aiLL 6/28/23 - The Innovations Transforming Our Future, By Kyle Shannon

Transcript

1. 0:00foreign [Applause]
2. 0:23with the AI learning lab welcome welcome welcome welcome happy Wednesday night yeah we'll get rolling here who's that Professor Sonya hello
3. 0:40what's shaking just getting started eight o'clock on the button here on Wednesday in Denver how is everybody
4. 0:58not a Missy what's that yield What's that name went away welcome everyone my name is Kyle Shannon this is the AI learning lab welcome
5. 1:11welcome welcome I was wondering when you would start I know I got a little bit late start I was watching the news and
6. 1:16fell asleep so I had a little cat nap after dinner good evening good evening good evening cheers I got some uh some iced coffee to us
7. 1:26Stave off the nap don't want to nap while I'm live but you know that could be entertaining you know someone screen records it sends
8. 1:35it to me hello hello all right oh wow 100 people here holy crap my name is Kyle Shannon this is the AI learning lab I'll talk about these
9. 1:46URLs behind my head in a second here um a little bit about my background I am an entrepreneur and thanks for tapping
10. 1:54the screen I think that'll let people know we're here I'm starting to figure out how this thing just what
11. 2:00what the kids call this tick ticker talker thing starting to figure out how this ticker Tucker works tap the little
12. 2:07screen that'll give us some little hearts hearts go up goes out people come in they say hi we hang out
13. 2:13uh this is the AI learning lab uh I'm an entrepreneur I'm the current uh co-founder and CEO of a company called
14. 2:21storyvine which is an automated video storytelling platform I am completely rethinking um the future of the product and the
15. 2:31value proposition thanks to AI so thanks Ai and also AI scarier than [ __ ] um I do swear occasionally so if you don't like that I'd go now
16. 2:44um what else I was a an early uh founder in the early days of the World Wide Web I was a co-founder of a company called
17. 2:51agency.com which is one of the first digital agencies and uh oh good looks like someone Lauren Lauren lay looks like you're coming to
18. 3:00the uh to the salon I'll talk about the salon in a second as well um I was there for the early days of the
19. 3:06world wide web and why I mentioned that is that the early days of the web and the early days of this generative AI stuff which
20. 3:14I'm sort of marking uh the launch of chech EPT as kind of this the new launch point of this generative AI stuff um
21. 3:22very very similar a lot of parallels from the early days of the web in the early days of uh of this generative AI
22. 3:28stuff so that's what this channel is about these URLs behind my head uh chat.openai.com that's the official chat GPT website bing.com
23. 3:38is Microsoft's search engine but if you click on the chat button that's also chat gbt it's gpt4 and you get to use it
24. 3:46for free it's got some nuances from the official chat GPT I'm a fan of the top one but if you don't want to spend the
25. 3:53money for gpt4 up here go there and it's free and then you can also if you go to poe.com that gives you access to six
26. 4:01different large language models that you can kind of compare and contrast um models from anthropic and from open
27. 4:08Ai and from Poe which is quora um so that's that and then prompts.chat that is it's really just a document
28. 4:17that teaches you a little bit about prompting and then it's a whole long list of different personas that you can
29. 4:23assign to Jet gbt like act like a screenwriter act like a business coach act like a life coach act like a I don't
30. 4:31know poet comedian whatever and it will go off and do that good evening Emilio's wife welcome welcome welcome if you've
31. 4:37got I'm echoing hmm I don't know what to do about that well um hey Tick Tock stop making me Echo
32. 4:52any suggestions for designing a clothing line with Ai No Echo you're good okay great it might be on your end there
33. 4:59Emilio's wife oh you might want to turn off the echo feature get it it's comedy we do we do stand-up comedy even though I'm sitting oh
34. 5:09I get no respect all right I'm a gen xer I'm bitter we have fun and uh let's let's get into this so if you've got questions about AI
35. 5:20pop them in the questions below I will do my best to answer them all I will fall behind that's a promise and let's get started
36. 5:28what prevents Bad actors from making Skynet um they can't get cast in any films [Laughter] or joke see I got a Hamilton shirt on that's
