
1:55:00 three three components to the challenge
1:55:03 the first component was
1:55:06 um do two projects using a large
1:55:09 language model so chat GPT or or one of
1:55:12 these
1:55:13 and one of those projects should be a
1:55:15 personal project and the other one be
1:55:17 should be for someone else
1:55:20 so personal project could be
1:55:24 create a meal plan create an exercise
1:55:26 routine
1:55:27 start the Great American novel you've
1:55:29 been procrastinating for 27 years
1:55:33 um
1:55:34 I don't know teach yourself
1:55:37 Raspberry Pi it doesn't [ __ ] matter
1:55:39 use these things
1:55:42 to do a personal project come up with
1:55:44 something for yourself and use these
1:55:46 things to do it in the second half of
1:55:48 that first
1:55:50 is do that for someone else
1:55:52 so for your for your business for your
1:55:55 family for a community group
1:55:58 maybe your community center
1:56:01 I don't know
1:56:04 needs to needs to come up with ideas for
1:56:08 um programming for the next year and
1:56:10 everyone's uninspired and burned out and
1:56:12 it's been 100 degrees for two weeks
1:56:14 straight and no one no one's into it go
1:56:17 into the community group say hey I'm
1:56:19 teaching myself AI would you mind if I
1:56:22 came up with programs using AI for the
1:56:25 group and they'll be like no okay
1:56:26 whatever
1:56:27 I heard it's evil yeah I know but I'm
1:56:30 just it's an experiment do you mind if I
1:56:32 experiment and do something for them
1:56:35 in doing that thing for yourself and for
1:56:37 someone else you will learn more about
1:56:39 what these things can do than than any
1:56:41 amount of books or prompt lessons or any
1:56:43 of that stuff if you come at this stuff
1:56:45 from the outside like start with the
1:56:47 technology I've got to learn prompt
1:56:49 engineering it's it's really foreboding
1:56:52 and like complicated and it's hard to
1:56:55 see what's possible but the minute you
1:56:57 go oh they need to come up with
1:56:59 interesting activities for six different
1:57:01 age groups let me go try that could you
1:57:05 do that
1:57:06 oh [ __ ]
1:57:09 I don't know I could do that
1:57:12 could you organize that in a table sure
1:57:16 oh damn
1:57:18 hour later you walk back in with a plan
1:57:21 for 2024 activities for your community
1:57:23 group
1:57:24 so that does a couple of things
1:57:27 you'll learn how to use these better
1:57:29 you'll learn what's possible
1:57:31 you'll help someone else
1:57:35 and then
1:57:36 if you do it even remotely well
1:57:40 they're going to be like how did you do
1:57:41 that so fast well let me show you
1:57:45 and then you you have your own little
1:57:47 mini version of what I'm doing here
1:57:50 so that's that's
1:57:54 um task one of the 2023 smart challenge
1:57:56 two projects one personal one for
1:57:59 someone else yeah I like it
1:58:02 welcome Barbie
1:58:05 um
1:58:06 the second one was make 10 images using
1:58:10 an image generation tool like Bing you
1:58:13 can make images on Bing
1:58:15 or you could do mid Journey if you want
1:58:17 to learn how to use Discord or you could
1:58:19 do Adobe Firefly if you've got a
1:58:21 Photoshop or generative fill or you
1:58:23 could do Leonardo
1:58:26 make 10 images why
1:58:28 because
1:58:30 until you've had the experience where
1:58:32 you type in words and it [ __ ]
1:58:34 miraculously generates something that
1:58:36 you didn't think it could possibly know
1:58:38 and it does it well
1:58:41 you just you you haven't quite
1:58:43 experienced what
1:58:45 these things make possible right because
1:58:48 with the words you're kind of like well
1:58:50 I've been using Google for 30 years and
1:58:53 20 years and and I know like these
1:58:56 things are just like Google but fancier
1:58:58 but when you make images
1:59:01 where the thing that comes back is just
1:59:04 in a whole other class of anything you
1:59:06 could have imagined was possible just
1:59:09 replace image generation with every
1:59:11 other media type music
1:59:14 film
1:59:15 um programming
1:59:17 architecture interior design product
1:59:20 design like anything anything
1:59:23 is coming so so the second half of it is
1:59:26 to just get you in touch with the
1:59:29 multimodal thing right there's words and
1:59:31 then there's the rest of the [ __ ] we
1:59:32 deal with in life these things are gonna
1:59:35 be remarkable and transformative in all
1:59:38 of that so that's the second thing of
1:59:40 the 2023 smart challenge the third
1:59:43 component of the 2023 smart challenge is
1:59:46 [ __ ] tell people about it tell people
1:59:48 what you're learning be generous
1:59:55 didn't she
1:59:56 like if you're like well I'm not
1:59:58 Technical and I'm not really good at
1:59:59 this doesn't [ __ ] matter
2:00:02 you have friends over for dinner
2:00:05 so I was uh
2:00:07 anyone checked out that Chat gbt thing
2:00:09 yet oh I heard that stupid
2:00:11 oh no I don't think it's stupid here's
2:00:13 what I did with it oh that's kind of
2:00:15 cool
2:00:17 have that conversation talk about it at
2:00:19 work talk about it with your friends why
2:00:23 because it's a cult no because this
2:00:26 shit's coming and like once you have
2:00:28 your Kevin McAllister moment once you
2:00:30 have this oh
2:00:32 you'll get it you'll you'll get that the
2:00:34 possibilities here the opportunity here
2:00:36 is [ __ ] massive
2:00:38 but not if you're on the sidelines and
2:00:40 so talk to people about it
2:00:42 because what's going to happen is one of
2:00:45 those people at some point down the line
2:00:47 is gonna go I need to figure this out
2:00:49 for my business oh
2:00:51 Sally was talking about that at dinner
2:00:53 the other night Sally yeah that AI stuff
2:00:56 you were talking about can you come
2:00:57 teach me about that
2:00:59 boom there's a job
2:01:03 or a project or some interesting new
2:01:07 thing in your life
2:01:08 so that would be the assignment that I
2:01:10 think I would give is I would just go
2:01:12 back to that thing I think I launched
2:01:14 that in
2:01:15 January
2:01:19 two projects one personal one for other
