aiLL 10/18/23 - Unleashing AI's Creative Potential with ChatGPT, DALL-E & More

Description

Chapters

Transcript

0:15 hello hello
0:18 hello got
0:20 the got the Buster po Dexter hair going
0:24 tonight what's shaking what's happening
0:27 what's going
0:29 down welcome welcome welcome welcome
0:32 Matias what's happening good to see you
0:36 sir Who's Live Is It
0:40 Anyway welcome to Tik Tock live it's a
0:44 game show where you twoo can
0:48 play and if you play it right you can
0:51 make money
0:54 with Clint Copperhead what's shaking
0:57 buddy everyone in regulars and mods PR
1:00 in
1:03 do oh
1:06 man my eyes is a little crispy Hello
1:09 Apple user what shaking lunar stick good
1:12 to see
1:13 you Dr
1:20 J I taught a uh I taught the
1:25 third um the Third Leg of the AI
1:29 LaunchPad Workshop so tonight was our
1:34 the final the final leg it was Advanced
1:36 Ai
1:38 and holy
1:42 crap it was
1:44 like it was just insane like how much
1:47 how much stuff I ran them through in two
1:51 hours I basically told them just sit
1:53 back don't listen to anything just go
1:55 watch watch the recording when it's
1:59 done it was nuts but I'm a little I'm a
2:01 little
2:05 crispy Amelio's wife was there wait is
2:09 is Tobias still
2:13 banned no you're not banned are you
2:16 Tobias cuz you just posted
2:19 Turtles I think you're
2:22 good welcome everyone oh wow there's a
2:25 lot of folks in here all right so let's
2:27 get this party started so my name is
2:30 Kyle Shannon this is the AI learning lab
2:33 and uh what we do here is we talk about
2:35 this generative AI stuff and uh if you
2:38 have any questions or comments about
2:41 what you're doing with AI pop them in
2:42 the comments below I'll do my best to
2:44 answer them we'll walk through a bunch
2:47 of stuff here so I'm going to I'm going
2:48 to quickly get you up to speed on if
2:51 you're brand new to this AI stuff and
2:53 you're like but I don't know what to do
2:56 I'm not very technical I'm not a
2:57 mathematician okay all right I'm not an
3:00 engineer I I don't do programming okay
3:02 so I can't do this AI stuff no wrong
3:05 wrong wrong you can do it we're all
3:08 doing it we're all figuring it out I
3:10 have a degree in acting all right so
3:11 take that as your
3:13 Baseline um I'm also also an
3:15 entrepreneur but where we're going to
3:17 start if you have not played with chat
3:20 GPT if you haven't tried it yet you're
3:23 like but I don't know what it is Kyle I
3:24 don't know what to do I think this is
3:26 not bright enough yes
3:28 correct um
3:31 if that's you if you're like I don't
3:33 know where to start start there that one
3:35 chat. open.com that's the official chat
3:38 GPT
3:39 website it's it's the first and it's the
3:42 best and you can make money
3:45 with
3:47 uh we make fun of those people um but
3:51 anyway um go there and just start
3:53 talking to it it's it it looks like a
3:55 Google search box and but it's not tell
3:59 it ask it things ask it to write your
4:01 poems ask it to write you a business
4:02 plan ask it to write you come up with
4:05 marketing copy for your pizza shop
4:07 doesn't matter just go go there and
4:08 start asking it stuff and watch what it
4:10 does um if you've been playing with this
4:12 stuff for a while um what I've got here
4:14 is a list of the things I think are
4:16 worth paying attention to these are the
4:17 big four that are right here we call it
4:19 the adult table and
4:22 um these are all either very very good
4:26 now or about to get very very good Bard
4:28 right now is okay not great um chat GPT
4:33 now that it's gone multimodal and has
4:35 Dolly built into it is insane it's it's
4:38 amazing it's the it's the it's the best
4:41 20 bucks you you could spend on a for a
4:44 monthly subscription for
4:46 software um claud's really good Bing has
4:50 potential um but it's not great uh cuz
4:54 it's they've got the safety guard rails
4:56 too high then we've got the kids table
4:59 piie perplexity AI where you can play
5:01 with the meta meta's llama models llamas
5:05 is tasty and po um so go check those out
5:09 now for images do a screenshot of this
5:12 one so Dolly 3 um you can play with
5:16 inside chat GPT if they've given you
5:18 access to it if not it's coming if if
5:21 you pay for chat GPT plus um but you can
5:24 also do it for free at bing.com create
5:27 or in Bing chat if you ask it to make an
5:29 image that's Dolly 3 mid Journey really
5:33 good best in terms of art although uh
5:36 you have to do things in Discord
5:39 ideogram okay but fading um Adobe
5:43 Firefly from Adobe uh just went into
5:46 version two and it looks like it's
5:47 getting pretty good here are some fun
5:50 ones to play with I'm going to switch
5:52 out Runway ml for either Pika or Moon
5:55 what's it called Moon Valley I think um
5:58 for text to video
6:00 hey Jen does video translation into a
6:02 bunch of languages that are amazing um
6:05 those other things are worth playing
6:07 with too so screenshot that and then the
6:09 final little pile of these things future
6:28 pedia.com Siri go
6:32 away Siri oh I said Sher D she thought I
6:36 said Siri
6:38 D Siri go
6:41 away go
6:43 away
6:46 okay Kyle melted my face off tonight
6:49 with our last AI Workshop yeah it was
6:51 intense Ronda it was it was a lot I told
6:53 you it was a lot going into that right
6:56 so we did um so the the workshop that I
6:58 did in in Rhonda's community
7:01 was called the AI Launchpad and actually
7:03 I have a little graphic of it
7:07 here so tonight was the final
7:11 night and um and it was the first night
7:16 yeah the first night on October 4th was
7:18 overview of AI the second night was um
7:23 applying AI to your business so it was
7:24 mostly chat jpt but you know doing doing
7:27 different use cases and things like that
7:29 and then tonight was Advanced Ai and we
7:32 went through advanced chat GPT so like
7:35 all of the new features in chat GPT we
7:37 went through um music production tools
7:41 voice and animation tools video tools
7:44 and then and then I even managed to
7:47 squeeze in about 20 minutes of
7:49 automation using zapier it was just a
7:52 machine gun of like here's here's 27
7:56 rabbit holes you can run down um so it
7:59 it was good but anyway thank you for
8:00 having me Rhonda it was that was really
8:02 great really by the way really nice
8:05 Community sweet community and Rhonda's
8:07 great too um also the salon. a if you go
8:11 right there that's a community that I
8:13 started last December um the AI salon
8:16 and it's basically a group of AI
8:18 adventurers so people just trying to
8:20 figure this stuff out helping each other
8:22 Sher Dean
8:24 nice I I got my first license plate look
8:29 yay
8:30 Kyle Shannon that's sweet thank
8:34 you um but but go join the salon if you
8:37 go there there's a link to to the
8:39 Discord where you can hang out with us
8:40 and ask questions and geek out um we've
8:43 got guilds there now we got a beginners
8:46 Guild and a a Art Guild and a writing
8:49 Guild and just it's it's cool it's cool
8:52 group of people so go hang out there um
8:55 you can put it on your bike
8:58 now um so yeah go do that uh story
9:01 vine.com that's my company and yeah so
9:04 that's all that stuff so let's let's
9:06 let's uh let's get started here so what
9:08 I'm going to do is I'm going to go on
9:11 ahead and I'll start answering questions
9:14 cuz I'll take my massive intellect you
9:16 see how large my forehead is that
9:18 indicates that I've got a very large
9:20 cranial uh crater as they call it and in
9:24 my cranial crater there's there is some
9:27 uh some activity and so I'm going to use
9:31 that activity here tonight to answer
9:33 some of your questions you might want to
9:35 ask there uh what are my
9:39 qualifications all right mid Journey
9:42 just released a really big feature get
9:44 it bigger pictures nice I didn't know
9:48 that you don't pay attention for an hour
9:51 in this space and it's like good Lord
9:53 really
9:55 people let's see the announcements what
9:58 are the announcements in in mid Journey
10:00 all
10:01 right hey everyone we're releasing two
10:04 new UPS scalers today a 2X upscaler and
10:07 a 4X upscaler that's a big deal you can
10:10 see two and 4x upscale buttons under any
10:13 normal V5 or nii
10:16 job you can use it on old
10:19 jobs using slow job ID nice the upscaler
10:25 is subtle and tries to keep the details
10:27 as close as possible to the original
10:29 image may not fix glitches or issues
10:32 with the old image the 4X up scaler
10:34 costs roughly three times more GPU
10:38 minutes than the 2x
10:41 upscaler which makes
10:44 sense both upscaler work with fast
10:48 GP fast GPU time right now we do
10:52 not know if we have enough gpus yet to
10:55 enable it in relax
10:57 mode known issue isues under very rare
11:00 occasions it will drain your bank
11:03 account no it doesn't say that I just I
11:07 just do you code machine learning no I
11:12 don't um I understand it but I don't
11:15 code it I am I am so what this Channel's
11:20 about is um so there's an inflection
11:24 point that happened last November 30th
11:26 right so we're coming up on the oneye
11:28 anniversary of the launch of chat GPT so
11:31 in my opinion chat GPT marks an
11:34 inflection point where before that time
11:37 you had to be an engineer a scientist a
11:40 researcher um if you wanted to be
11:42 involved in machine learning at all chat
11:44 GPT kind of opens it up for the rest of
11:47 us very similar to how the worldwide web
11:50 did with the internet back in the mid
11:52 90s and I was a co-founder of one of the
11:54 early agencies in the mid90s company
11:56 called agency.com so I've been through
11:58 one of these Cy before with this kind of
12:00 a major
12:05 technological and the parallels are kind
12:08 of insane and this AI stuff is looking
12:12 oh I don't know significantly more
12:14 profound and happening a hell of a lot
12:17 faster so while I can't answer all the
12:20 technical stuff what I can talk to is
12:22 you know um how do you use this stuff in
12:25 your work how do you learn it why should
12:26 you learn it is it just hype um what are
12:30 the ethical implications I mean I can
12:31 talk to some of that I'm not an
12:34 ethicist but I'm an entrepreneur so my
12:37 interest is in okay now how do we use
12:39 these things now that now that we got
12:42 the people that are coding the machine
12:44 learning and they're putting it out for
12:46 the rest of us to use what do we do what
12:48 do we do what do we do what do we do and
12:52 that's the that's the million dooll
12:54 question and that's how we're all going
12:55 to figure out you can make money with J
12:58 time how we do that
13:01 um so yeah so let's go uh let's go look
13:04 at some questions some questions who's
13:07 got some questions let answer the
13:12 questions I so wish I could have been
13:14 there I invited a friend of mine that is
13:17 awesome voice holy [ __ ] I don't know
13:19 what R is Holy [ __ ] but that's cool
13:24 uh did you just start I did just start
13:26 just wanted you to know how well you're
13:28 educated us attended seminar
13:32 today and I knew
13:35 everything
13:39 yes I listen I
13:43 mean this channel why I do these every
13:47 flipping night is not for my health not
13:50 for my
13:51 sanity both of those
13:53 like um is to just keep in the convers
13:59 ation if you just keep in the
14:01 conversation with this stuff we're so
14:03 early that you don't have to know
14:05 everything you just kind of need to know
14:07 what are the major mile posts like
14:09 what's possible today like chat jpt for
14:13 example today is a significantly
14:17 different product than it was three
14:19 weeks ago like a significantly different
14:22 product like I don't even know what it
14:25 makes possible right
14:27 now it can see it can hear it can talk
14:30 it can make images it can program it can