37. 5:40it's uh that's thespian comedy um I think what you're talking about however is um horrible people from using AI to destroy all of humanity
38. 5:53um this is this is perhaps a naive answer hahaha I think that we are early enough in whatever this stuff's going to become
39. 6:07and and it's the reason this channel exists right so if you haven't played with chat EBT I strongly encourage you
40. 6:12go there and start playing while you hear me ramble I think that any anything that we do right now that bends the use
41. 6:19of AI toward the good is going to have a disproportionately positive effect on the long-term trajectory of generative AI
42. 6:28um are there risks that bad actors are going to do bad [ __ ] I think it's more than risks I think Bad actors will do bad [ __ ]
43. 6:35um but I also think that other people are going to do remarkable stuff like remarkably positive stuff and so go start playing go start
44. 6:47um like like here's the thing even though Bad actors might make the next Skynet and might take down Humanity
45. 6:56um don't let that stop you right now from getting curious about this stuff and going and learning what it's capable
46. 7:01of and what it can do for you right now for your community for your business for your new company for your friends go start playing with it
47. 7:12um learn about it talk about it um share what you're learning you're the CEO of Vine I love Vine no I'm the CEO
48. 7:21of storyvine which uh funny story we got a little spat with the old Twitter lawyers over that one uh because we were
49. 7:30there first and after I don't know about a year of back and forth with their lawyers they're like yeah you probably
50. 7:37on the trademark but we're Twitter so go ahead and sue us if you want so yeah it's life in a big city but anyway we're still storybine
51. 7:48but it is a video platform which is why we got in a trademark lawsuit with Twitter okay drinking cocktails like on the show
52. 7:56yeah well if there were booze in the in in Guild coffee sure but I'm I'm uh I took a little nappy right after dinner so I'm still waking up
53. 8:07um okay thanks for your reply today you're welcome Troy I want to design a clothing line with AI any any suggestions on
54. 8:15where to start so let's just go play one of the things I like to do here is just let's just flip the camera around
55. 8:23turn off the old wacky URL image which by the way that was generated in stable diffusion and let's go do some live uh stuff let's
56. 8:34go do some live come on uh some live what do they call it I don't know uh fashion designing what are we what are we doing
57. 8:48suggestions for designing a clothing line okay we're gonna design a clothing line all right so first thing we're gonna do we're gonna say
58. 8:56um I want you to act like a world class clothing designer tell me oh wait let me
59. 9:12I'm gonna go use gpt4 connected to the internet because maybe just maybe it will connect to the internet tell me what are the latest
60. 9:25fashion trends and we'll get something here maybe it it may possibly because I put latest in there it may fire off a little plug-in
61. 9:35to go search the internet it may not it may just hang because I'm doing this live live from Denver playing on chat gbt what'd you do
62. 9:46tonight I don't know I watched some old guy try to load a website how'd that work out for you yeah it's pretty awesome
63. 9:57browsing the web see what did I tell you I told you it was gonna do that am I good or what clicked on a link what did I click on
64. 10:06where did it go the spring 2023 the Vogue all right the Vogue editor shopping list all right so vogue's gonna tell us what the latest
65. 10:14trends are it's gonna be fantastic it's thinking it went to some other site 2023 fashion trends so it's reading multiple sites that's
66. 10:24kind of cool so if you haven't if you haven't seen plugins or if you haven't seen chat gbt connected to the internet
67. 10:29this is kind of what it's doing in the background when it goes and searches the internet that one said reading content failed
68. 10:35that might have been behind a paywall or something like that so it's trying it again it's probably gonna fail again but we'll see
69. 10:45click three times now all right this is getting boring come on chat gbt just give me an answer come on come on come on you can do it
70. 11:00you can do it uh by the way welcome here my name is Kyle Shannon uh what we're doing right now is is watching uh Chachi PT in a
71. 11:08seemingly endless loop trying to read uh a website and uh this channel this live what you're watching right now