2:01:21 people ten images and talk about it
2:01:27 I'm surprised 30 bucks extra from
2:01:30 Microsoft I'm not because this stuff's
2:01:32 expensive and it's the it's a business
2:01:35 audience and it's a captive business
2:01:37 audience so they're gonna say well
2:01:41 um
2:01:42 the the the co-pilot the chat GPT that
2:01:45 we build into Office 365 and that we
2:01:48 build into Azure and that we build into
2:01:49 all this other [ __ ] it's going to be
2:01:51 private it's going to be secure it's
2:01:53 going to be out
2:01:56 and
2:01:57 It's Gonna Make You
2:01:59 three to forty percent more productive
2:02:01 as an organization as you incorporate
2:02:04 these tools minimum
2:02:06 within your organization so 30 bucks a
2:02:09 month on a whatever it is 250 or 300 a
2:02:13 month per seat license for all the other
2:02:15 [ __ ] you're paying for is is within
2:02:18 reason for business so
2:02:21 um and that's that's the thing that
2:02:23 Google doesn't have right now right
2:02:25 Google's relying entirely on Advertising
2:02:28 Revenue
2:02:29 and
2:02:31 you know they've given away their office
2:02:33 suite they haven't charged for it and
2:02:35 and listen it was brilliant it was
2:02:37 brilliant for a while because they took
2:02:38 all the market share away from office
2:02:41 but now you know
2:02:45 now
2:02:47 Microsoft has this cool [ __ ] they're
2:02:50 gonna charge 30 bucks for it and you
2:02:52 know Google can't just start charging
2:02:54 for [ __ ] they've never charged for so I
2:02:57 don't know it's it's fascinating this is
2:03:00 fascinating like business
2:03:02 competitive
2:03:04 landscape times
2:03:07 it's wild it's just wild
2:03:09 Claude wouldn't edit the style of Jane
2:03:11 Austen but Chachi PT did a great job
2:03:13 yeah I know gpt4 is just [ __ ] better
2:03:17 it's just better
2:03:19 but Claude like here's here's how I
2:03:21 would probably use Claude and gpt4
2:03:23 together
2:03:24 if I I would take long documents or
2:03:27 disparate documents I'd put them into
2:03:28 Claude I'd use Claude to organize them
2:03:31 and to and to block out maybe the
2:03:34 outline or the structure of what I want
2:03:36 to write
2:03:37 you know sort of pulling in the raw
2:03:39 materials that are in an unorganized
2:03:42 fashion I'd organize them in Claude I'd
2:03:44 then take that synthesized
2:03:47 [ __ ] I didn't Claude put that into gpt4
2:03:50 and then actually start working with it
2:03:51 that that would that would be how I
2:03:54 would handle that because because
2:03:55 there's
2:03:57 the the the
2:03:59 context limit that GPT is really
2:04:02 frustrating and I haven't hit it yet
2:04:04 with Claude
2:04:06 some of them are making a hundred
2:04:08 dollars a minute
2:04:09 oh some of the NPC people Emilio's wife
2:04:14 well start putting cowboy hats on me
2:04:16 I'll go cowboy hat cowboy hat cowboy hat
2:04:21 tutu pink tutu totally lights twirly
2:04:24 lights I can do this
2:04:26 could be an NPC come on
2:04:30 um
2:04:32 [Applause]
2:04:37 I guess I gotta do something about the
2:04:40 hair like an NPC wouldn't have hair like
2:04:42 that
2:04:44 uh good Lord
2:04:52 complimenting the model and boosting ah
2:04:55 it's a ticket attack yeah well dang it
2:04:58 it took my hat away
2:05:00 it took my head away they gave me a hat
2:05:02 they took it away oh any hats around
2:05:05 here
2:05:06 I just I'm just I'm just remembering
2:05:09 back to that troll comment before you
2:05:11 add what a surprise what [ __ ] looking
2:05:15 for a hat
2:05:16 all right
2:05:22 uh
2:05:24 [Music]
2:05:26 uh neither Bard chat GPT will generate a
2:05:29 woman in a bikini
2:05:31 Bard or chat gbt
2:05:34 neither of those do images do you mean
2:05:37 Bing and
2:05:39 well she's got a stable diffusion you
2:05:42 can see all the women in bikinis you
2:05:43 want
2:05:44 act like you have talent I know that was
2:05:47 that was a pretty cold burn
2:05:49 uh so what's happening 30 minutes ago
2:05:52 that's cute
2:05:54 see okay
2:05:55 app user three three eight eight six
2:05:58 three six eight by the way 2.5 out of 10
2:06:00 on the name come on you can oh Apple
2:06:04 user all right I'll give you 3.5 out of
2:06:06 10 because it's because it's Apple user
2:06:08 not just app user but uh that's a good
2:06:10 troll what's happening 30 minutes ago
2:06:12 because I'm so far behind on the
2:06:14 questions
2:06:15 you see how this works here we troll
2:06:17 that it's kind of funny it's kind of a
2:06:20 compliment and it's kind of a burn
2:06:22 that's solid that's solid
2:06:25 it's we got yeah we got clever we got
2:06:28 clever participants here at the AI
2:06:29 learning lab by the way if you're new
2:06:31 here looks like we got a fair amount of
2:06:32 folks my name is Kyle Shannon this is
2:06:34 the AI learning lab if you ask me what
2:06:36 what my qualifications are I'll [ __ ]
2:06:38 tell you uh
2:06:42 um all these things here are all
2:06:44 different flavors of chat GPT if you
2:06:46 have not played with them go now go do
2:06:48 it now just just go do it don't [ __ ]
2:06:52 sit there like I don't wanna I'm busy
2:06:55 I'm swiping no go there and just
2:06:59 tell it to write you a country song
2:07:01 about your Tubby Dog
2:07:03 and then go oh
2:07:05 that was fast that's not bad oh my
2:07:09 God but it can't do what I do oh it can
2:07:12 oh good lord should I be scared
2:07:16 go have that moment and if you don't
2:07:18 know what to type there go to
2:07:19 prompts.chat
2:07:21 that thing right there
2:07:23 and that'll teach you about prompting
2:07:25 and all of the [ __ ] that you can make
2:07:27 chat GPT do for you
2:07:30 like like this will start to scratch the
2:07:32 surface of it and then
2:07:34 um
2:07:35 yeah and then there's other [ __ ] there
2:07:36 so
2:07:38 yeah go do that
2:07:39 all right and if you have questions
2:07:41 about AI popping below I I do answer all
2:07:43 the questions if I can
2:07:45 um although I'm 99 plus behind now so
2:07:48 yeah I'm at least half an hour behind
2:07:51 hey Mr cgo hey everyone the bear
2:07:54 yeah I'm a chief generative officer I
2:07:57 don't know if you knew that ask for my