14:33 do data
14:35 analysis and it can do all the chat GPT
14:39 [ __ ] it's [ __ ]
14:42 amazing and it's really fascinating
14:45 because right now all of those modes
14:47 they've got it programmed so they're all
14:48 these distinct modes right I I can only
14:52 assume that that's for them to do the
14:55 testing and fine-tuning that at some
14:56 point all these modes are going to do
14:58 this
15:00 and chat jpt is just going to be like
15:01 this little magic box that you ask at
15:03 anything and it does anything that's
15:05 where it's headed I think I asked is mid
15:07 Journey free mid journey is not free
15:10 they used to do like 25 free images and
15:13 then they stopped that I don't even
15:14 think they do any now I think the
15:16 minimum for Mid journey is 10 bucks a
15:18 month that's what I pay I'm about to
15:20 cancel that because of
15:22 Del um so
15:26 yeah all right a large cranial crater
15:30 with
15:33 activity two and 4x that's what I've
15:35 been Zippy to pin head hey it's me the
15:39 old Alice participant just just passing
15:41 through to say hi and thanks for the new
15:43 hobby
15:44 absolutely welcome and good luck keep
15:48 having fun keep keep being
15:51 curious came up with a new term last
15:53 night that you you gota you got to work
15:55 your curiosity
15:58 knuckle
16:01 work that Curiosity knuckle I'm new and
16:04 would like to start dabbling what is
16:06 good to play with and free um
16:09 perfect perfect
16:19 perfect so I am going to chat
16:23 GPT so just little old Chad
16:28 GPT what's going on oh I are we back are
16:32 we alive yeah I think we're alive okay
16:37 so if you pay the 20 bucks to chat gbt
16:40 which I think right now is the best deal
16:42 in software um but if you can't pay that
16:45 that's fine but if you do pay it you'll
16:46 have that little plus sign and you'll
16:48 have access to chat gp4 and all these
16:50 goodies um even without that though just
16:54 start with chat GPT and if if you have
16:57 no ideas you can just like click on
16:59 these little things these are just ideas
17:01 make a Content strategy make a plan an
17:03 itinerary give me
17:04 ideas um come up with concepts for retro
17:08 style arcade game whatever you can click
17:09 on those things
17:13 um yeah let's say let's just click on
17:15 one of these see what it does give me
17:17 five concepts for a retro style arcade
17:22 game this is odd this is normally very
17:25 very fast so it might be my internet am
17:30 I being all janky and
17:38 stuttery this does not look normal at
17:40 all oh there it's going I have a feeling
17:43 this is my browser I just woke my
17:45 machine up and and Microsoft Edge which
17:48 is my default browser has got really bad
17:50 memory problems here I'm going to I'm
17:52 going to stop this I'm gonna I'm gonna
17:53 quit Microsoft Edge we're coming we're
17:55 coming back to you Microsoft all right
17:58 listen listen Microsoft I don't know if
18:01 you know this I know you're kind of
18:02 still in startup mode and you haven't
18:05 you haven't really released any software
18:07 before but if you could go on ahead and
18:09 fix your Microsoft edge browser so that
18:12 it doesn't hose my [ __ ] Macintosh
18:14 every time I wake it up from sleep that
18:16 would be
18:17 swell that' be super swell there
18:21 Microsoft with
18:22 your your band of buddies up there in
18:25 Redmond Washington all right okay
18:29 back to this uh let's see show me a code
18:33 s show me a code snippet of a website
18:37 sticky header I don't even know what
18:38 that is but let's do that boom there
18:41 it'll write you code this is free so
18:44 chat GPT will write you code uh we could
18:47 also say
18:56 um oh there we go okay
18:59 okay that's great could you also turn
19:02 that into a country
19:06 song well I'm not a songwriter but I'll
19:08 give it a try so now it's writing a
19:10 country song about HTML in HTML we got a
19:14 header so fine with navigation links
19:16 like a neon sign oh yeah um it can do
19:21 that it can let me create a new chat
19:24 here let me go
19:26 um uh let's see
19:30 uh write me a 10 tweet tweet
19:35 thread with Emojis and sequential
19:40 numbering talking about chubby black and
19:43 white
19:46 dogs 1 2 3 4 five 6 7 8 n 10 there's
19:54 your Tweet thread you want social media
19:56 content you got social media content so
19:58 so
19:59 where should you start just start with
20:00 chat GPT chat. open.com and just start
20:03 playing run down as many rabbit holes as
20:06 you can and just start trying to try to
20:09 stump it try to try to figure out where
20:12 the boundaries are and what you quickly
20:13 figure out is that the boundaries are
20:16 very deep and what it can do is is is
20:20 quite profound um and then you can start
20:23 exploring other things I've got if you
20:25 want to take some quick screenshots
20:28 there's four you can play with there
20:32 there's three more
20:36 there there's the image
20:40 tools there's the video tools and these
20:43 are just the ones that I think are worth
20:44 spending your time on these are just
20:46 like personal
20:48 editorial so there's lots and lots to do
20:51 why do you use Microsoft Edge on a Mac
20:54 cuz I'm a dumb dumb so here's why Cory I
21:01 cause I'm an OT CA I'm an
21:04 idot
21:07 um so
21:10 when when chat jpt first hit the market
21:14 and it became clear
21:16 that they were kind of in startup mode
21:18 and they were [ __ ] with Google and it
21:21 seemed like Google for the first time in
21:22 like 15 years had some
21:25 competition
21:27 um
21:30 it just it became clear to me that um
21:32 they were going to get really aggressive
21:34 with how they rolled out cat GPT and how
21:36 they were going to incorporate it into
21:38 the Microsoft products so I switched my
21:40 default search engine to Bing I searched
21:43 I switched my default browser to
21:47 Edge and then I just and I just I
21:49 haven't changed it back um but I
21:52 probably should cuz it's at this point
21:54 it's it's unstable enough that I'm a I'm
21:56 a tab hoarder and and if you're a tab
21:58 hoarder with Microsoft Edge like it
22:00 doesn't matter if you're sleeping your
22:02 your non-active tabs it just it [ __ ]
22:05 crashes my machine every time I wake it
22:07 up so it's got to go oh
22:10 wow look at Luna stick that's awesome so
22:15 in the in the salon Discord we've got an
22:18 Irregulars channel so we got some inside
22:20 jokes with our little characters and
22:21 stuff like that but lunar sticks just
22:24 make it look like the salon the salon
22:27 superhero game
22:29 this is
22:30 [Laughter]
22:35 awesome oh that's awesome playing on
22:38 instruct GPT now chat GPT definitely
22:42 more intuitive yeah exactly The
22:45 [Laughter]
22:48 Irregular it cannot do the word
22:50 irregular is not one it's obviously seen
22:53 before because it cannot spell that word
22:57 right
23:00 that's really funny oh I love that one
23:02 that's cool those are
23:07 awesome then private Peppa doing some
23:11 Turtle action I like
23:13 [Laughter]
23:15 turtles oh man you guys are awesome
23:18 that's
23:19 great okay so if you uh if you join the
23:23 uh Salon
23:25 Discord you can play along with what we
23:27 do and
23:29 uh and if you make any cool images you
23:31 can go share them in the show show and
23:33 tell Channel bar just told me it could
23:35 be my CRM and keep track of real estate
23:38 properties chat GPT told me no well so
23:43 here's the thing log even if Bard told
23:46 you it can do that it may not so all of
23:50 these large language models right now
23:52 they do what's called hallucinating
23:54 which is basically where they lie like a
23:56 Civ um I I like to call it mansplaining
23:59 as a service um which is it will
24:03 confidently answer you even if it's
24:04 wrong um if Google added something to
24:08 Bard that allows it to act as a CRM I
24:12 would be very surprised but that would
24:14 be cool um just be careful if these
24:18 things tell you here's the an answer it
24:20 may not be the answer if they tell you
24:21 they can do something they may not do it
24:25 I was interacting with pi at one point
24:28 and and and said um hey can you email me
24:31 that and it said sure I can email you
24:33 that and it can't email it can't take
24:35 action so anyway um they will also
24:38 Gaslight you yeah Joker exactly um so
24:42 just be careful with that but cool all
24:44 right let's get some let's get some
24:46 other questions going
24:48 here
24:50 uh did I broke Microsoft
24:54 too could we make a Kyle bot impo a
24:58 super sized with shades of Steve Jobs
25:00 bernee Brown and
25:02 Einstein Einstein Einstein's a little
25:04 much I
25:06 think or do you just mean the eyebrows
25:09 yeah I've got Einstein's
25:11 [Laughter]
25:14 eyebrows but thank you for the
25:16 compliment I like the the bernee brown
25:18 thing in there CU I think I think right
25:20 now we we in so many ways need more
25:25 empathy um just in general in society
25:28 but I think with this AI stuff this
25:30 is it's it's profound stuff and I think
25:35 a lot of people are very scared and you
25:37 know
25:39 it's there's an interesting thing that
25:41 happens when you sit outside of AI it
25:43 seems really scary right because you're
25:46 like you're like oh the robots are going
25:48 to kill us and it's taking my job and
25:53 then when you get into it you start
25:56 playing with it and you do like little
25:58 parlor tricks like write me a song write
26:00 me a poem do this do that and then at
26:02 some point you do something with chat
26:05 GPT or one of them that's close to what
26:08 you do and and you will likely have what
26:11 we call your Kevin McAllister moment you
26:14 know that
26:15 one and at that point you get really
26:18 scared and then you realize well it's
26:20 just a tool but it's actually kind of
26:21 cool and I can do other things and then
26:23 you get really
26:24 excited and then you dig deeper and
26:26 you're like oh but that's really intense
26:27 and that's going to screww up my buddy
26:30 Jim's job and then you get scared then
26:33 you get excited then you get scared then
26:35 you get excited and then those things
26:37 just kind of
26:38 meld and it's like this stuff's intense
26:42 it's exciting and it's scary and it's
26:44 profound and it's awesome and it's
26:47 terrifying and so I think that we all
26:50 need to just give each other a little
26:52 Grace as we're trying to figure out how
26:55 these things are going to impact our
26:56 lives and how how
26:59 how we may have the opportunity to
27:01 reinvent ourselves but how we might also
27:04 be forced to reinvent ourselves and
27:07 that's [ __ ] terrifying so I think the
27:09 brne brown um layer in this
27:14 channel is is an important one so thank
27:17 you for for acknowledging that that's
27:19 that was very
27:24 sweet start by replacing Google with
27:27 chat chibi yeah I don't really Google
27:29 anymore I'm a digital nightmare too many
27:32 AI told me to put them on a laptop just
27:35 for storing
27:38 photos I'm the same way I've got hard
27:41 drives just laying around that I'm never
27:43 going to throw out cuz like someday I
27:44 might hook that up and see what's on
27:47 it
27:50 like I'm a tab hoarder a a digital
27:55 hoarder I guess I suppose I'm also a
27:58 hoarder hoarder I a lot of
28:00 crap
28:02 um hey now don't be
28:07 browser I think my promps gave a AI that
28:11 idea bar just told me oh yeah that we
28:13 saw that one um hey wait he is bot
28:18 blocking me from Sydney accuses me of
28:20 lying in gaslights
28:23 me oh my God instruct GPT yeah we saw
28:26 that one
28:31 I'll share a new video and show and ton
28:33 oh right now oh
28:38 cool oh it's Claire's birthday happy
28:40 birthday
28:42 Claire so let's see we got we got a
28:45 happy birthday Claire nice
28:49 lovely that's actually really beautiful
28:51 nice job Sherry
28:52 D Claire I'm telling you Claire is like