72. 11:17it may or may not get more exciting than this chances are are uh are good that it won't okay sure I understand the following are some
73. 11:26major Trends in men's fashion I didn't say men's fashion but that's okay relax tailoring outdoor inspired pieces
74. 11:32sustainable materials bold prints monochrome outfits unfortunately okay that's good enough so I'm gonna say okay great write me a
75. 11:44visual prompt of a um male model on a fashion shoot in New York City and
76. 11:59describe his uh whatever fashion clearly I I purposely made that kind of a shitty prompt because I'm going to show you something fun here
77. 12:12oh no it's going back out to the internet I'm gonna stop that stop that we're gonna turn off the internet
78. 12:17because I just oh wait but I'll lose my context ah dang it okay we'll do this again I know this is horribly boring
79. 12:23people are leaving here in droves because they're like why are we just watching a screen of someone waiting for
80. 12:28a website to load seriously it's gonna be so worth it you don't even know how cool this is gonna be man it's gonna be so stinking awesome
81. 12:38lucky name that movie that was also a really bad uh impression of yawn Dynamite
82. 12:52something is wrong yeah something is wrong as you're taking forever tonight because apparently people like playing with chat chippy tea
83. 13:00by the way if you're here uh because you watched a video that said you can make money with Chachi pite
84. 13:07that's not what you're gonna find here oh talk about some [ __ ] and I'll show you some cool [ __ ] but
85. 13:14and we can talk about entrepreneurship and starting businesses and making money in general but AI is there's no get rich quick scheme
86. 13:24with AI any more than there's a get rich quick scheme with okay I'm done with this okay enough about that so that was
87. 13:32a bad example all right we are now at po-po.com is one of the sites that I tell people to go check out and one of the things that Poe has is
88. 13:41it's got these little um these Bots and and these are just Bots that people made so this this one right here MJ Captain is a bot that
89. 13:52helps write mid-journey prompts so I'm gonna say um male model um um photographed in New York City wearing
90. 14:09relaxed fit out Dorsey uh High and fashion I don't know not great but what
91. 14:24this prompt will do is it will take that and very very quickly give me something better a photo of photo of a male model
92. 14:30walking down a busy Street in New York City wearing a high-end relaxed fit outdoorsy fashion so it kept that in the
93. 14:37style of Richard avadon nice shot using a 35 millimeter camera with the title Urban Explorer focal point on the
94. 14:45model's face black and white color palette with high contrast capturing the energy of the city around him okay I'm
95. 14:52going to change that a little bit but I'm going to go copy that so now what we're going to do is we're going to hop over to Discord
96. 14:58and if you've never seen Discord before I'll give you a little lay of the land in the left-hand side of Discord these
97. 15:04are all different servers so those are different communities the community I'm in right now is the AI Salon which is a
98. 15:10community that I started about just a bunch of people geeking geeking out on AI stuff and within it these little
99. 15:17things here are channels so different people having like here's General discussion and they're talking about
100. 15:23plugins and chat GPT and here's video tool talk and then here's mid Journey so this is a mid Journey channel so if
101. 15:31you've never used mid Journey it is in my opinion by far the best of all the image things so if you're doing
102. 15:38if you want to do a fashion line and you want to generate things I would I would definitely recommend mid-journey because
103. 15:44it's just better and then how you make an image is you type slash imagine and then it gives you a little thing for
104. 15:53a prompt and I'm just gonna paste The Prompt in there that we had before and I'm gonna say I don't want this to be
105. 15:59I'm going to say instead of black and white color palette I'm going to say cross processed color palette with high contrast that's
106. 16:09good capturing the energy of the city a three by two image no we want this to be yeah that's good we went three by two now we want two by
107. 16:18three we'll do two by three it'll make it taller than it is so that's aspect ratio AR no rigid upright
108. 16:26or boring uh version 5.2 so we're using version 5.2 of mid journey and I'll hit return and so now this is going to go off and um