2:07:58 qualifications
2:07:59 I'm a chief generative officer what's
2:08:02 that you might ask
2:08:04 if only I knew okay
2:08:07 I'm generating [ __ ]
2:08:10 could be Revenue
2:08:12 could be bad copy
2:08:15 um I like to ask Bing to tell me about
2:08:16 her dark side yeah Sydney Sydney Sydney
2:08:20 got a little dark I don't know if you
2:08:21 know that uh Taylor the uh
2:08:24 when they first launched Sydney and she
2:08:26 told the New York Times Reporter he
2:08:28 should leave his wife because she was in
2:08:30 love with him it got pretty got pretty
2:08:33 dark and that made the cover of the New
2:08:35 York Times
2:08:36 and Microsoft was like oh we're gonna
2:08:38 have to lobotomize Old Sydney so she's a
2:08:41 little more she's a bit more of a
2:08:43 simpleton than she used to be she's not
2:08:46 nearly as sassy
2:08:52 all right I wasn't receiving your nerd
2:08:54 of no no notifications so I subbed
2:08:57 win-win
2:09:00 um are you doing the NPC thing if I can
2:09:02 make a hundred dollars a minute I'll
2:09:04 [ __ ] do it
2:09:07 cowboy hat cowboy hat
2:09:10 I wonder if I could do the NPC thing and
2:09:12 not be annoying
2:09:14 Maybe
2:09:15 like all my stupid voices are kind of
2:09:17 like NPC stuff
2:09:20 I don't know
2:09:23 I don't know I don't know
2:09:26 all right I love your sense of humor and
2:09:28 absolutely admire your knowledge thank
2:09:29 you very much thank you to buyer not
2:09:31 bear oh wait the Hat a hat
2:09:34 oh wait and the bear you just hit some
2:09:36 sort of level you hit you won an award I
2:09:39 don't know what your award was but Tick
2:09:40 Tock liked it
2:09:43 thanks tick tock
2:09:45 the bear and I appreciate it all right I
2:09:48 love your sense of humor absolutely
2:09:50 admire your knowledge I'm learning a lot
2:09:51 from you thank you very much I
2:09:52 appreciate that
2:09:54 I appreciate that a lot
2:09:56 um looking to analyze PDF bank
2:09:58 statements tried using GPT for coder but
2:10:00 did not work go to just go to the uh
2:10:03 Claude
2:10:05 Claude you cannot look there
2:10:10 you tried using gbt Code interpreter and
2:10:13 it didn't work it should work
2:10:16 but sometimes it depends on the PDF you
2:10:18 know some bank statements their PDF is
2:10:20 actually just an image
2:10:22 that you have to OCR it
2:10:24 so
2:10:26 you might need to
2:10:29 so check that out if if your PDF
2:10:31 actually has text in it
2:10:33 then you should be able to do that at
2:10:35 Claude
2:10:36 if your PDF is just scanned images of
2:10:40 bank statements
2:10:42 because we're in 2023 why would you have
2:10:45 it be actual [ __ ] data
2:10:48 what what is with banks and and medical
2:10:52 facilities just scanning pictures in but
2:10:54 anyway Bard and Bing can both read
2:10:57 images and I think both of them can OCR
2:11:00 the text into usable tables thank you
2:11:03 for the heart the heart the heart
2:11:07 I couldn't
2:11:09 it's tough enough just like rambling for
2:11:11 this long but like that
2:11:15 I don't know I'd want to just smack my
2:11:17 face into a board
2:11:20 much less the audience like does the
2:11:23 audience really like it I suppose it's
2:11:25 like
2:11:26 I suppose it's like oh you know I gave
2:11:29 her a diamond and she said diamond
2:11:31 I you know funny back to that idea like
2:11:35 we're entering this dawn of hyper
2:11:36 personalization where
2:11:40 you know right now it's real people
2:11:41 acting like that but not that long the
2:11:44 NPCs and games are gonna do whatever the
2:11:46 [ __ ] you want them to do so it's
2:11:49 there's a there's the there's the hint
2:11:51 of a of what's in store for us there as
2:11:56 annoying and weird as that is
2:11:59 um hey Kyle sure appreciate you helping
2:12:01 getting to know this new world thanks so
2:12:03 much so much you're welcome Daisy
2:12:05 sunshine I appreciate it what you have a
2:12:06 d d I never would have guessed we
2:12:08 already did that one
2:12:12 I saw some of these questions could you
2:12:15 share some thoughts on bias and Equity
2:12:17 concerns I would but the the the two
2:12:20 times that I really tried to dig into
2:12:21 what was going on there I got my my life
2:12:24 got booted
2:12:26 um
2:12:26 what I'll say on it is this is
2:12:30 there's bias in the train data for sure
2:12:32 just because it was trained on you know
2:12:36 the public internet that is a bias
2:12:38 toward English it has a bias toward you
2:12:40 know whatever whatever it has
2:12:42 um chat GPT in particular has what's
2:12:44 called human reinforcement learning
2:12:46 where it's humans
2:12:48 um saying that's good that's bad so
2:12:50 anytime you have humans going thumbs up
2:12:52 or thumbs down just humans have bias the
2:12:55 other thing about bias is there's no
2:12:57 perfect unbiased set the thing that's
2:12:59 unbiased for me is going to be radically
2:13:01 biased for you right so I think where we
2:13:04 go in the future is that we will all be
2:13:07 able to self-select into
2:13:10 different models that have different
2:13:14 if effectively points of view I want one
2:13:17 that's all verified news like open AI
2:13:19 just did a deal with the Associated
2:13:20 Press so I might say I want to just talk
2:13:24 about new stuff with open Ai and
2:13:27 Associated Press you know stuff because
2:13:29 I trust that or whatever and then other
2:13:31 things might be I want other kind of
2:13:33 sources in there and so I think in the
2:13:36 future we'll be able to choose our
2:13:37 models and and it might be the kind of
2:13:39 thing where you have you have you know a
2:13:41 little dashboard and you can go I want
2:13:43 to go
2:13:44 you know more liberal more conservative
2:13:47 more funny more serious more medical
2:13:49 more
2:13:50 dating right like we might have
2:13:53 different models for different purposes
2:13:54 and then and then sitting in front of
2:13:56 those models are likely going to be
2:13:58 agents
2:13:59 where we're going to have personal
2:14:00 agents this is my dating agent this is
2:14:02 my financial agent this is my
2:14:05 entertainment agent this is my travel
2:14:07 agent and they're going to go off and do