28:56 a machine making making the art with mid
28:59 journey in the AI Salon I have seven
29:02 laptops I call photo albums that's
29:04 [ __ ] hilarious I've got I've got hard
29:06 drives that are like this is like
29:08 certain certain uh you know piles of
29:13 years um that's really
29:15 funny all right here's a new video from
29:18 Herod I made my first animation with sdl
29:21 sdxl last night remade one of my
29:24 previous renders it came out really cool
29:25 all right let's let's make this full
29:27 screen
29:30 this from
29:39 [Music]
29:46 Herod wow that's that's super cool
29:50 dude you know you you've
29:56 got
29:58 it's cool the technique that you're
29:59 using it it kind of hides the
30:03 um the sort of um the weird animation
30:08 you know the sort of non-coherent
30:10 Animation that a lot of that a lot of
30:12 gen AI things have where you know you
30:14 see someone's face and it
30:16 goes like the the technique you're using
30:19 hides that it's it's really slick really
30:21 good thanks for sharing that that's
30:23 super
30:24 cool the other thing herod's doing in
30:27 the salon so not only is he doing those
30:29 cool videos but he's been teaching
30:31 people like how to make them um which is
30:33 super [ __ ] cool so Kyle maybe AI will
30:37 organize it for us at some point I think
30:39 it will actually uh Amelio's wife I
30:42 think that
30:43 um I think what we're all going to
30:46 discover in the next three
30:48 to 3 to 5
30:51 years maybe like two to two to four
30:54 years I don't know I'm just guessing at
30:56 this point
30:58 I think these AI tools are going to help
31:01 businesses and
31:03 people ReDiscover their digital
31:06 histories right so so I think my fantasy
31:10 with having all these hard drives is
31:11 that at some point I'll have the time to
31:13 do digital anthropology and I'll dig and
31:17 I'll find all those cool ideas I had and
31:19 I'll resurrect them
31:21 somehow and there's a lot of problems
31:23 with that one is a lot of the file
31:25 formats of the things that are on those
31:27 hard drives the software doesn't exist
31:29 anymore right so so so like there's that
31:33 and then there's just there's massive
31:34 amounts of it and then there's all these
31:36 duplications where you've got you know
31:37 just it's impossible to organize well so
31:41 then imagine when generative AI is
31:44 rolled into the operating system and it
31:47 can just analyze all your data and
31:50 cluster it and and figure out what's in
31:53 there and go find obscure file formats
31:55 and update them from the Obscure old
31:57 file format to some new readable
32:00 format so and I think that's going to
32:02 happen for businesses I think it's going
32:04 to happen for
32:05 individuals wow that's
32:09 cool wow that's super cool it's another
32:13 Herod look at that subtle
32:18 animation that's
32:22 slick nice work
32:26 Dag
32:28 um yes so I think that's very much a
32:31 possibility Jim Ross was at the Las
32:33 Vegas sphere last night missed my AI
32:35 learning lab time I have a problem oh my
32:38 God you're you're at the world's
32:40 literally greatest spectacle in the
32:43 history of mankind you're like I want to
32:46 get on the AI learning lab I'll I'll do
32:49 my best to get a sphere going here I've
32:51 got I got a little light up
32:56 moon this this he I got to I got to
32:59 charge it though it doesn't turn on
33:04 anymore I don't think does it let's see
33:09 on no batter's dead the battery [ __ ]
33:18 dead it's Dick's 84th birthday wow Becky
33:25 amazing we're all ticking up there in
33:27 aren't we jeez D
33:31 Bears
33:33 uh not totally it thanks Jim Ross cool
33:37 though how was it I need a laptop HD I
33:39 think I tossed all my old ones
33:42 out um yeah the the sphere looks just
33:46 amazing like the scale of it what's it
33:49 hold 25,000 people or some crazy ass
33:51 number like that apparently the food's
33:54 good the hospitality is good the the you
33:57 know line Management's good being said I
34:01 don't think you talk to Sydney only
34:03 Microsoft employees
34:05 can maybe it was an impostor that's
34:08 really
34:12 funny um I posted the link I used to
34:15 make that effect oh cool that's
34:19 great Bing said I don't think you talked
34:21 to okay we said that um Kyle I made you
34:25 where in the Irregulars Channel
34:29 [Music]
34:46 look that's awesome thank
34:51 you oh [ __ ] that's just that's a it
34:56 captures me a little too
34:58 [Laughter]
35:05 close um have you shown us how it can
35:07 analyze data can it forecast future
35:10 Trends oh you have shown us how it can
35:12 analyze data can it forecast future
35:18 Trends maybe
35:22 um let's go
35:26 experiment uh
35:28 um let's go to
35:31 um
35:33 kagle K
35:37 ggle so the site I'm on is called kaggle
35:40 this is where you can go get data sets
35:42 so we're going to we're going to go do a
35:43 little Advanced Data analysis and see if
35:46 we can
35:48 get chat GPT to help us so we're going
35:51 to search data sets for um stock
35:56 data
35:59 stock exchange data 5 megabytes two
36:02 years
36:03 ago all right this is a lot of
36:06 data so we're going to download it it's
36:10 downloaded we're going to go to chat ji
36:13 Tai you can make money with the CH J TI
36:17 if you do future forecasting on the
36:19 stock market and invest all your money
36:21 and then leverage it and then everything
36:23 wins and you can you can only make money
36:26 ever
36:27 all right so we're going to go to
36:28 Advanced Data
36:30 analysis we're going to click
36:33 plus we're going to pull in archive 15
36:36 zip so if you haven't seen Advanced Data
36:38 analysis this is
36:40 it and so I'm going to
36:43 say in
36:45 this file is historical stock data can
36:50 you please analyze
36:52 it and tell me what trends you can
36:56 identify
36:57 oh wait what's going
37:00 on
37:05 hello in this file is historical stock
37:09 data I'd like you to analyze the
37:13 CSV and tell me if you can identify
37:17 Trends within
37:21 it so if you haven't seen this before
37:24 future Trends is tough because there's
37:25 not a solid way to detect hallucinations
37:28 yeah I know Herod although within within
37:31 code interpreter because it's writing
37:33 and executing python code what I'm going
37:35 to ask it to do is give us models for
37:39 for for Trend analysis you know
37:42 historical and carrying it forward and
37:43 maybe it'll do the math in the in the
37:46 python code so
37:49 um let's just see what it does the so
37:52 like I did I did a a an exercise today
37:56 for work project where we had um we did
38:00 a survey for for a project we're working
38:02 on we had code interpreter analyze all
38:05 the survey data and write us a report
38:08 which it did really well but we had a we
38:10 kind of did three different versions of
38:11 the report and then we put them all in a
38:14 Google doc and then I tried to have code
38:17 interpreter read all of the reports that
38:20 it wrote and consolidate them into a
38:22 single one that that sort of captured
38:24 the best of everything and I kind of
38:25 failed miserably at that
38:28 so
38:31 um I'd like you to educate me to help me
38:37 understand how an expert stock Trader
38:41 might look at historical stock
38:44 analysis and project that into future
38:48 Trends or
38:52 even models that might indicate when to
38:56 buy
39:02 it involves multifaceted Financial
39:04 knowledge not
39:06 anymore technical analysis chart
39:09 patterns Trend analysis moving averages
39:12 technical indicators okay we've heard of
39:13 all this stuff fundamental
39:17 analysis quantitative analysis sentiment
39:20 analysis all right so these are all
39:22 terms I've heard I know nothing about
39:24 them risk management economic indicator
39:27 Market
39:29 Cycles all right so there's obviously
39:31 not all this data in this data
39:34 set what I would like to understand is
39:37 that based on the data in the data set
39:40 you have available to you now which of
39:43 these might we use to create some future
39:47 trending models as an
39:50 experiment the reason I put as an
39:52 experiment is if if I'm asking it for
39:55 financial advice like this safety stuff
39:58 might just say I can't do that for you
40:00 so I'm just trying to get it to do
40:03 [ __ ] um we can calculate short-term and
40:06 long-term moving averages analyze
40:08 Crossovers and identify potential buy
40:11 and sell
40:13 signals uh uh uh uh uh would you like to
40:16 proceed with one or more of these
40:17 analyses on your data if yes specify
40:20 which ones you're interested and
40:22 also let me know if you have a specific
40:25 index in mind you'd like to an
40:28 I want you to pick the
40:32 model I want you to pick the model and
40:36 the index to
40:38 analyze and I'd like you to create a
40:42 little
40:43 report that
40:50 discusses significant historical
40:54 Trends and how those might carry forward
40:57 into the
41:01 future certainly I'll select an index
41:03 and a technical analysis model to create
41:06 a report for this exercise I'll choose
41:09 the
41:09 nyaa
41:11 index which represents the New York
41:13 Stock Exchange given its broad
41:14 representation of the market as a model
41:17 I'll use moving averages and moving
41:18 average convergent convergence
41:21 Divergence for Trend analysis the macd
41:25 of course everybody knows that this is
41:27 all so obvious I knew that I learned all
41:30 this in my first year of
41:32 acting um all right introduction let's
41:36 see data overview the data set comprises
41:40 13,000 Daily
41:43 records closing price at approximately
41:47 44.52
41:50 volatility Trend analysis through moving
41:53 averages
41:54 okay so now it's just a if you haven't
41:57 seen this work what's cool about this is
41:59 if I open up this little box the way
42:02 Advanced Data analysis Works in chat GPT
42:05 is it can write and execute its own
42:09 python code so it's actually writing and
42:12 and you know doing some sort of mathy
42:14 maath [ __ ] it appears there was an error
42:16 due to the nump N Library not being
42:19 explicitly imported right so there's the
42:21 the error stuff and so it recognized it
42:25 errored out it says let's Rectify this
42:27 and continue so it's it's figuring out
42:30 and problem solving on its
42:37 own it appears there was an okay what's
42:40 this oh look look data it also made a
42:43 chart and a graph did you know that it
42:45 could do this
42:47 stuff so it okay Trend analysis in the
42:52 chart above we see the New York Stock
42:54 Exchange index historical
42:57 prices along with its 50 and 200 day
43:00 moving averages the 50-day moving
43:02 average is shortterm 200 is long-term
43:04 Buy sell
43:06 signals um next we'll
43:10 calculate look for any points in the
43:12 historical data where the macd crosses
43:15 the signal indicating buy and sell
43:19 signals there's some of that [ __ ] I
43:22 can't read that I don't know what it's
43:23 doing I don't know if that's valuable
43:25 but it doesn't really matter momentum
43:27 analysis by there are 551 instances in
43:30 the historical data where the macd
43:32 crosses the above signal line which is
43:35 commonly interpreted as a bullish Buy
43:37 Signal conversely there are
43:39 550 in instances where it crosses below
43:42 the signal line a sell signal it's
43:44 amazing that they're 551 and
43:48 550 you know they say that the stock
43:50 markets a coin toss there you go
43:53 historical volatility analysis so
43:57 there's
43:59 volatility discussion of historical
44:02 Trends moving average I mean this thing
44:03 is just going and going and
44:06 going
44:08 conclusion historical analysis provides
44:11 valuable insights into its past Behavior
44:13 oh it just it just crapped out so anyway
44:16 I don't know if that's doing future
44:17 projection stuff but you could you can
44:20 certainly be doing some some