109. 16:36start thinking a little bit hopefully it'll work a little faster than chat gbt did that that was excruciating earlier
110. 16:42uh welcome everybody looks like there's some new people here if you're new here uh someone asked me about how they might
111. 16:48get started doing uh high-end fashion a clothing line with AI designing it so I've popped over here to uh what's it
112. 16:57called uh mid-journey which is a generative AI image tool that in my opinion is one of the best now you'll
113. 17:04notice as these things as as this is rendering it kind of started out as this really four blurry blobs and then they
114. 17:11got slightly less blurry and slightly less blurry what you're actually watching there is you're watching the
115. 17:17the this diffusion model it's called a diffusion model it's the way it actually generates images it kind of averages
116. 17:24probably thousands of images together um and and uh into basically kind of a a highly noised image and then it denoises
117. 17:35that kind of random noise image over 20 or 30 or 50 or 70 steps and so right now we're at 93 percent and so there it is so
118. 17:47I can open this in a browser we can zoom in on these we'll see like that one's kind of interesting that one on the top
119. 17:55left there mid-journey's pretty good with the fashion stuff that dude is intense there's I don't like the color of that that
120. 18:04one's cool all right so so there's our images so let me show you a couple other things you can do with mid-journey which are kind of fun
121. 18:12so when you create a set of images it gives you four options and then you've got these four buttons U1 through four
122. 18:17and V one through four V means variation u means upscale and then and then that redo button is a redo button so let's
123. 18:24take I guess uh let's do the the yellow one that one's I don't know uh let's let's do the third one so it goes one
124. 18:31two three four so I'm gonna do V3 and I'll do V1 just for the heck of it so that's going to give us four
125. 18:41um alternates for each one and three and it should do that pretty quickly we'll see all right so here we go this is the yellow one
126. 18:57obviously you can see a yellow blob there and I mean this is kind of fascinating stuff where you know it starts out
127. 19:04completely blurry and like now this he's got like a face behind him here it'll be interesting to see if that turns into a
128. 19:10face he's got a face behind him not sure what that's about maybe they're like billboards that's wild
129. 19:22not quite sure why that happened but you know that's part of the part of the joy here is just experimenting and figuring
130. 19:27stuff out and seeing what works all right so that one is really creepy let's go with the creepy one so I'm
131. 19:37gonna upscale number two here just because it's creepy and if you're asking why did it put the picture back there I
132. 19:43don't know because it did and then you can do things like tell it not to put other faces in the picture
133. 19:49and things like that but I'm going to hit YouTube fine band fine fine band and now it's going to upscale that
134. 19:59and then there's going to be some one other cool thing I'm going to show you which is the unzoom feature so part of
135. 20:04mid-journey is this new zoom out feature or not unzoom zoom out Okay so here's our here's our high-end thing
136. 20:14let's go look at it that's pretty nice this dude's head is way too small he's out of proportion so that's that's kind
137. 20:22of comically bad but that's okay we'll go with it all right so now let's say so so I did this Photograph
138. 20:28and let's say that I was hired to make a vertical photograph but all of a sudden I realized oh crap I need a landscape
139. 20:36photograph I need this to be wider than it is tall and I can't really crop in on him because it would just be too weird
140. 20:41so I can do zoom out two times zoom out 1.5 times or or do a custom Zoom so I'm going to do a custom Zoom
141. 20:49and what I'm gonna do is I'm gonna do aspect ratio of 16x9 which will give me a good cinema kind of thing and I'm going to do Zoom 1.5 so
142. 21:03I'm going to zoom out 50 beyond what we've got now but I'm going to do it in 16x9 and then I'll hit submit and so
143. 21:10this is now going to give me three different or four different versions of this that basically fill in the
144. 21:17background that doesn't exist right now so we just created that image out of nothing and now we're going to create
145. 21:23um a completely different aspect ratio from this out of nothing you can also do this in Photoshop in the Photoshop beta