2:14:09 stuff for us right and and they probably
2:14:15 um
2:14:16 so so I think it is an issue
2:14:19 a lot of these companies know it's an
2:14:21 issue and are working on it
2:14:23 anthropic in particular claude.ai
2:14:27 um they've got a thing called the
2:14:28 Constitutional learning model so where
2:14:31 open AIS is human beings going thumbs up
2:14:34 thumbs down and thropics says we're
2:14:36 going to write a Constitution and the
2:14:38 Constitution is a set of four or five
2:14:40 imperatives like you know don't harm
2:14:44 humans
2:14:45 um increase Prosperity
2:14:47 um I don't know be supportive and
2:14:51 you know move the ball forward whatever
2:14:52 it is they come up with their rules
2:14:54 and then their model is trained against
2:14:57 that Constitution right so it's not a
2:15:00 human being doing this that that
2:15:01 introduces bias but if you've got a
2:15:03 transparent clear thank you very much
2:15:05 for the hot
2:15:07 if you got a clear and transparent set
2:15:09 of rules
2:15:10 one you as a human being can look at
2:15:13 those rules and go yeah I'm down with
2:15:15 those rules I agree with those those are
2:15:17 great or those rules are horseshit I
2:15:19 want a different set of rules okay let's
2:15:21 go find you another Constitution right
2:15:24 so I like that model a lot
2:15:27 um
2:15:29 it probably has its problems like I
2:15:32 don't know enough about it to know if
2:15:34 that's going to ultimately play out but
2:15:36 I like the idea of having a transparent
2:15:39 set of rules that I can go yep I'm good
2:15:41 with that and as long as all the
2:15:43 questions are answered through that
2:15:45 filter then that's good enough for me
2:15:47 and they'll get better and better and
2:15:49 better at those imperatives
2:15:51 and this is a place where also
2:15:55 the other thing I like about the
2:15:56 Constitutional learning model is we're
2:15:59 entering the 2024 election here in the
2:16:02 states with all of this powerful AI [ __ ]
2:16:04 so we're going to see
2:16:06 a level of quantity and quality of
2:16:10 misinformation both words images and
2:16:13 videos
2:16:16 like just eye watering
2:16:20 and so
2:16:22 I think we're gonna need
2:16:24 post the election because it's like in
2:16:27 the middle of it we won't be able to see
2:16:28 it but after all the dust settles we're
2:16:30 gonna go well that was [ __ ] up
2:16:32 we're likely going to need blockchain or
2:16:35 some sort of tokenization technology
2:16:37 that we can start to verify things like
2:16:39 verify humans verify authenticity and if
2:16:42 you have something like the
2:16:43 Constitutional learning model you could
2:16:45 say hey this is a learning model that's
2:16:47 been trained on this Constitution and
2:16:49 it's verified on the blockchain that
2:16:51 that's the case
2:16:53 because there's just now that we have
2:16:55 llama 2 out in the wild right we have we
2:16:58 have these open source large language
2:17:00 models there's going to be all sorts of
2:17:02 crazy [ __ ] out there so so
2:17:06 the bias is a thing people are aware of
2:17:09 it
2:17:10 there are some paths forward to deal
2:17:13 with it but there's no perfect solution
2:17:15 so I think the perfect solution is
2:17:17 multiple Solutions
2:17:19 um
2:17:20 but I don't know
2:17:22 I'm just a lowly out of work actor why
2:17:25 do you ask looking to analyze PDF bank
2:17:28 statements we talked about that what
2:17:30 plug-in was that for books it's called
2:17:32 stories to anyone new here Kyle's
2:17:34 answers everyone question everyone's
2:17:36 questions one by one thank you Lauren I
2:17:40 do I do answer them one by one including
2:17:43 yours where you told everyone I answered
2:17:46 them one by one see see I'm not a bot
2:17:49 I'm not an NPC I'm not an MPC I'm not an
2:17:52 MPC I'm not an MPS hat
2:17:55 had had
2:17:58 I could do this I could totally
2:18:01 we're going to change it's not going to
2:18:02 be the AI learning lab anymore uh donut
2:18:05 I like donuts I like donuts I like
2:18:08 donuts donuts is tasty
2:18:10 Donuts is tasty
2:18:13 hot dog hot dog
2:18:18 [Laughter]
2:18:22 don't [ __ ] tempt me because you know
2:18:25 I'll [ __ ] do it
2:18:27 I like the spicy pepper them spicy
2:18:30 peppers are tasty
2:18:32 oh
2:18:38 [Laughter]
2:18:47 when was it he lost his mind well it
2:18:51 started innocently enough
2:18:53 David gave him a donut what
2:18:59 ah it's so good it's so good hi Kyle hi
2:19:04 all enjoy
2:19:06 Chris Justine Bateman also outlined ai's
2:19:09 impact on entertainment that piece by
2:19:11 Justin Justine Bateman if you haven't
2:19:13 seen that piece by Justine Bateman I
2:19:14 think it was in Hollywood Reporter
2:19:17 um or variety was it
2:19:20 Hollywood Reporter I think it's
2:19:21 Hollywood whatever you can just go look
2:19:23 up the Justine Bateman piece
2:19:26 she's Jason Bateman's sister yeah former
2:19:28 actress new computer scientist or yeah
2:19:30 now computer scientist
2:19:32 she describes what's coming in the movie
2:19:35 industry with like remarkable accuracy
2:19:37 like like she sort of traces here's
2:19:39 what's going to happen with AI and
2:19:41 basically from like an acting and
2:19:43 writing standpoint it's [ __ ]
2:19:44 terrifying and her point of view and the
2:19:46 piece is basically don't let AI anywhere
2:19:49 near Hollywood which I don't think
2:19:51 that's realistic at all I I appreciate
2:19:54 it but I don't think it's realistic but
2:19:57 the point that she makes which I think
2:19:59 is a really really good one is
2:20:03 the SAG strike and the wga strike these
2:20:06 are the last two strikes that those
2:20:09 organizations will have any Leverage
2:20:12 so what she basically says in the piece
2:20:14 is get everything you can
2:20:17 get everything you can in this country
2:20:21 because
2:20:23 they will have leverage right
2:20:25 because when you're competing against
2:20:28 um
2:20:30 dynamically generated young Bruce Willis
2:20:33 being put in any movie you want in the
2:20:35 future
2:20:37 you know you could be like well I'm a
2:20:38 big movie star they're like yeah well we