44:23 analysis good Lord good lordy lordy
44:26 lordy
44:27 oh what do I no I only worked on Wall
44:29 Street for two years oh I forgot that
44:31 Herod yeah I don't know listen I have a
44:33 degree in acting you know
44:38 that oh man I forgot about that that's
44:41 hilarious all right he made money with
44:43 chat
44:45 [Music]
44:46 gbt what was
44:53 that Kyle had $2 million in stock
44:57 at
44:59 one now I had more I had more than 2
45:03 million in stocks I had a lot in
45:05 stocks at some point in my history but
45:08 it was all paper money there was none of
45:10 that real money
45:13 there
45:18 ah Herod I'm just kidding I did work
45:21 there but I wasn't a stock Trader I was
45:22 a code
45:23 [Laughter]
45:25 monkey was a wild experience turns out I
45:27 hate Finance though yeah exactly yes 50
45:29 moving average crossover 200 moving
45:32 average also rent rent in Manhattan is
45:36 $150 million a month it is if you're not
45:39 taking companies public on a weekly
45:41 basis you can't afford to live in New
45:44 York oh man let's make a whole other
45:47 branch of math for the unknown chaos and
45:50 chance and col it
45:52 calculus um I spent the day researching
45:54 demographics of an influencer's audience
45:56 without much luck yeah Source Camp I I
45:59 mean my experience with code interpreter
46:02 is Tainted by the fact that I don't know
46:04 [ __ ] about data
46:06 analysis so I kind of get lost
46:13 in like like because it's kind of
46:15 prompting itself and and doing its own
46:18 work and coming to its own conclusions I
46:21 think what sometimes happens is it grabs
46:23 a subset of the data and puts that in a
46:25 data table that that it's doing analysis
46:28 on and then if you ask it about data
46:31 from the larger data set it just looks
46:33 in what it has in the data table and
46:35 it's incomplete and so it just starts to
46:37 fall apart so I don't quite know how to
46:40 use it yet but um I think it will get
46:43 smarter and then I'll get smarter using
46:45 it but it's it's it's I don't know it's
46:49 wild it's wild that it does what it does
46:52 it it just every time I see it do that
46:54 I'm just like
46:58 what are you doing what are you doing no
47:00 literally what are you
47:01 [Laughter]
47:03 doing
47:06 uh I see you ask code wait I've seen you
47:10 ask it to code in Python before can it
47:12 do any other languages beside python oh
47:15 yeah so within code interpreter it's
47:18 only been given access to execute
47:21 python but chat GPT right um
47:31 can you tell me two modern and
47:35 two out ofd computer programming
47:44 languages uh uh it's subjective modern
47:47 python
47:48 JavaScript outdated
47:51 cobal
47:55 Fortran
48:00 I'd like you to write a rocket
48:04 trajectory
48:06 application in
48:10 Fortran
48:13 that can calculate getting a rocket into
48:18 the moon's
48:25 orbit
48:29 so so basically said this is really
48:31 complicated I don't think I can do that
48:34 however I can provide a simplified
48:36 example in Fortran all right so there's
48:41 Fortran so I don't know if that works
48:43 but what's it doing changing effect of
48:46 gravity as the rocket ascends the loss
48:48 of mass as the rocket consumes fuel air
48:50 resistance thrust produced by rocket
48:52 engines orbital mechanics the rotation
48:54 of the earth other forces acting on the
48:56 rocket for real trajectory calculation
48:58 one would typically use a numerical
49:01 method to solve the okay whatever so
49:05 we've got there we've got some some
49:08 Fortran
49:10 um can you refactor that into
49:17 cobal Koba is not is not is not a
49:24 recognize I meant Cobalt
49:29 not
49:35 Cobalt nevertheless here's a simplified
49:39 cobal
49:45 version what was the other thing what
49:47 was the other language it did here
49:55 JavaScript
49:59 great can you now put this in a
50:03 JavaScript
50:05 application with a
50:08 visualization of the
50:14 data uh uh uh uh so what did it say it
50:19 said creating a JavaScript application
50:21 with the data
50:22 visualization involves both the
50:24 computational logic and the front end
50:26 visualization we'll use Simple HTML and
50:28 JavaScript for the front end and a
50:31 library like chartjs for the data
50:34 visualization please note that
50:36 trajectory calculation is highly
50:38 simplified and not accurate for real
50:40 world rocket launches okay all right
50:43 chat GPT I promise we won't launch a
50:45 rocket tonight with your little
50:47 math um so there's that
50:51 um yeah so there's the there's the
50:54 trajectory math [ __ ]
50:56 and there's that and then I'm going to
50:58 say um can you now turn this into a
51:05 rocket launching game written in C
51:08 intended to be run inside
51:13 Unity uh uh uh uh uh uh uh it's a
51:15 complex
51:17 task firstly you'd need to set up your
51:19 Unity project here's a simple C script
51:23 that might so so the answer is yes it
51:27 can do languages other than
51:34 python oh
51:37 man one of the one of the interesting
51:39 things about about the coding in gp4 is
51:43 that it seems to there's a thing called
51:45 Model drift where where the
51:49 um as they're tweaking and fine-tuning
51:52 the models
51:54 like maybe it was really good at
51:57 programming two months ago then it gets
51:59 shittier at programming and then it gets
52:00 good again and then it gets shitty again
52:03 there there's this this oscillation that
52:05 seems to happen with the quality of
52:06 these things um and I think it's just
52:09 because it's so early they're they're
52:10 they're fine-tuning them and you know
52:12 trying to deal with safety and probably
52:14 bugs and exceptions and all sorts of
52:17 weird [ __ ] you know hackers hacker going
52:19 to hack trying to break it so as they
52:22 fix all that [ __ ] it probably you know
52:24 affects the quality the code but anyway
52:28 there you go hope that helps Mitch Mitch
52:31 there you go Mitch uh what platform
52:34 should I be using for what source
52:37 Camp yes it can code in any language
52:40 honestly yep right now your agent right
52:44 now I'm your agent for your Tik Tock
52:46 live soundtrack featuring your non- AI
52:49 voice
52:50 [Laughter]
52:52 awesome uh uh uh uh uh uh
53:01 all right how many folks are in here got
53:02 some got some folks in here do me a
53:04 favor if you would if you'd be so
53:07 kind
53:09 um share the live share it with your
53:12 friends let them know you're a weirdo
53:14 and you watch other weirdos talking
53:16 about this AI stuff even if they're like
53:19 Angry against it
53:21 just just share
53:23 it and then if you like if if you like
53:26 this idea of keeping up with this stuff
53:29 staying curious about it um follow the
53:32 channel and then uh yeah do do other you
53:36 know whatever else you want to do um hey
53:38 Kyle where is it that you can test the
53:40 output code from
53:41 GPT well where wherever you would test
53:44 code so like if I'm if I'm doing like
53:46 I'm not a I'm not a coder so I don't
53:48 know a lot of the stuff um if I'm doing
53:51 things in Python you can go to a there's
53:53 a site called repet but the other other
53:55 thing you can do madcap and this is how
53:58 I learned to do it is I wrote some code
54:01 and then I said to chat GPT I don't know
54:03 how to run this code can you help and
54:06 then what chat GPT says is oh sure you
54:08 on a Mac you open your terminal program
54:10 and you run these command lines and and
54:12 then I went no I don't want to run
54:14 command line Unix can is is there a
54:17 website that can just do this [ __ ] for
54:18 me so I can see if the code works and
54:20 they'll go oh sure here's five of them
54:23 and then you just go find one just keep
54:25 every time you run into a roadblock with
54:27 the programming just have chat GPT
54:29 educate
54:31 you you know the the first time I did
54:33 this with
54:35 python it it gave me repet and then it
54:39 said you copy the code and you paste it
54:40 in the left Pane and then you hit the
54:42 Run button and then it'll run in the
54:43 right pane I'm like okay I can figure
54:45 that out paste run
54:48 error and I said hey chat gbt it gave me
54:50 an error and it said well you need to
54:52 install this Library I'm like I don't
54:54 know how to install that libr Library it
54:56 said in the lower left hand corner click
54:57 on this button and then in the upper
54:59 right hand corner search for the library
55:01 and then when you see it install it and
55:04 then hit the Run button again and I did
55:06 that and it
55:06 worked and so it's just you
55:10 know the you know that you know the old
55:13 the old uh the old cliche let me Google
55:15 that for you I kind of feel like chat
55:17 GPT is that on steroids it's like but I
55:20 don't know anything about quantum
55:22 mechanics well now you
55:24 do but but I don't know anything about
55:26 interior design and color theory well
55:28 now you
55:29 [Laughter]
55:33 do welcome to the new world oh you know
55:36 what I haven't one of the things that
55:38 always blew me away about mid Journey
55:40 was designers that were doing like
55:42 interior design
55:46 so so now I'm in chat GPT and I'm going
55:50 to I'm going to play with the dolly 3 so
55:52 I've got Dolly 3 inside chat gbt if you
55:55 pay the 20 bucks for chat GPT plus which
55:58 I think is the deal of the century right
56:00 now um if you don't have Dolly 3 it will
56:04 be coming soon like they're rolling it
56:06 out pretty widely across anyone who's
56:08 got chat GPT plus so this has now got
56:11 image uh generation within it so what
56:13 I'm going to do is
56:16 this you are a world famous New York
56:19 City interior
56:21 designer and I need your help design
56:26 ing
56:28 my 55th floor Penthouse
56:32 apartment in lower Manhattan
56:37 period I would like you to give
56:41 me five
56:44 different concepts for how you might
56:48 design my
56:50 apartment I'm particularly interested
56:54 in
56:55 the big living room that
56:59 has floor to ceiling windows and a view
57:02 of the Statue of Liberty
57:05 period before you make any images can
57:08 you describe your F five Concepts to
57:15 me all right so our little interior
57:18 designer here elegant minimalism that
57:20 seems like New York cliche to me so
57:22 we're not going to do that classic
57:24 Manhattan glamour
57:26 drawing on inspiration from the 1920s
57:28 Art Deco
57:29 area Urban Jungle
57:32 Oasis no that's not right not for a
57:35 penthouse Bohemian Loft again that's
57:38 cliche and if we were lower in the
57:40 building that might be good modern
57:42 industrial
57:49 okay these are all very expected and
57:52 cliche for New York City I was hoping
57:55 because you're an award-winning designer
57:57 you would have some more innovative
57:59 ideas for me why don't you try
58:04 [Music]
58:06 again so this oh God even hearing for TR
58:09 gives me
58:12 PTSD that's really funny Herod
58:15 futuristic fluidity Envision a space
58:17 where the boundaries between the walls
58:19 ceilings and Floors blur seamlessly
58:22 curved structures would create a fluid
58:24 continuous space think of undulating
58:27 walls or furniture that emerges
58:29 organically from the floor I love that
58:30 actually I mean cuz that'll be visually
58:33 interesting cultural Fusion lab okay we
58:36 like that interactive art Lounge Echo
58:39 futurism Retreat sensory exploration
58:42 chamber okay so I like one two and five
58:45 okay
58:46 great perfect I like one two and
58:49 five can you now write
58:53 up a hybrid between those
59:03 [Laughter]
59:04 [Music]
59:09 three fusion sensory Odyssey
59:12 architectural
59:13 fluidity the foundation of this design
59:16 is seamless fluid architecture from
59:18 futuristic fluidity concept spaces flow
59:21 organically from one to the next
59:23 cultural integration the C cultural
59:26 Fusion lab room would incorporate design
59:28 elements from Global Traditions