146. 21:31right now with a thing called generative fill and you'll you'll notice here the part of the image that was in Focus
147. 21:39before is still in focus and we've still got this weird blurry stuff in the background so as these images develop
148. 21:45that'll come more and more into Focus that looks like Macy's right there that's kind of cool that makes that
149. 21:51thing look much more like a billboard than a creepy dude looking over his shoulder it's given us some nice perspective with
150. 21:59these there's a Yellow Cab right so we know we're in New York City right now we're at 62 percent we're going we're going Bueller Bueller
151. 22:13any bonus points for anyone here who's a gen xer laughs because is that is that done that we're at 93 percent
152. 22:23and then okay so let's look at these so oh that one's kind of cool this is like in the the meat packing District in
153. 22:28lower Manhattan with this with the brick streets um I think that's the one I like that kind of looks like Soho a little bit and
154. 22:36Brick Street so let's do that so we're gonna upscale number one this should go pretty quick and there you have it within
155. 22:47within a few minutes whatever however long this has been 10 minutes of me doing this you know we went from I don't know
156. 22:54anything about fashion to there's our fashion dude our New York City shoot so you could see that if if you actually
157. 23:01had an eye for this and you knew the terminology and you knew what you wanted to do in in very short order you could
158. 23:07put together you know at a minimum mood boards right uh potentially even significantly beyond that so let me flip
159. 23:15my camera around welcome everybody my name is Kyle Shannon this is the AI learning lab uh I am learning this stuff
160. 23:22kind of in real time while talking about it with y'all if you have questions about AI pop them in the comments below
161. 23:28I do my best to answer everything um I'm an entrepreneur by training I've started 12 or 15 companies in my day and
162. 23:36that's why we're here so uh do that I'll talk talk about stuff below and let's get moving on what is the bot name that
163. 23:48created that prompt am I toward the bottom where am I did I lose my place again the the bot name was uh
164. 23:56MJ MJ Captain so if you go to poe.com MJ captain or poe.com mid-journey both of those are Bots that just regular old
165. 24:08people like you and I created on poe poe.com poe um you can make your own bot you could make your own mid Journey bot if you
166. 24:17like figure out an interesting way to do mid-journey prompts you could turn that into a bot so someone gives you a piece
167. 24:22of crap prompt and you turn it into something good that is in your power I'll show you how to do that in a bit
168. 24:29let me go find out where I was in my little questions here and so because there were probably there's a lot of questions here thank
169. 24:37you all for participating I really appreciate it all right here's where I was Dolly too is not bad for visual creation
170. 24:47dolly two is okay if you go to bing.com create that's dolly two so that's open ai's image generation tool um
171. 24:56of all the inner image generation tools I think dolly is is kind of fourth from a quality standpoint um first is mid-journey in my opinion
172. 25:06second is stable diffusion stable diffusion is way more um there's a much healthier developer community so there's lots more features
173. 25:14but you got to be a [ __ ] geek to use it number three is uh Adobe Firefly there are image Generation stuff is
174. 25:23really good and it's all commercially safe um and then they've got generative fill within Photoshop that's really good and
175. 25:30then Dolly if you go to bing.com create that's Dolly and that's pretty good you can also call Dali from API calls
176. 25:39um I'll even come back okay beautiful I can never evoke a reaction out of you am I just not funny oh Eric wait what did
177. 25:47you say did you crack a funny joke that I missed I don't know I don't think you did I don't think you did a funny Eric I don't
178. 25:55think it's true wait oh no that was to RM Walter or good actors for that matter oh I get it it was an okay joke Eric it wasn't
179. 26:06great it was okay now Echo New York you're good thank you very much um yes you should sound like God yeah yeah I'll turn on my echo
180. 26:18saw I am Thor no you're not dolly two is not good I'll even come back okay here we go um I'm like 90 AI LOL you too