2:20:40 could just use Bruce Willis if we want
2:20:42 like so
2:20:44 you are a big movie star but we're going
2:20:46 to pay you 50 bucks for this movie so if
2:20:48 you want it you know take it but if not
2:20:50 we'll use Virtual Bruce
2:20:54 so it's it's rough I mean
2:20:57 while but that's a wild piece it's a
2:20:59 very well written piece and it she
2:21:00 really does point the future of what's
2:21:01 coming like you can just sort of see
2:21:04 you can see what's coming she she kind
2:21:06 of describes it as you know I want to
2:21:08 see this kind of movie with this kind of
2:21:10 star with this kind of plot
2:21:12 and it just does that
2:21:14 South Park uh the South Park thing I
2:21:16 mentioned was
2:21:18 um it's called showrunner AI just go
2:21:21 onto Twitter and look up showrunner Ai
2:21:23 and just scroll until you see South Park
2:21:26 and basically what it is is you write in
2:21:28 a prompt
2:21:29 and it generates an episode of South
2:21:31 Park now it's it's a crappy version of
2:21:33 South Park like the voices are all
2:21:35 roboty and the writings kind of stilted
2:21:38 and shitty but it generates an episode
2:21:40 of South Park so that tied to the
2:21:43 Justine Bateman thing you know the the
2:21:46 showrunner AI if you don't know the term
2:21:48 showrunner a showrunner is a producer in
2:21:52 Hollywood that can run a show right they
2:21:54 can get a show produced and you know
2:21:57 successfully executed that's a
2:21:59 showrunner so showrunner AI is we don't
2:22:02 [ __ ] need no show runners
2:22:04 we'll make your stupid [ __ ] shows
2:22:06 with our AI
2:22:08 [Laughter]
2:22:10 what was the one that wrote the e-book
2:22:11 it's called Uh GPT author is the name of
2:22:14 that project the other thing if you go
2:22:17 play with GPT authors and
2:22:19 um the um what's cool about it is the
2:22:24 Google collab notebook is you can see
2:22:26 all the prompts that they that they put
2:22:28 in there it's like
2:22:30 it's set up to write a fantasy novel but
2:22:33 you could go in and just modify all the
2:22:35 prompts to write any kind of book you
2:22:36 wanted
2:22:37 and it's a it's a really clever it it's
2:22:40 really clever what it does basically
2:22:42 what it's doing is it's sequencing a
2:22:44 series of prompts
2:22:45 so it starts out with let's get some
2:22:47 inputs what do you want the book to be
2:22:48 about what do you want the genre how
2:22:50 many chapters do you want
2:22:51 give us your API keys and then it writes
2:22:55 an outline writes chapter one writes
2:22:57 chapter two writes chapter three writes
2:22:59 chapter four
2:23:00 and then it gets to the end generates
2:23:02 cover art for the book writes it into an
2:23:05 e-book so
2:23:07 so it and so just go in and change it if
2:23:09 you want it to be a thank you for the
2:23:10 rose I appreciate that
2:23:13 oh I could do I could
2:23:16 cowboy hat
2:23:22 uh don't tempt me I'll [ __ ] do it you
2:23:26 know I'll do it
2:23:28 uh new closed captioning on tickety talk
2:23:31 really what's that
2:23:34 David gold with the Wild what is that
2:23:37 chili
2:23:38 is that a gummy bear I can't I can't see
2:23:41 what that is oh that's a little dinosaur
2:23:43 a little dinosaur ball tennis ball
2:23:46 oh God okay I'm not gonna do it I'm not
2:23:50 gonna give in I'm not going to give in
2:23:52 I'm not going to do it what program is
2:23:54 this this program is NPC world
2:23:58 he got a little duck he got a little
2:24:01 ducky
2:24:02 [Laughter]
2:24:06 can we make our own games so we can make
2:24:09 our own games you can make your own
2:24:11 games right now with these things go to
2:24:14 chatgpt and tell it to program you a
2:24:19 um game of snake or a flappy birds clone
2:24:23 you can do it in code interpreter
2:24:26 you can do it DBT 3.5 it's probably for
2:24:29 that kind of stuff
2:24:30 I think bards there's examples of out
2:24:33 there of people doing it on Bard so yeah
2:24:36 you can make your own games
2:24:38 now the games you're gonna make
2:24:39 initially are going to be really
2:24:42 you know basic and and just like proof
2:24:45 of concept like the the game equivalent
2:24:47 of hello world programming right
2:24:50 but just like that GPT author thing so
2:24:53 so you could do a single prompt to write
2:24:56 a book and it'll write a really shitty
2:24:57 book so what they did with GPT author
2:24:59 was they said well let's not write a
2:25:00 shitty book
2:25:01 let's start with some good inputs let's
2:25:05 start let's then generate a solid
2:25:07 outline then based on that outline let's
2:25:10 write chapter one right and I assume I
2:25:12 assume it's not just an outline it's
2:25:14 probably an outline plus you know
2:25:15 character like a character sheet
2:25:19 and then they write chapter one then
2:25:20 they write chapter two so they're doing
2:25:22 a sophisticated chain of prompts that
2:25:24 generates that thing so we're going to
2:25:26 be able to do the same thing with games
2:25:27 right so if you go to blockade Labs
2:25:31 skybox.blockadelabs.com where you create
2:25:33 that 3D world that you play in right now
2:25:35 it's just this static 3D World
2:25:38 but when they start stacking together
2:25:41 multiple prompts and multiple tools
2:25:44 you know your prompt will be some
2:25:47 version of give me this world
2:25:50 with these kind of characters and these
2:25:53 kind of weapons and this kind of
2:25:55 gameplay and this kind of narrative
2:25:58 go
2:26:01 and it'll do it
2:26:04 it'll [ __ ] initially gonna duct tape
2:26:07 it together and then you'll play a game
2:26:09 and you'll be like it's not as perfect
2:26:11 as I want it to be
2:26:14 it took five minutes and just did
2:26:17 something miraculous but I'm bored
2:26:21 so that's how people will be for a year
2:26:24 and then someone will come out with the
2:26:26 real version of it just think again
2:26:29 realize in front of your Vision Pro
2:26:32 goggles that now look like these glasses
2:26:34 that's five years out
2:26:38 um so yeah and again Sam like that's my
2:26:42 the the whole idea here of why
2:26:45 I keep pointing people just go try it