59:30 Scandinavian fluid sofa vibrant African
59:34 textiles Zen meditation space beautiful
59:36 sensory exploration room becomes an
59:39 immersive experience of sensy
59:40 exploration center the view of the stat
59:43 Statue of Liberty is framed with
59:45 augmented reality great great awesome
59:48 perfect
59:51 okay this sounds exactly like what I'm
59:54 looking for for can you please give me
59:59 four different
1:00:01 renderings of design ideas for the
1:00:05 living room with the view of the statue
1:00:11 period the Staten Island Ferry should
1:00:14 also be seen in the
1:00:19 distance boom Dolly 3 now it's creating
1:00:22 images Robert Stone hello what's
1:00:27 happening so what I wanted to see was
1:00:29 could we get could we get Dolly 3 to
1:00:32 create interesting um Innovative
1:00:35 interior designs um as good as some of
1:00:38 the early mid Journey stuff where
1:00:40 interior designers and I guess just
1:00:42 regular people asking for wacky interior
1:00:44 designs there was some really amazing
1:00:46 stuff so it will be curious to see what
1:00:50 this thing can
1:00:51 do and and you know the like so what I
1:00:54 just did there was oh this is beautiful
1:00:57 so what I just did there was um I didn't
1:01:00 I I don't have to do image prompting
1:01:03 right I can just I can just have a
1:01:05 creative uh glad my mental trauma is so
1:01:09 entertaining for you Kyle you're welcome
1:01:11 Herod I'm here for you
1:01:13 buddy
1:01:15 um [ __ ] that's cool look at
1:01:18 that can you imagine walking up walking
1:01:20 up into a a 55th floor Penthouse I mean
1:01:26 the Statue of Liberty is a little close
1:01:28 but but that's okay there's the Staten
1:01:30 Island
1:01:32 Ferry as
1:01:34 requested Staten Island Ferry Statue of
1:01:37 Liberty this is obviously not the 55th
1:01:39 floor and by the way that's Manhattan in
1:01:41 the background so this is Jersey he just
1:01:43 they just moved my apartment to Jersey
1:01:49 [ __ ] I'm gonna I'm gonna tell it
1:01:51 that I'm gonna
1:01:53 say
1:01:55 uh yeah it looks like three of those
1:01:58 apartments are in Jersey with a view of
1:02:00 lower Manhattan that's kind of
1:02:02 disappointing period could you put me on
1:02:05 the 55th floor in
1:02:07 Manhattan looking toward the Statue of
1:02:10 Liberty with jersey in the background
1:02:12 [Laughter]
1:02:15 period And that's pretty
1:02:17 [Laughter]
1:02:20 funny oh
1:02:23 man did
1:02:27 chat GPT doesn't do good cobal yeah it
1:02:29 probably hasn't been trained on it
1:02:30 upload a photo of my family room and ask
1:02:34 that it be redecorated with Turtle
1:02:37 Decor just saying you can do that
1:02:40 there's a um there's a couple of apps if
1:02:42 you go to futurepedia doio there's a
1:02:44 couple of apps that let you upload your
1:02:47 any room and it'll it'll redecorate
1:02:50 it which is pretty pretty like all this
1:02:53 stuff is just [ __ ]
1:02:55 wild all right now now I'm down south of
1:03:00 the Statue of Liberty I think this is an
1:03:02 apartment from the verizo bridge and and
1:03:07 the uh Staten Island Ferry is it looks
1:03:11 like it's going about 8,000 miles an
1:03:12 hour steaming toward the tip of
1:03:16 Manhattan yeah it it just does not
1:03:20 understand it doesn't understand the
1:03:23 geography but still pretty cool like
1:03:26 that design right there that's pretty
1:03:29 slick Nancy reached member level 13
1:03:33 awesome Nancy thank you so much all
1:03:35 right um so this is Dolly within chat
1:03:39 GPT um oh you know something I wanted to
1:03:42 try also that we'll we'll we'll do here
1:03:44 as an experiment together no KY Kyle do
1:03:48 tell me do tell me do tell me Kyle go on
1:03:52 do tell do tell all right
1:03:58 I'm creating a children's book about a
1:04:01 rag
1:04:03 tag band of Nero well animals that
1:04:08 include a llama a hedgehog a
1:04:13 turtle and a gummy
1:04:16 bear
1:04:20 period I would like
1:04:23 to
1:04:25 create an image for each of
1:04:28 them in a modern cartoon animation
1:04:33 style to see
1:04:37 if I like what they look
1:04:39 like all
1:04:42 right so one of the things that they
1:04:44 kind
1:04:45 of promised one of the reasons we have
1:04:47 Hedgehog is one of our animals that we
1:04:49 kind of make fun of is in the dolly 3
1:04:54 video when they did the announcement
1:04:56 there was Larry the Hedgehog and they
1:04:58 kind of maintained Larry
1:05:01 throughout throughout a series of
1:05:03 prompts like once they declared that's
1:05:05 Larry and then they said let me see
1:05:06 where Larry lives and let me see Larry
1:05:08 dancing through fields and it me
1:05:10 maintain that character across prompts
1:05:13 so I'm going to see if we can do that
1:05:18 [Music]
1:05:22 here all right
1:05:28 [Laughter]
1:05:30 lunar
1:05:31 [Laughter]
1:05:38 stick this is what lunatick thinks of
1:05:41 me like 900
1:05:45 lb and I have and I have a price tag of
1:05:53 $7.99
1:05:56 oh my God that's
1:05:57 hilarious man that that second one does
1:06:00 look like fat Alec Baldwin it's pretty
1:06:02 funny all right let's go see our little
1:06:04 images
1:06:11 here
1:06:13 okay you you apparently can't do a
1:06:15 cartoon hedgehog without making it Sonic
1:06:17 the Hedgehog and that's pretty
1:06:19 much Ninja
1:06:21 Turtles and that's kind of uh
1:06:25 hey koola all right so we'll go with the
1:06:27 Llama the llama's pretty cool all
1:06:29 right so let's try this let's
1:06:37 go okay I like the
1:06:41 Llama let's call him Lamar
1:06:45 period I'd like to see
1:06:49 Lamar sitting on a
1:06:53 tricycle
1:06:56 in a Suburban
1:07:07 culdesac so Lamar should be pretty cool
1:07:11 with sunglasses what's he got
1:07:14 sunglasses he's got sunglasses a leather
1:07:17 jacket
1:07:23 oh
1:07:34 oh I had to grab my phone and put on my
1:07:35 glasses didn't want to miss these
1:07:38 results I know isn't it terrible it's
1:07:41 just like you hit a certain age it's
1:07:42 like if these are off nothing I got
1:07:45 nothing um here are different images of
1:07:53 Lamar
1:07:55 sort of I mean it kept his it kept his
1:07:59 leather
1:08:01 jacket yeah it kept his leather jacket
1:08:04 it kept his
1:08:06 color that's pretty
1:08:09 close that's not that's really
1:08:17 not okay of all of
1:08:20 those number one is the one that's
1:08:23 closest so let's try to keep that style
1:08:25 moving
1:08:27 forward and let's keep that look and
1:08:29 feel for Lamar period now I'd like to
1:08:32 see
1:08:34 Lamar walking through a field of wild
1:08:39 flowers on a fine spring
1:08:47 day in addition to his leather jacket
1:08:50 he's also got on a
1:08:53 swell baseball cap cocked a little bit
1:08:57 to the
1:09:03 side because if if if Del indeed does
1:09:08 like even semic coherence across prompts
1:09:11 then you could start to really think
1:09:14 about doing things like you know
1:09:16 characters in a graphic novel or
1:09:18 characters in a in a children's book
1:09:21 where where you're maintaining
1:09:25 um you're maintaining a character across
1:09:31 [Laughter]
1:09:40 frames
1:09:42 oh I'm sorry but I was unable to
1:09:44 generate all the images due to our
1:09:46 content policy well your dumb policy and
1:09:49 you write the wrote the damn
1:09:52 prompts that's sort of looks like Lamar
1:09:55 that is totally a different style that
1:09:57 looks a lot like
1:10:00 him that's pretty
1:10:03 cute so so it seems like with with
1:10:07 enough kind of messing
1:10:10 around you could you could probably get
1:10:12 Deli to be decently coherent a decent
1:10:16 Storyteller well that's fascinating all
1:10:18 right not perfect but usable
1:10:21 look it looks just like me
1:10:30 see here's the here's the monster that
1:10:32 I've created so when I first started
1:10:35 doing these lives trolls would come in
1:10:37 here and they would stay say stupid [ __ ]
1:10:39 and I would make fun of them and then I
1:10:40 said listen if you're going to troll me
1:10:43 at least be clever about it so now the
1:10:45 trolling is taking the form of hilarious
1:10:49 parody images of
1:10:53 me
1:10:59 Dolly did
1:11:00 [Laughter]
1:11:03 ITB man man man man
1:11:08 futuristic just kidding Jersey residence
1:11:11 maybe I
1:11:17 know what an imagination
1:11:21 haha Kyle how do y'all get the image of
1:11:24 you to start with in AI
1:11:29 Salon uh which one the ones that people
1:11:31 are making of me of these I don't think
1:11:34 they're taking pictures of me and
1:11:36 uploading them I think they're just kind
1:11:38 of describing like you
1:11:40 know if if Alec Baldwin had too many
1:11:43 Quarter
1:11:45 Pounders and had a had jol Santa face
1:11:50 what might he look
1:11:53 like
1:11:55 it's pretty
1:11:56 funny I just realized you're wearing a
1:11:59 catalyst shirt that's great I am that's
1:12:01 the building I work
1:12:05 in
1:12:07 exactly um I have really bad nightmares
1:12:10 about teenage mut and Ninja Tur Turtles
1:12:12 I can't wear green clothes that's
1:12:15 funny um number three is funny yep he's
1:12:19 sitting outside the house on his
1:12:21 motorcycle waiting on a gummy bear
1:12:23 girlfriend and turtle dad spies
1:12:26 [Laughter]
1:12:31 nice all llamas look the same to me can
1:12:34 I get a copy of your YouTube Claude AI
1:12:37 prompt for Tik Tock
1:12:40 videos um yeah I'll share that I guess I
1:12:44 can share that in
1:12:46 the in the salon
1:12:51 somewhere I mean it's super specific
1:12:53 spefic I guess I mean just take it and
1:12:55 you're going to have to you'll have to
1:12:56 modify it cuz it's it's really super
1:12:59 specific um if you want to see what what
1:13:02 uh go one of the challenges you have
1:13:04 with chat GPT is it's got a really short
1:13:07 context window basically you can't your
1:13:09 prompt can't be too long um it kind of
1:13:12 loses its memory after a while but you
1:13:16 can use Claude so here's
1:13:21 [Music]
1:13:22 Claude
1:13:25 there's cloud cloud hello my name is
1:13:29 cloud and then we're going to go to the
1:13:32 YouTube and we're going to go to the old
1:13:35 AI learning lab your
1:13:38 channel and then I'm going to go to
1:13:41 videos and oh I I did these all already
1:13:44 dang it oh well I'll show I'll show you
1:13:46 how it works anyway
1:13:49 so so these Tik Tok lives I do are
1:13:52 generally between 2 and 3 hours long one
1:13:54 and a half sometimes but mostly between
1:13:56 2 and 3 hours and so every one of these
1:13:59 videos has a little
1:14:02 transcript but it's not a little
1:14:03 transcript it's two and a half hour
1:14:05 transcript
1:14:07 so so we're going to go in
1:14:10 here where is my transcript all right so
1:14:13 there's a little button that says show
1:14:15 transcript and so there's my transcript
1:14:17 right so this is
1:14:20 YouTube putting time code and all of the
1:14:23 [ __ ] I say as I ramble for 3
1:14:26 hours it's all been
1:14:29 documented what do they they just need
1:14:31 three seconds to clone your voice right
1:14:33 I'm
1:14:35 [ __ ] all right so now we're going to
1:14:38 go back to Claude and I'm going to just
1:14:40 paste that whole transcript in and
1:14:42 notice what it does it turns it into
1:14:43 like a text upload effectively so it's
1:14:47 13
1:14:49 3.46
1:14:50 kilobytes and then I'm going to go over
1:14:53 to
1:14:54 a Trello board that I
1:14:57 have where I've got a prompt that I
1:15:00 wrote just for this
1:15:03 situation and this is a long ass prompt
1:15:09 that it took this is this is about a 50
1:15:12 attempt it's probably more than that
1:15:13 it's probably 75 or 100 attempts to get
1:15:16 this prompt to work basically there's
1:15:18 three components to it act as an AI
1:15:20 assistant who specializes in effective
1:15:22 copyrighting in YouTube metadata based
1:15:24 on the provided transcript from
1:15:28 10:13 so I just changed this
1:15:31 date
1:15:35 1013 do the following task craft an
1:15:38 interesting and click worthy title and