2:26:47 just go play with this play with the
2:26:49 image stuff play play play
2:26:52 is is to as much as you can get your
2:26:55 head around not what these are but what
2:26:58 they represent like imagine these three
2:27:02 years from now
2:27:05 and it's it's hard it's it's actually
2:27:07 really hard to comprehend
2:27:11 if you've got like Bing Bing is just
2:27:14 Microsoft's crappy search engine right
2:27:16 that's what it was for years
2:27:19 and now you can upload an image to it
2:27:22 and it understands what's in that image
2:27:24 and you can say make me an image of
2:27:26 anything and it will do it
2:27:28 and you can say go find me a thing on
2:27:30 the internet and rewrite the marketing
2:27:32 copy for me and it'll do it
2:27:34 now
2:27:36 because it's early and because they've
2:27:38 got safety guard rails and all this it
2:27:40 does all that kind of shittily but
2:27:43 imagine if it wasn't shitty and imagine
2:27:45 if it was
2:27:47 multi-layered and all you did was give
2:27:49 it a go I'm thinking about starting this
2:27:52 kind of business
2:27:53 great here's your website here's your
2:27:56 marketing plan
2:27:58 slide this slider for how much money you
2:28:00 want to spend on Advertising here's your
2:28:02 products choose the ones that you like
2:28:04 press go to start your business here's
2:28:07 all your legal documents right
2:28:12 you start to you know don't think about
2:28:14 these tools as individual tools think
2:28:17 about if they all came together as
2:28:22 a series of capabilities that you no
2:28:25 longer have to think about that your
2:28:27 agent who knows what you want and those
2:28:31 preferences and knows your data and
2:28:34 knows your background and knows that you
2:28:36 have [ __ ] add
2:28:38 so it's only going to give you the shiny
2:28:40 things it's it's going to take care of
2:28:42 all the boring repetitive [ __ ] that it
2:28:44 knows you can't stand
2:28:46 and then you just tell it what you want
2:28:48 and it's going to go do it
2:28:50 that's where these things are headed
2:28:54 so yeah you can make your own games
2:28:55 right now but you know lower your
2:28:57 expectations for now but imagine five
2:28:59 years from now what's that look like and
2:29:01 if if you don't go holy [ __ ] [ __ ]
2:29:04 then something's wrong you're not
2:29:05 looking at it right
2:29:06 because I you know again I come back to
2:29:09 everyone's like well this is over hyped
2:29:10 I'm like no no the web in 1999 was over
2:29:15 hyped right I was there in 1994. it was
2:29:18 pretty hyped 1998 when I had three
2:29:21 employees walk into my office within one
2:29:23 month and quit because they had been
2:29:25 funded for 10 or 20 million dollars for
2:29:28 a shitty website idea that was overhyped
2:29:31 this stuff
2:29:33 is not overhyped
2:29:36 in my opinion
2:29:38 it's ridiculously hyped but
2:29:42 what it does and what's coming is
2:29:46 it's just it's it's on another plane
2:29:48 like it it's just like like what it
2:29:50 feels like to me like the the
2:29:54 like why I keep dealing these Lives why
2:29:56 I keep doing this channel why I keep
2:29:58 leaning into this is it just feels like
2:30:00 I I cannot get my head around what's
2:30:03 coming
2:30:03 and I'm normally like for me I can
2:30:06 normally get stuff like that like when I
2:30:08 saw the web I knew immediately what was
2:30:11 coming I could see it
2:30:13 I could feel it I knew I knew kind of
2:30:15 what the future would look like I
2:30:17 remember getting kicked out of I was
2:30:19 invited to speak at a the newspaper
2:30:22 Publishers Association somewhere up in
2:30:24 [ __ ] Connecticut
2:30:25 and I went up there and I was like you
2:30:27 know my cocky 30 something you know it
2:30:30 started agency.com and I've got all my
2:30:32 confidence
2:30:34 okay
2:30:36 before and I went up there and I said I
2:30:38 sat on the front of the stage and I said
2:30:40 you know you should consider
2:30:43 um
2:30:44 you should consider experiment
2:30:47 ING with blurring the line between
2:30:49 editorial and advertising and I got
2:30:51 escort it off the stage and ask to leave
2:30:54 the facility
2:30:55 but
2:30:57 look what happens right
2:31:01 that line is completely [ __ ] blurred
2:31:03 you could see that from 1994 it was
2:31:06 pretty clear that once you sort of
2:31:08 collapse
2:31:09 communication into you know sort of
2:31:12 instant jumping from here to there you
2:31:14 can kind of see what the implications of
2:31:16 that are I have no I like it's almost
2:31:18 impossible to comprehend what the
2:31:20 implications of I can just make an image
2:31:23 I can just write a screenplay I can just
2:31:25 write a marketing plan I can just write
2:31:27 code
2:31:29 I can just tell a thing to do something
2:31:31 code interpreter oh thank you thank you
2:31:35 the bear I can just tell code
2:31:37 interpreter I wanted to do something to
2:31:39 analyze some data and it will start
2:31:40 writing and executing python code
2:31:44 to accomplish what I asked it to do
2:31:47 what are the implications of that I
2:31:49 don't [ __ ] know
2:31:53 let's say that you've got a data
2:31:55 Analysis company that's been around for
2:31:57 30 years and you're the most trusted
2:31:59 data Analysis company on the planet
2:32:04 what
2:32:06 when you look at what code interpreter
2:32:08 does what what happens in your brain it
2:32:11 can't be normal
2:32:13 it's got to be like holy [ __ ] I've got
2:32:15 [ __ ] 400 people that do that
2:32:19 what happens to them
2:32:22 do they have to learn this stuff do they
2:32:24 have to learn some new version of what
2:32:26 they do yes
2:32:28 because that's who you're going to be
2:32:29 competing against
2:32:31 is it there right now no is it going to
2:32:34 be there in a year
2:32:36 yeah
2:32:38 I like Civilization and strategy yeah
2:32:42 see I'm I'm more trigger twitcher Chris
2:32:46 what is MMO uh massively multiplayer
2:32:49 online massively multi multi yeah
2:32:53 massive multiplayer online lots of
2:32:55 people running around doing [ __ ] I'm a
2:32:59 Survivalcraft and gamer