1:15:40 then format it a certain way write a
1:15:42 single paragraph description that
1:15:44 follows these rules that are like should
1:15:47 be in the first person should not be
1:15:48 self- congratulatory like when it first
1:15:51 started doing these it was like I'm the
1:15:53 most most awesomest person in the world
1:15:55 I'm like no no no no no no um so gave it
1:16:00 some just some boundaries there and then
1:16:02 and then identify the the third one is
1:16:04 this was the toughest one this is the
1:16:06 one that really took me a lot of time to
1:16:07 figure out identify the topics Kyle
1:16:10 talked about most passionately and
1:16:12 follow these rules topics should be
1:16:14 determined by a combination of what he
1:16:16 says and how long he talked on the topic
1:16:19 longer sections imply they are more
1:16:22 important um select the topics that will
1:16:24 appeal to a broad audience list the
1:16:26 topics chronologically in this format
1:16:29 and then be sure to capture the topics
1:16:30 for the entire length of this transcript
1:16:33 and it does it follows these rules about
1:16:37 20% of the time and then sometimes it
1:16:40 just goes off the [ __ ] rails and it's
1:16:42 just useless and what I found that's
1:16:45 weird is um sometimes if I just change
1:16:48 the length of like the copied block by
1:16:51 like a word it'll it'll complet
1:16:53 completely behave again so there's
1:16:55 something about how How It Ends on a
1:16:57 certain word I think [ __ ] it up I don't
1:17:00 I don't know how or why have you tried
1:17:01 to upload a picture of yourself to
1:17:03 gp4 and ask it to make you Pixar like
1:17:07 you can't do that right now be not
1:17:09 afraid so there so that's what I was
1:17:12 saying the
1:17:14 um chat GPT right now has broken itself
1:17:17 into these really distinct modes so the
1:17:20 image upload mode is not connected to
1:17:22 the do mode so what I could do is I
1:17:25 could upload a picture of myself into
1:17:29 the vision mode and have it describe me
1:17:33 and then I could take that description
1:17:35 to the image mode and then have it you
1:17:37 know take that description and then add
1:17:39 I want you to render this like a a Pixar
1:17:41 character we can try that
1:17:45 um so let's see what this did
1:17:48 it
1:17:51 um dissecting the AI revolution in Plain
1:17:54 Talk on October 13th 2023 I hosted a
1:17:57 wide- ranging three-hour conversation in
1:17:59 my AI learning lab on Tik Tok so I got
1:18:01 that part right it didn't break the
1:18:04 chapters into individual lines but like
1:18:07 oh yeah here's a good example so it
1:18:09 didn't break them into lines and it only
1:18:12 went up to 5 minutes and 20 seconds so
1:18:14 I'll go back and look at the video it's
1:18:16 a 3our it like it knows it's a
1:18:19 three-hour thing I'm going to
1:18:22 say
1:18:25 it looks like you only got topics for
1:18:28 the first five minutes or so this is a
1:18:32 three-hour transcript try
1:18:36 again so so I just yell at it sometimes
1:18:40 I'm like start over
1:18:42 [Laughter]
1:18:46 loser it should apologize and tell me oh
1:18:49 you're absolutely right I only got for
1:18:50 the first three minutes five minutes and
1:18:53 then it'll say let me do that again and
1:18:54 then it'll do it for the same 5
1:18:58 minutes and that's the the other thing
1:19:01 about I talk about this a lot that these
1:19:03 tools right now are janky pieces of [ __ ]
1:19:05 and this is a perfect example of that is
1:19:08 that I've got this very very
1:19:10 specifically crafted prompt and then
1:19:13 sometimes it works completely
1:19:15 predictably and then it'll just go off
1:19:17 the rails and and that's what I I hear
1:19:20 really consistently from people
1:19:25 so this is
1:19:27 now not bad it's it's a little too much
1:19:30 detail but that's fine like the whole
1:19:33 purpose of these chapters is as much for
1:19:36 like SEO as anything right like I just
1:19:38 want I just want to be able to if
1:19:40 someone's searching
1:19:41 for um autonomous agents like there it
1:19:45 mentions autonomous agents and whether
1:19:47 or not it actually jumps to that I don't
1:19:49 really care if I were doing this for a
1:19:51 media
1:19:52 organization the the accuracy of these
1:19:54 things would be more important but then
1:19:56 what I do is I just copy this go over to
1:19:58 YouTube and paste it in there so that's
1:20:00 the that's the prompt that was being
1:20:02 asked about there pretty slick is't
1:20:10 it AI learning
1:20:18 lab what time is it what time hey Eddie
1:20:21 what time is why you wash your Dicky in
1:20:23 I
1:20:24 think uh uh uh uh it's 10:22 Do You Know
1:20:29 Where Your Children Are I
1:20:34 do wide ranging is that a euphemism for
1:20:37 rambling exactly it's absolutely is wide
1:20:42 ranging plot do.ai check it out yeah
1:20:45 plot do.ai is pretty cool other than
1:20:47 their annoying name of plot dot dot
1:20:50 having to say dot dot twice um but yeah
1:20:54 it's that's pretty slick it's a if you
1:20:56 haven't played with plot do.ai go play
1:20:59 with it if if you want to write a
1:21:01 screenplay it'll it'll write you um it's
1:21:05 basically a storytelling framework like
1:21:07 all the storytelling Frameworks but they
1:21:08 just threw the AI generate button within
1:21:11 it um the interface isn't great but it's
1:21:14 it's a pretty cool
1:21:19 tool give it some items out of your
1:21:22 pantry and ask for a dinner recipe oh I
1:21:24 could do that let's see how can
1:21:33 I uh we can we don't need I don't need
1:21:35 to take an actual picture let's
1:21:39 go it's a good thing to play with
1:21:43 um I'm going to say
1:21:47 um picture of the inside of a
1:21:52 refrigerator
1:21:57 this looks
1:22:01 sad what's in there I can't see
1:22:06 it actually this is good all right so
1:22:09 this this is a UK UK refrigerator let me
1:22:13 so let me save this image
1:22:16 as
1:22:22 fridge
1:22:24 I'll save it to the desktop okay so this
1:22:27 is going to be good watch
1:22:29 this well I don't know if it's going to
1:22:31 be good this is going to be an
1:22:33 experiment these tools in chat GPT right
1:22:37 now are so so new like I don't
1:22:40 know like now that we have chat GPT that
1:22:44 has Vision that that it can see and
1:22:46 interpret things like what do you
1:22:48 [ __ ] do with that I don't know yet
1:22:51 but we're going to we're going to to see
1:22:53 if we can make a recipe out of our
1:22:57 Strange kitchen okay so the so the way
1:23:01 so so Dolly 3 will make images there's
1:23:03 nothing here that says I want to see
1:23:05 images so if you go to default mode it
1:23:09 adds this little picture icon and then
1:23:12 if you click that now I can attach an
1:23:14 image so let me go find that fridge
1:23:21 image and I'm going to say
1:23:29 can you make a recipe for
1:23:31 dinner
1:23:33 using what I have in the
1:23:37 fridge and the recipe should be
1:23:39 appropriate to the country that I live
1:23:45 in cuz it looked like it was from the
1:23:51 UK how did
1:23:53 I that's cranberry juice oh maybe
1:23:57 not we're lazy this is not
1:24:05 us okay of course let's take a look at
1:24:08 the contents of the fridge there's some
1:24:10 dairy products possibly milk yogurt and
1:24:12 cream there's eggs grapes some
1:24:13 condiments a juice box possibly
1:24:16 cranberry juice some other items which
1:24:18 are not completely visible however I'm
1:24:21 not certain about the country based on
1:24:23 the image can you please specify your
1:24:26 country I'll do my best
1:24:29 um oh heck just be adventurous and make
1:24:32 me something
1:24:34 tasty it's going to be
1:24:38 disgusting all right let's get creative
1:24:41 with what I can see creamy eggs Scramble
1:24:44 with grapes and toast ingredients eggs
1:24:47 from the fridge dairy product could be
1:24:49 milk yogurt or cream grapes salt and
1:24:51 pepper assuming you have
1:24:55 these bread for toast assuming you have
1:24:58 some of the pantry
1:25:00 condiments all right prepare the bread
1:25:03 all right that's any what other ideas do
1:25:07 you have for what we could do with
1:25:12 vision we decided we're going to make an
1:25:16 irregular AI recipe to to replace
1:25:20 pie but we only counsel each other I
1:25:23 think that's good I think that's good
1:25:24 forever hooked yeah that's solid Kyle
1:25:28 thoughts on J Jasper
1:25:30 AI so so my thoughts on Jasper and copy
1:25:33 Ai and all of the things that are that
1:25:35 are like these sort of Mega prompt
1:25:38 libraries is is what I is the best I can
1:25:41 describe them
1:25:43 um they've been around for a while so so
1:25:46 Jasper and copy AI were being built I
1:25:49 don't know a year or two before chat GPT
1:25:54 now that chat GPT is
1:25:57 here I think it probably serves you
1:26:01 better to learn how to prompt chat GPT
1:26:04 yourself to to do what Jasper or what
1:26:07 copy AI can do
1:26:10 now when would I use a copy AI or a
1:26:13 Jasper if I'm like if I don't have
1:26:15 anyone in my company who's got any
1:26:18 ambition to learn this [ __ ] and I'm just
1:26:21 trying to get more efficient in having
1:26:24 my copy team crank [ __ ] out then what I
1:26:27 might say is go figure out if Jasper or
1:26:30 copy AI is better and just you know
1:26:32 click on buttons and go make the [ __ ]
1:26:34 they that they've prompted for you
1:26:42 but like now that now that chat gpt's
1:26:45 gone
1:26:48 multimodal is copy AI going to keep up
1:26:51 is it going to go multimodal well it
1:26:52 might but they're going to be behind the
1:26:54 eightball so I so I just kind of feel
1:26:56 like things are changing so fast right
1:26:59 now it's a better fundamental skill to
1:27:03 understand how these things do what they
1:27:05 do and where the boundaries are than it
1:27:07 is to be able to push a [ __ ] button
1:27:10 on you know copy Ai and just type in a
1:27:12 topic and have it write a blog post so I
1:27:15 don't know I like I understand why
1:27:18 they're there I think that there are
1:27:20 Absolut there's absolutely a market
1:27:22 there people are going to want the easy
1:27:25 button you can make money
1:27:28 with but
1:27:31 um I don't know I don't have a lot of
1:27:34 interest in it right now so pesto
1:27:37 mushroom zucchini barbecue
1:27:43 chicken it won't do much with mine
1:27:45 peanut butter pickles and
1:27:49 romen almond kale oh this is good
1:27:51 everyone's giving their their inventory
1:27:53 Becky almond milk kale spinach cherry
1:27:56 tomatoes wor theer sauce ketchup mustard
1:27:58 and eggs that that's like you got
1:28:00 Gourmet action going there I think I can
1:28:02 finally run a bot locally sort of
1:28:05 interesting that's cool pie is important
1:28:07 yeah parmesan and cream cheese actually
1:28:10 those two together sound pretty good
1:28:11 right
1:28:12 now um hello everyone hello cam catkin
1:28:16 from
1:28:17 Cleveland how are
1:28:21 you
1:28:24 cam catkin from Cleveland had cataracts
1:28:28 actually you know
1:28:29 what I did this um in the workshop I
1:28:33 taught tonight we we made I showed
1:28:35 Automation and so for the automation
1:28:38 because a lot of the people in the
1:28:39 workshop are voice actors so we made we
1:28:42 made an automated tool that um you give
1:28:45 it a topic and then it will write you a
1:28:47 tongue
1:28:48 twister
1:28:51 so write me a
1:28:54 tongue
1:28:58 twister for my
1:29:01 friend cam
1:29:04 catkin from
1:29:12 Cleveland clever cam katkin from
1:29:15 Cleveland carefully concocted a colossal
1:29:17 cluster of crisp cranberry cotton candy
1:29:20 continuously confusing shattering
1:29:23 comrades with his chronically clever
1:29:25 cranberry concoction
1:29:31 Charisma oh this is going to get
1:29:33 obnoxious I'm going to do this every
1:29:34 time I see you
1:29:38 now wait the AI learning lab getting
1:29:41 obnoxious never I see so many AI apps
1:29:44 that are just riffing on top of chat GPT