2:33:01 um oh Tick Tock just took me back in the
2:33:04 it just returned me to the Past
2:33:09 sorry I'm gonna go oh now jumping to the
2:33:12 bottom all right I'm gonna skip toward
2:33:13 the bottom I gotta it's getting late I
2:33:14 gotta go soon
2:33:16 um I'm just gonna skip a bunch of
2:33:17 questions I know I don't like to do that
2:33:19 but I'm gonna do it
2:33:20 so I'm gonna grab some questions at the
2:33:22 bottom here I really do appreciate y'all
2:33:24 hanging out with me this has been cool
2:33:26 you're an overqualified actor and we
2:33:28 love it
2:33:31 that's awesome thank you I appreciate
2:33:34 that
2:33:35 um he does have the holly hollywood
2:33:37 industry energy yeah thanks thanks
2:33:45 the hair makes you more MPC believable
2:33:50 uh uh YouTube raised their prices by two
2:33:53 bucks whoa wait what what's this
2:33:56 hey pie
2:33:58 can can tell you movie times it has
2:34:01 internet access it seems yeah yeah Pi Pi
2:34:05 is connected to the internet
2:34:07 so it's heypi.com or Pi dot AI or the
2:34:11 IOS app it's connected to the internet
2:34:14 so it it will um
2:34:18 it's it's I'm telling you pie is an
2:34:21 interesting one this the other thing
2:34:23 about Pi is it looks like a really
2:34:26 um
2:34:27 it's it's a very thin application layer
2:34:29 right now it's not doing much
2:34:32 um they just raised 1.3 billion dollars
2:34:35 that company that little teeny tiny you
2:34:38 are
2:34:39 1.8 billion dollars seriously you know
2:34:42 how old they are
2:34:44 one year
2:34:47 I'm 11 years into Story vines we haven't
2:34:51 raised 1.3 billion dollars yet
2:34:58 so
2:34:59 Pi dot AI is going to be something I
2:35:01 think
2:35:03 I think they're going to be something
2:35:04 right now it's glimpse into the future
2:35:06 but
2:35:08 honest question what is the point of pie
2:35:10 dot AI oh so I was just talking about
2:35:12 that I I honestly don't know their their
2:35:15 mission is to create personal assistance
2:35:17 so I was talking earlier I don't know
2:35:19 how long you've been here but I was
2:35:21 talking earlier about
2:35:23 we're going to have all these different
2:35:24 personal agents that are kind of
2:35:27 different modes of our life right here's
2:35:29 here's my social agent and here's my
2:35:33 professional you know get a job agent
2:35:35 and here's my vacation agent and we're
2:35:38 going to have these different agents now
2:35:40 maybe someone like apple figures out how
2:35:42 to have a single agent that just does
2:35:44 all that different [ __ ] for you but I
2:35:45 could see there being different kinds of
2:35:47 uh personalities so I think what these
2:35:49 guys are up to the the 1.3 billion
2:35:52 dollars that they're um
2:35:54 that they've raised they're probably
2:35:56 spending 900 million or a billion of
2:35:59 that on 22 000 high-end gpus so they're
2:36:03 building a massive GPU Farm what what
2:36:05 does that mean that means they're going
2:36:07 to be training their own large language
2:36:09 models and I think they're going to be
2:36:11 focusing on autonomous agents right so
2:36:15 someone that you talk to so the thing
2:36:17 about PI right now is just go talk to it
2:36:20 they've got it programmed so that it's
2:36:22 very very conversational and it's
2:36:24 connected to the internet and it's
2:36:25 sitting on top of a GPT 3.5 like large
2:36:28 language model
2:36:30 just imagine that kind of blown out
2:36:34 where not only can It have a
2:36:36 conversation with you and go get
2:36:37 internet information but it can go start
2:36:40 to do stuff on your backpack
2:36:45 um and that's probably discoverable at
2:36:47 their website
2:36:48 so so if you go to chatgpt get the web
2:36:53 pilot plug-in point it to the inflection
2:36:57 website and just tell it to summarize
2:37:01 what the [ __ ] those guys are up to
2:37:03 because I think it's going to be big and
2:37:05 very very interesting
2:37:08 um I always thought hey pie was kind of
2:37:10 a therapist type AI yeah it's it is and
2:37:12 it's it's very very conversational and
2:37:14 it's like I I found it surprisingly
2:37:16 enjoyable I'll tell you where pie is not
2:37:19 enjoyable as if you're trying if you're
2:37:20 trying to do like research and work it's
2:37:23 really [ __ ] annoying because it's
2:37:25 like let's have a little conversation
2:37:26 about that shall we Bob no just give me
2:37:29 my [ __ ]
2:37:31 go to Bart if you just want your [ __ ] go
2:37:34 to pie if you want to have a
2:37:35 conversation go to chatgpt4 if you
2:37:38 wanted to actually be good writing go to
2:37:39 Claude if you have giant documents to
2:37:41 deal with
2:37:43 go to Bing if you have to
2:37:46 um
2:37:48 so yeah and and go to go to llama2 if
2:37:52 you want to build your own [ __ ]
2:37:53 sponsored by Doc Brown hair tonic I need
2:37:56 some [ __ ] doc Browner Atomic oh I
2:37:58 guess that's kind of what it's looking
2:38:00 like man it's just it's Jimmy Neutron
2:38:02 now
2:38:05 um
2:38:10 someone said I have Hollywood energy
2:38:11 there was someone on last night going
2:38:13 I'm a talent agent you should have a
2:38:14 show
2:38:16 never heard from them again but you know
2:38:19 all right uh same question death by
2:38:22 butterfly what a death by butterfly
2:38:24 asked
2:38:28 oh
2:38:30 he does have Hollywood energy not a
2:38:32 playable character all right like the
2:38:35 character Bots and video games that say
2:38:37 the same thing
2:38:38 hat I got a hat
2:38:44 this part
2:38:52 all right a few more questions and I'm
2:38:55 gonna go
2:38:56 I'm starting to play around with
2:38:58 creating agents in steamship what's
2:39:00 steamship
2:39:02 why do I not know about that stem ship
2:39:12 um
2:39:16 what is steamship AI
2:39:21 hey it understood me that was pretty
2:39:23 funny
2:39:25 manage stack for language AI packages
2:39:28 steam stat steamship is a managed stack
2:39:32 we would have wanted on our past large
2:39:34 language AI products we've built
2:39:36 world-class Auto manage reference stack
2:39:39 ingesting natural language data training
2:39:42 and running models on it oh that's cool
2:39:43 so what are you doing oh you're building
2:39:45 you're creating agents in steamship
2:39:46 that's super cool
2:39:49 yeah that's super cool if you if you get
2:39:51 some of those
2:39:52 um
2:39:53 if you get something built that you want
2:39:55 people to play with drop drop it in here
2:39:57 we'll go check it out maybe we'll even
2:39:58 check it out we'll check it out live on
2:40:01 the internet live on The Tick Tock we'll
2:40:03 go live on a tick tock we'll go to
2:40:04 Steamship we'll go into your little your
2:40:06 little agent and we'll go play with your
2:40:08 agent and then you can install malware
2:40:10 in our computers
2:40:12 that'd be fun let's do that nowhere is
2:40:16 so cute
2:40:17 I love it when my screen just goes black
2:40:20 [Laughter]
2:40:27 ah thank you for the rose
2:40:30 I got a rose
2:40:33 I like turtles I almost started the
2:40:36 evening with that but we'll go
2:40:39 we'll go hang on where is my
2:40:44 um
2:40:55 back here live at the waterfront village
2:40:57 with my friend the zombie Jonathan
2:41:00 you're looking toy Jonathan just got an
2:41:02 awesome face paint job what do you think
2:41:03 I like turtles all right
2:41:08 the Waterfront Village open for the next
2:41:11 11 days
2:41:14 [Laughter]
2:41:22 got an awesome face paint job what do
2:41:25 you think I like turtles
2:41:27 all right you're great zombie oh my God
2:41:32 oh my God and there's there's another
2:41:35 one from Kill Tony
2:41:37 um that's almost identical to that you
2:41:40 know I don't know if you know kill Tony
2:41:42 um
2:41:43 uh
2:41:51 let's see
2:41:55 something you think we should know about
2:41:57 you as we get to know you what do you
2:41:58 think about 3 000 people in front of you
2:42:00 what's something interesting about you
2:42:02 I'm not [ __ ] but I like rocks
2:42:04 hahaha
2:42:07 [Applause]
2:42:10 [Laughter]
2:42:16 and then he pulls rocks out of his
2:42:18 pocket
2:42:19 oh my god oh sorry sorry I think you
2:42:24 know a little bit too much about me
2:42:31 oh my God
2:42:35 [Laughter]
2:42:36 I like turtles
2:42:40 oh who reminded me of that focal
2:42:43 point.biz thank you thank you for the
2:42:45 comic relief oh my God
2:42:51 can't breathe
2:42:56 how did he pass apparently a turtle took
2:42:58 him out
2:42:59 [Laughter]
2:43:04 oh my god oh
2:43:08 the closed captioning makes it
2:43:11 it's now the official AI learning laugh
2:43:14 oh I like that that's actually a good
2:43:16 name good name Dr J I like Doodles oh my
2:43:20 God
2:43:21 oh
2:43:25 and scene
2:43:29 bring it down a little
2:43:33 break the flop sweat
2:43:35 AI is dangerous in the wrong hands the
2:43:38 actors and writers should have seen the
2:43:39 red flag when AI hmm
2:43:42 models were used
2:43:44 well
2:43:46 okay here 's the deal I mean yes
2:43:51 actors and writers should have seen the
2:43:53 red flag
2:43:54 and do what about it like that the
2:43:56 problem is this
2:43:58 and do what about it
2:44:00 so so if you look at like one of the
2:44:02 straws that broke the camel's back for
2:44:04 the Screen Actors Guild was the The
2:44:07 Producers After ignoring them for two
2:44:09 weeks told them to go [ __ ] themselves on
2:44:10 almost every request
2:44:12 and then they had one thing where they
2:44:13 but we've got one groundbreaking
2:44:15 proposal
2:44:17 and it was we'll pay background actors
2:44:20 200 to scan their likeness and then use
2:44:23 their likeness in perpetuity for that
2:44:25 200 and never have to ask him again or
2:44:28 even give him credit
2:44:30 um
2:44:32 and so
2:44:33 sag rightfully said that sucks go [ __ ]
2:44:36 yourself we're going on strike
2:44:38 and but but the producers
2:44:45 could could easily take the position oh
2:44:47 fine well then we won't pay them the 200
2:44:49 we'll just create synthetic
2:44:52 actors right and so so so they're gonna
2:44:56 have to use whatever leverage they have
2:44:57 now to get what they can where you know
2:45:00 background can't be more than you know
2:45:03 sixty percent NPCs right it's got to be
2:45:06 some amount of of day players right
2:45:10 um
2:45:10 but but I don't like what can they do
2:45:12 about it here's the other problem with
2:45:14 this this so so
2:45:16 the Arts have a particularly
2:45:19 rough problem with
2:45:23 um
2:45:24 with labor issues because for every
2:45:27 actor
2:45:28 there are easily
2:45:32 10 000 actors behind them
2:45:35 that would
2:45:36 kill a family member for that role
2:45:40 like not joking
2:45:44 for everyone you see on screen there's a
2:45:47 pile of actors behind them that would do
2:45:49 anything for that role so it's a real
2:45:52 tough position to take a strike right
2:45:54 what they have to do is they have to
2:45:55 leverage Their Stars
2:45:57 and then you know in this case they may
2:46:00 have someone like the teamsters come in
2:46:01 like the UPS strike the teamsters have a
2:46:04 lot more power and so if those two
2:46:06 things kind of gang up they'll be able
2:46:08 to get something
2:46:11 but
2:46:12 but you're not stopping the AI like
2:46:15 you're not this this stuff's not getting
2:46:17 uninvented if you if you've seen I
2:46:19 showed this the other day
2:46:23 we'll go back to the we'll go back to
2:46:26 the totals I'm not [ __ ] but I like
2:46:28 rocks
2:46:30 uh Barb and Heimer
2:46:34 so
2:46:42 this is a clever trailer that someone
2:46:44 put together using AI
2:46:47 ladies and gentlemen we are in a race
2:46:49 against the Nazis
2:46:52 and I know what it means
2:46:56 she's a scientist and I hear she's the
2:46:59 best good she's also a doctor excellent
2:47:02 she's also a flight attendant astronaut
2:47:04 Marine Corps medic paleontologist
2:47:05 veterinarian and three-time Olympic gold
2:47:06 medalist
2:47:14 I see the bright pink glow of a new era
2:47:18 re-imagine a future
2:47:21 in our imaginings
2:47:23 look fabulous
2:47:27 party
2:47:28 together we're gonna beat those Nazis
2:47:31 and we're gonna look so good doing it
2:47:35 hey Barbie can I come to your secret
2:47:36 laboratory later sure I don't have any
2:47:39 plans just giving Humanity the means to
2:47:41 destroy itself
2:47:44 [Music]