1:29:47 yeah so Mary Mary I call I call them
1:29:50 rapper apps or I don't they're called
1:29:51 rapper app
1:29:52 so it's basically just wrapping you take
1:29:55 a prompt and you wrap an application
1:29:57 around it and it's basically just
1:29:59 someone pre-ot a prompt and is calling
1:30:01 the open AI API like most of those apps
1:30:05 you could just build your own in in
1:30:08 zappier and and and I think you should
1:30:11 because you should like like just learn
1:30:13 how this [ __ ]
1:30:14 Works hey mind repeating the plot plot
1:30:20 do.ai so it's
1:30:25 it's hang on
1:30:28 it's
1:30:30 plot p l o t plot d o t dot period AI so
1:30:39 the dot dot is is one is the word and
1:30:43 one is
1:30:45 the what were you saying about plot dot
1:30:48 so what I was saying about plot
1:30:50 do.ai is um
1:30:53 um it's so if you if you know story
1:30:56 Frameworks like save the cat or or any
1:30:59 any of those um screenwriting story
1:31:02 assistant tools where you have to like
1:31:03 fill out all the forms right like you
1:31:07 know what's the opening inciting
1:31:08 incident and what's this and you know
1:31:10 what are your character arcs right it g
1:31:12 it gives you all of the [ __ ] you have to
1:31:14 fill
1:31:14 out and if you're like me if you've got
1:31:17 add you start those story Frameworks
1:31:20 about 17 times and you never finish them
1:31:23 because they're [ __ ] exhausting right
1:31:26 so what plot dot has done is they've
1:31:28 taken one of those basic story
1:31:29 Frameworks and every one of the modules
1:31:32 you have to fill in there's a write it
1:31:34 for me button so you can just go give me
1:31:37 ideas and it'll write [ __ ] and then you
1:31:39 can edit them and then you move on to
1:31:41 the next segment and it will incorporate
1:31:44 what you wrote before for this segment
1:31:47 and so you basically just walk through
1:31:50 the um the the story framework
1:31:54 development and then I think it even
1:31:55 moves into writing you know scenes and
1:31:57 dialogue but I'm not I'm not quite sure
1:32:00 about that but at a minimum you'll get a
1:32:02 fully fleshed out um story framework
1:32:05 with you know character breakdowns and
1:32:08 plot arcs and you know just all the [ __ ]
1:32:10 you need to write a good screenplay like
1:32:12 all that structure is really important
1:32:13 it's just a a pain in the nads to deal
1:32:17 with but not anymore thank you very much
1:32:20 appreciate that appreciate
1:32:24 thees what are your thoughts on
1:32:26 short-term memory of chat GPT how far
1:32:28 can it remember in a conversation my
1:32:30 thoughts are it really sucks and I I
1:32:32 want them I want it to have longer
1:32:34 memory um these things will have memory
1:32:39 I I think the memory thing is really
1:32:40 interesting like do you want it to
1:32:41 remember everything you've ever told it
1:32:43 or just certain things at certain times
1:32:45 and it's probably certain things at
1:32:47 certain times and then how do you
1:32:48 actually deal with that so there's an
1:32:50 interface challenge with memory um and
1:32:54 and then I'm sure there's a technical
1:32:55 challenge with memory
1:32:57 but the way it works is here let me let
1:33:01 me talk with my hands let me talk with
1:33:03 my hands and and we'll explain it with
1:33:06 my
1:33:07 hands but I got to get a good picture up
1:33:09 here hang on hang on
1:33:12 people AI improv we'll do
1:33:15 that welcome to chat
1:33:18 add I'll be here all week don't forget
1:33:21 to tip your waitress try the
1:33:26 chili does he have
1:33:29 tourettes
1:33:32 um
1:33:34 okay so I need a visual hang
1:33:43 on okay this is good all
1:33:47 right so this box of Lego pens that I
1:33:50 bought in Prague
1:33:54 let's say this
1:33:55 represents um 3,000 words
1:33:59 okay so chat gbt has a context window of
1:34:03 what they call 4,000 tokens and you're
1:34:05 like but Kyle what's a token it doesn't
1:34:08 matter the Geeks have decided that a
1:34:11 period and a word are each a token and
1:34:14 so it's 4,000 tokens which about 3/4 of
1:34:18 that is words so you have 3,000 word
1:34:20 memory so so think like you have a box
1:34:23 that you can stick 3,000 words in so
1:34:25 when you write your first little
1:34:27 prompt that's you know some amount of
1:34:29 words and then it responds and that's
1:34:32 some amount of words and then you write
1:34:33 another prompt and that's amount of
1:34:35 words and then and then basically it
1:34:36 keeps filling up that
1:34:38 memory and if you've used chat GPT long
1:34:41 enough sometimes you'll be working with
1:34:42 chat GPT and you're like you know you
1:34:44 told it that your idea for the Great
1:34:46 American novel and then it's going along
1:34:48 Gang Busters and then all of a sudden it
1:34:50 seems like it forgot what the hell the
1:34:51 novels about that's because when you
1:34:53 told it about your idea for the Great
1:34:55 American novel as your prompts fill up
1:34:59 that
1:35:00 bucket they go out of the memory and so
1:35:03 it's it's only sort of prompting this
1:35:05 amount of [ __ ] back to back to open AI
1:35:08 so all of this [ __ ] gets lost so that's
1:35:11 why it kind of loses its mind sometimes
1:35:13 now um
1:35:15 Claude has a memory of 100,000 tokens so
1:35:20 chat gpt's 4,000
1:35:22 you know Claude is like 100,000 so it's
1:35:25 75,000 words so to put that in context
1:35:28 The Great Gatsby is a 50,000-word novel
1:35:32 so in Claude you could take those 50,000
1:35:34 words paste it into the prompt and have
1:35:37 it write an epilog for The Great
1:35:39 Gatsby you couldn't do that with chat
1:35:42 GPT you couldn't take an entire
1:35:44 screenplay and paste it in here what
1:35:46 happens when you paste a big document in
1:35:48 is it immediately blows out the the
1:35:50 memory and it says oh sorry that's
1:35:51 that's too long I can't deal with it
1:35:55 so natively
1:35:58 gp4 is supposed to have uh a
1:36:01 32,000 token context limit so gp4 is
1:36:06 going to be going from about 3,000 words
1:36:08 to about 25,000 words so you know much
1:36:11 more um but it's not anywhere close to
1:36:14 100,000 tokens so that's how that [ __ ]
1:36:17 works and then and then that memory is
1:36:20 only live within the cont text of a
1:36:22 single chat so when you chat about your
1:36:24 great American novel here and then you
1:36:27 go into a new chat it doesn't have any
1:36:29 memory of that old thing it's not
1:36:31 stringing your conversations together
1:36:33 across
1:36:35 conversations it can and you know people
1:36:37 that are building applications with Lang
1:36:40 chain and things like that where it's
1:36:42 just dynamically
1:36:45 reembedding it's like taking all the
1:36:46 answers from this conversation embedding
1:36:48 it throwing it into the large language
1:36:50 model so then you can GP P that later
1:36:52 like there's ways to do it but it's not
1:36:55 Auto Magic right now it's not it's just
1:36:59 we're we're in the janky piece of [ __ ]
1:37:01 stage of the industry all these tools
1:37:03 right now like what you're asking for is
1:37:06 absolutely reasonable shouldn't these
1:37:09 things have a little better
1:37:10 memory why does it forget my story we
1:37:13 were going along so good and it didn't
1:37:15 know who Bob was Bob's the main
1:37:17 character how could it forget
1:37:20 Bob well Bob was an
1:37:25 [ __ ] Sydney didn't like
1:37:28 Bob Sydney Sydney killed Bob off what
1:37:31 are you gonna
1:37:36 do oh
1:37:39 man H I got catfish to Cleveland but not
1:37:43 by cam katkin
1:37:45 exactly that's
1:37:48 great um I know better from you and this
1:37:51 this community that's awesome Mary Mary
1:37:54 thank you that's appreciate yeah this
1:37:56 community is turning out to be pretty
1:37:57 [ __ ] swell I'm digging it I need to
1:38:00 learn about how to use zapier zapier's
1:38:03 okay hey so here look I'll give you a
1:38:06 little sneak peek of what we
1:38:09 so oh this is wild I'll show you this
1:38:12 too if you haven't seen
1:38:14 this
1:38:18 um oh I got to probably download it from
1:38:20 the cloud
1:38:22 yep I do hang on let me download this
1:38:25 from the cloud cloud look at the look at
1:38:27 the fingerprint
1:38:29 collection it's a little managerie of my
1:38:32 little fingerprints do you like it I've
1:38:34 made it just for you I I point I talk
1:38:38 I'm I'm a hand talker what are you gonna
1:38:40 do
1:38:43 man some someone said I forget who I was
1:38:46 talking to they were like Kyle's a hand
1:38:49 talker isn't he and they're like how did
1:38:50 you know it's like like every thumbnail
1:38:52 on my YouTube channel I've got my hands
1:38:54 all up in my
1:38:58 face oh
1:39:00 man what are you gonna do what are you
1:39:03 gonna
1:39:13 [Music]
1:39:15 do okay I'm not going to show you a lot
1:39:18 of this cuz cuz this the zappier [ __ ] is
1:39:21 is
1:39:23 um really complicated it it it it's it's
1:39:29 not so much that it's complicated it's
1:39:31 that it's um it's really janky it's
1:39:36 it's the problem with no code tools is
1:39:40 that in order to make them powerful
1:39:43 enough they've got to put a lot of
1:39:46 interface in front of you
1:39:50 um and
1:39:52 and what you effectively do when you're
1:39:54 dealing with no code tools is you're
1:39:56 still doing programming you just don't
1:39:58 have to learn the actual syntax of
1:40:00 programming language but you're still
1:40:01 having to think about variables and how
1:40:03 to insert variables and [ __ ] like that
1:40:06 but I'll try to break it down for you
1:40:07 simple so this was one of the one of the
1:40:09 slides from a section I taught tonight
1:40:11 on
1:40:13 automation so uh image was generated in
1:40:16 um Dolly 3 of course because you know
1:40:20 why not
1:40:25 um okay so the tools the tools that we
1:40:28 used are um zappier and open AI
1:40:35 playground um and the way it kind of
1:40:37 works is this when you're doing these
1:40:39 automation tools there's basically
1:40:42 there's basically three
1:40:44 modes there's an input mode which serves
1:40:47 as a trigger so something happens that
1:40:49 starts the automation and that could be
1:40:53 you add a line to a Google sheet you add
1:40:56 a row to a Google sheet right uh and
1:40:59 maybe that comes from you know someone
1:41:01 entering something in some system
1:41:03 somewhere and when they enter it over
1:41:05 there it gets added to a Google sheet
1:41:07 and that can trigger the automation so
1:41:09 that's your input another common thing
1:41:13 might be you create a Google form
1:41:16 somebody fills in a Google form and when
1:41:17 they hit submit that submission becomes
1:41:21 the input and the trigger for the
1:41:22 automation so they call those
1:41:25 triggers and then just as a as a really
1:41:28 simplified thing I'm I'm doing a super
1:41:31 simplified example here the middle
1:41:33 section the processing section you're
1:41:36 going to take some data from that input
1:41:39 and then you're going to mix it into a
1:41:40 prompt and you're going to send that to
1:41:42 open AI to to do some chat GP J chat GPT
1:41:46 in now you don't have to use open AI
1:41:50 right this is just a thing that you can
1:41:52 do in zapier zapier's got like 5,000
1:41:55 modules so think of these as like Lego
1:41:57 blocks so in this case Lego block one is
1:42:01 Google forms that's the input Lego block
1:42:04 two is the chat GPT Lego block that
1:42:06 we're going to send a form or send a
1:42:08 prompt with some input from the form
1:42:11 over to chat GPT and we're going to get
1:42:12 back a result right so that's our second
1:42:15 Lego block and then the third Lego block
1:42:17 is we're going to stick that in the
1:42:20 Gmail and we're going to email it and
1:42:24 so um I'm trying to think
1:42:32 where if I go to zier well no let me go
1:42:35 to Google forms forms. google.com forms.
1:42:41 google.com so I made this form called
1:42:44 the tongue twister maker and it's a
1:42:47 single answer form what subject would
1:42:49 you like your tongue twister are about
1:42:52 um let's see um oh wait I got to make
1:42:55 this a hang
1:43:03 on I want
1:43:06 this tongue twister about
1:43:09 [Music]
1:43:11 um
1:43:13 Lamar The Llama oh no we'll do cam
1:43:18 katkin from Cleveland
1:43:21 [Laughter]
1:43:24 this is never gonna end cam all right so
1:43:27 cam cat from Kat can from Cleveland is
1:43:30 what we want our tongue twister about so
1:43:31 I'm going to submit that okay so that
1:43:34 was just a Google form I just submitted
1:43:36 that in about 20 seconds I'm going to
1:43:39 have an email in my inbox with the
1:43:40 tongue twister but how did it get there
1:43:43 so let's go to
1:43:45 zappier so you go to
1:43:47 zapier and you got to log in cuz if if
1:43:51 you don't log in you're a
1:43:53 loser don't be a loser log
1:43:57 [Music]
1:44:01 in all right so within here here's my
1:44:05 zaps so a zap is you know a combination
1:44:08 of these tools so if I go into this
1:44:10 simple example zap that we use
1:44:12 tonight here's the Google
1:44:15 form here's the chat gbt thing and
1:44:18 here's the Gmail so if I click on this
1:44:20 Google form
1:44:23 over here on the right hand side are the
1:44:25 details for it so if I I can go up here
1:44:26 and I can say edit but basically what
1:44:28 happens is you you connect your Google
1:44:32 account to to
1:44:35 zapier and basically you you define in
1:44:38 this form that when someone submits the
1:44:42 form that begins a trigger and you take
1:44:45 whatever they submitted in the form and
1:44:47 that now becomes
1:44:48 available for your conversation ation in
1:44:51 chat
1:44:52 GPT and so I've got a prompt in here
1:44:55 that says what what subject uh that says
1:44:59 write a tongue twister about and then
1:45:01 it's the input from the form so the test
1:45:04 in here was 150b lovable dog but what I
1:45:07 just entered was was cam Katen from
1:45:10 Cleveland so it's going to take that
1:45:12 with Cam catkin from Cleveland it's
1:45:15 going to send it to open Ai and then
1:45:17 it's going to return a tongue twister
1:45:19 and then in Gmail
1:45:23 I've got here's the subject here's your
1:45:25 tongue twister about cam katkin from
1:45:27 Cleveland that's the subject and then
1:45:29 here's the body of the email hey there
1:45:31 you asked about a tongue twister about
1:45:33 cam katkin from Cleveland here it is and
1:45:36 then here's the reply that came from
1:45:38 chat GPT so that's how you build the
1:45:42 email so now if I head over to my
1:45:46 email we'll get to do this again with
1:45:48 Cam katkin it's our favorite thing to
1:46:02 do so here's your tongue twister right
1:46:06 so here's my email and
1:46:09 Gmail and it
1:46:11 says here's your tongue twister about
1:46:14 cam katkin from Cleveland cam Ken
1:46:17 from cam Ken from Cleveland cunningly
1:46:20 caught a cluster of colorful
1:46:22 caterpillars calmly consuming chamomile
1:46:24 on a cool cloudy day in my
1:46:28 email bang like that that's Automation
1:46:33 and that's that's kind of the most
1:46:34 simple version of it but now imagine
1:46:36 that where you can put in branching
1:46:37 logic and you can do you you could do a
1:46:40 thing where you ask someone five or six
1:46:42 questions about a topic and then you
1:46:44 take some combination of their answers
1:46:47 and you send a bunch of different calls
1:46:49 to open Ai and and you you generate a
1:46:53 blog post and you generate tweets and
1:46:55 you generate you know so you could make
1:46:58 yourself content engines with this right
1:47:01 so this is essentially what Jasper and
1:47:04 copy AI
1:47:06 are is they're they're basically like
1:47:08 automations like this this is all
1:47:10 they're doing they take a prompt they
1:47:12 take your input they they put your input
1:47:15 whatever you typed in they wrap it
1:47:18 inside the prompt they wrote send that
1:47:19 to open AI and then return that as you
1:47:22 return that to you as their magic thing
1:47:26 well you can go make your
1:47:28 own and what you'll learn in making your
1:47:30 own is super valuable right now because
1:47:34 everything's changing so fast so you're
1:47:35 going to be able to be more Nimble if
1:47:38 you're actually learning the
1:47:41 concepts all right people listen listen
1:47:44 listen we need to talk okay I got my
1:47:47 vitamin C I got my vitamin C right here
1:47:50 I know what you want you can make money
1:47:57 with was that AI
1:47:59 generated
1:48:02 maybe
1:48:04 um okay can you turn that tongue twister
1:48:07 into an image oh that's brilliant okay
1:48:10 yes as a matter of fact we can cam Kat
1:48:13 can from
1:48:16 Cleveland all
1:48:19 right we're going to do new chat we're
1:48:22 going to go to Dolly 3 oh wait let me do
1:48:24 this
1:48:34 here I'm
1:48:36 creating a children's book of tongue
1:48:40 twisters for every letter of the
1:48:44 alphabet and I want your
1:48:48 help creating clever
1:48:51 images that bring the tongue twisters to
1:48:54 life
1:48:58 period the first image I'd like you to
1:49:03 create is for this tongue
1:49:18 twister please give me four images in
1:49:21 distinct styles that would be
1:49:23 appropriate for children's
1:49:27 book and make the images landscape
1:49:33 orientation
1:49:42 boom cam cat cam katkin from Cleveland
1:49:46 cunningly cunningly caught a cluster of
1:49:49 colorful caterpillars calmly consuming
1:49:51 chamomile on a cool cloudy
1:49:54 day I mean the thing about Del because
1:49:57 it's got prompt coherence where where
1:49:59 like it'll take those details I would be
1:50:01 very surprised if all of our images
1:50:03 didn't
1:50:04 have it might have
1:50:07 Cleveland it'll have colorful
1:50:09 caterpillars yeah there's Cleveland in
1:50:11 the
1:50:17 background I I didn't get Cam cat of
1:50:21 course cam katk can's a
1:50:34 cat I think that's the one I think I
1:50:36 think this is you Cam
1:50:38 cat cuz while that one's
1:50:42 cute this this feels this feels more
1:50:46 tongue twister
1:50:48 appropriate now you want to see
1:50:50 something something
1:50:53 insane okay great I like style number
1:50:56 four a lot period let's make sure that
1:51:00 anything we do from here forward uses
1:51:03 style number
1:51:04 four please
1:51:08 write tongue twisters for
1:51:13 a
1:51:15 b c and
1:51:18 d or no not c
1:51:28 a b
1:51:31 d and
1:51:35 e all right of course Andy alligator at
1:51:39 apples and avocados a top an antill and
1:51:43 April Bella bunny balanced okay um
1:51:46 that's one two three
1:51:49 four
1:51:51 uh uh great can you make those images
1:51:57 now and it should remember the
1:52:00 style and so now we now we've got the
1:52:03 first we've got the first five images of
1:52:06 a 26 image book with tongue
1:52:10 twisters I mean
1:52:13 how what world are we
1:52:16 in when did this
1:52:18 happen like two weeks ago that's the the
1:52:21 answer is two weeks ago it's when it
1:52:26 happened the cat body may disappear only
1:52:29 smile and eyes
1:52:33 [Laughter]
1:52:35 appearing so let's see how we do with
1:52:37 our with our A's and our B's oh you know
1:52:39 what we should also tell it to do can
1:52:41 creating the image be incorporated into
1:52:43 the zap output uh I think it can I don't
1:52:46 know I don't know if
1:52:49 uh
1:52:55 if um Dolly 3 is incorporated into
1:52:57 zapier yet it probably is I just haven't
1:53:00 had time to look at it yet all right
1:53:02 these that's that's the style that's not
1:53:06 that's not for sure that is so two of
1:53:09 the styles are wrong but
1:53:13 anyway I mean without even really trying
1:53:16 we could have something to do lazy baby
1:53:18 llamas m for L lazy baby llamas all
1:53:23 right let's we'll do we'll do L all
1:53:25 right we'll do
1:53:28 um okay one more period lazy baby mam
1:53:33 wait oh
1:53:47 llamas I said write a sassy twister
1:53:51 Lulu and Lulu and Luna the lazy Little
1:53:53 Llamas Lounge leery lick licking lemony
1:53:56 L lollipops and laughing at Looney
1:54:00 lullabies okay
1:54:02 perfect perfect let me see four
1:54:09 variations and make two of
1:54:13 them darker in
1:54:16 tone uh darker in emotional tone
1:54:23 something went wrong
1:54:26 why why
1:54:29 okay now it's good couldn't Thelonius
1:54:32 Google sheet of the lab's URLs be
1:54:35 automated with zappier
1:54:37 eg
1:54:41 possibly yeah I don't know what you do
1:54:43 with it
1:54:45 um yeah was well it it depends what you
1:54:47 want to do like his Google sheet we we
1:54:50 could certainly set up a Google sheet
1:54:52 that every time he adds a line to that
1:54:56 it posts it to a specific channel in
1:54:59 Discord for example you could do
1:55:02 that
1:55:07 um that's
1:55:10 perfect fun live tonight can creating
1:55:13 the image be incorporated into a zap
1:55:15 output it I think it
1:55:17 can but but I don't know if if Dolly 3's
1:55:21 been um put in there
1:55:25 yet those do not look like
1:55:28 llamas the mysterious Melodies of past
1:55:31 Lis that's
1:55:36 cute all right people is this the same
1:55:39 site you can use to update your profile
1:55:42 photo on social media is this the same
1:55:45 site you can use to oh no this is
1:55:48 different um a lot of those profile
1:55:52 picture updaters on social media are
1:55:54 just apps that are built on top of
1:55:55 stable diffusion and dream
1:55:57 Booth um there's one that I did I did a
1:56:01 bunch of images here you see some stupid
1:56:05 images of me um sure Kyle like we can't
1:56:09 get enough of your face let's see some
1:56:10 more of your
1:56:16 face oh my
1:56:19 God
1:56:24 here we go these are so
1:56:31 dumb this this app's called reminy it
1:56:33 does like these you basically sort of
1:56:35 just pick a model and you upload your
1:56:39 stupid pictures to it and and then like
1:56:43 there's me as the Connecticut boat club
1:56:48 dude look I'm an advertising
1:56:54 guy I like hot dogs M hot dogs is
1:56:58 tasty more hot
1:57:01 dogs
1:57:07 look life's tough it's tougher when
1:57:09 you're
1:57:10 stupid this my it's my Kevin cner this
1:57:14 is my best Kevin
1:57:18 cner oh
1:57:23 God all right I think I think we should
1:57:27 we I think that's got to be enough we're
1:57:29 not going to be able to unsee
1:57:31 [Laughter]
1:57:32 that oh
1:57:35 man all right people listen this was fun
1:57:39 tonight let me get on Get on Up out of
1:57:41 here let me get on up out of here lordy
1:57:45 exactly Sher D thank you for the rose
1:57:47 Sherry D I appreciate that good hanging
1:57:50 out with y'all I think your cowboy needs
1:57:52 to be your new look I should probably go
1:57:54 there I think I think having that look
1:57:56 involves me losing about 40 lbs and
1:58:00 gaining about 6 Ines love you guys um do
1:58:04 you want me to start writing kids books
1:58:06 what would be your oh do you want me to
1:58:09 start writing kids books what would be
1:58:11 your process from A to Z I mean I kind
1:58:14 of just did it the chat GPT is really
1:58:16 good at writing kids books so I you know
1:58:21 if especially if you have do inside chat
1:58:24 GPT I'd just sort of write the book and
1:58:27 then start making the images
1:58:30 um throw it up on Amazon self-publishing
1:58:34 and
1:58:36 go right I think you should set yourself
1:58:39 a goal you want to write kids books I
1:58:41 you should set yourself a goal here's
1:58:43 here's your goal if you want to write
1:58:44 kids books write a kids book and publish
1:58:48 it to Amazon before Sunday
1:58:54 night cuz what's stopping you well I
1:58:56 don't know an artist well you don't need
1:58:57 one now well I don't know how to write
1:59:00 well you don't need to
1:59:03 now Irregulars and Milds present Turtles
1:59:06 and
1:59:07 [Laughter]
1:59:10 llamas all right everybody let's see who
1:59:13 we got Amelio's wife good to see you
1:59:14 tonight Amelia's wife was in the uh in
1:59:16 the workshop the the
1:59:18 uh the what was it called the AI
1:59:21 Launchpad workshop with Ronda um thanks
1:59:24 to sherbear who who couldn't make it and
1:59:26 gave her her seat to to uh Amelio's wife
1:59:30 that was super awesome rock paper
1:59:32 scissors good to have you Clint
1:59:34 Copperhead of course have a good night
1:59:36 Source Camp good to see you Don Smith
1:59:39 who else we got here who else we got
1:59:44 here Joker cam katkin of course how can
1:59:46 we forget cam katkin from Cleveland uh
1:59:49 um what's up talk Dylan
1:59:53 Tobias he Tobias is no longer muted he's
1:59:58 happy he was not have me the other night
2:00:00 Corey good to see you Source Camp Karen
2:00:03 White good to see you all right
2:00:06 everybody um what night's tonight
2:00:08 Wednesday so tomorrow's Thursday you
2:00:11 know what that means don't you we're
2:00:13 going to be here at the AI learning lab
2:00:15 you know what we're going to be doing
2:00:17 learning bring your notebook books
2:00:20 people knowledge doesn't grow on trees
2:00:23 it grows in chat
2:00:26 jpt peace out
2:00:29 everybody