181. 26:31um what Masters should I do in AI uh I have no [ __ ] clue what you're interested in um if you're interested in the Science and
182. 26:43Technology and Engineering side of AI I don't know go to Stanford go to MIT there's there's probably lots of places
183. 26:49to go if you're interested in all this generative AI stuff there's not a masters in it right now this is all too
184. 26:55new this is all it is evolving so fast no one knows anything do I love Denver we almost relocated from Irvine
185. 27:03um Denver is beautiful the people are super nice the weather's amazing um I miss the energy of the East Coast I
186. 27:13miss New York I've been out here for 14 years and um I miss Snappy people you know when you like so so here's the thing when you're
187. 27:23in New York and you're like hey I got this idea no matter who you talk to it could be on the subway it could be a
188. 27:29stranger it could be your best friend could be someone at work you're like hey I got this idea what happens in New York is you get
189. 27:36instant feedback oh that's a piece of [ __ ] or oh that's really good hey you know it would be really good
190. 27:42there's this there's just this drive this energy of like let's [ __ ] go let's [ __ ] get there in Denver I'm like hey I got this idea
191. 27:50oh that's nice it's just like the energy just goes hmm out of the room it's really hard to have an elevated conversation here and it
192. 28:02drives me [ __ ] batty but I do like living here I just I miss that energy I missed I didn't realize how much
193. 28:12that kind of back and forth and that energy to just [ __ ] do it the best how much it drove me it drove me a lot
194. 28:19so I miss it hard I miss it hard that's so hard well why don't you get up and leave Kyle I don't know because I like complaining
195. 28:31what could I complain about if I had everything I wanted huh um T is hype Google of Google okay gumba Limba
196. 28:47you clearly haven't used it so whole purpose of this channel go there launch that make an account if you're too much of a [ __ ] to give
197. 28:57them your phone number because they do ask for your phone number then go to that one go to bing.com and click on the
198. 29:02chat button and start playing with these things and if you don't know anything about prompting or what to do with them
199. 29:07or if you can't get it to do anything beyond what a Google search does then go here to prompts.chat that will teach you
200. 29:13about prompting and it'll start to give you it'll this document will help you start to scratch the surface of what
201. 29:21that thing [ __ ] makes possible this is some of the most remarkable technology ever put in the hands of mere mortals
202. 29:31this this represents like Decades of machine learning development put into a package that is if you use it right indistinguishable
203. 29:45from Magic that's an Arthur C Clarke quote I think it's Arthur C Clarke any significantly advanced technology is
204. 29:55indistinguishable from Magic if you think Chachi PT is just hype you haven't used it go there I honestly honestly I
205. 30:02feel like everyone owes it to themselves to go use this thing long enough to have a Kevin McAllister moment which is it's
206. 30:10a U.S film for uh called home alone and the little kid was left alone and he had this moment you will have that moment if you play with this
207. 30:21if you sort of get out of your way and stop judging it and stop diminishing it all right years ago when AI started
208. 30:27running the financial system things got really bad well well yeah so so that an important thing to note ai's been around for [ __ ] decades
209. 30:38the difference now what the inflection point that we have passed that that started on November 30th 2022 six months ago is Chachi PT
210. 30:49made machine learning accessible to the rest of us right I don't disagree that a lot of the AI implementations that have been done so
211. 31:00far haven't been great haven't been as successful as predicted you know uh Watson from IBM their Watson Health Initiative kind of flamed out
212. 31:11um so so yeah there's some things it does really good there's some things that does really bad but this and and
213. 31:20you know the similar tools like it and how fast that got adopted by just regular people and business people is going to transform
214. 31:28the way we all work so so again don't just sit back and wait for this to get better jump in right now ideally IA will reach a place where it
215. 31:42can defend itself against Bad actors yeah hopefully so and I thought listen I think that there's going to be lots of
216. 31:47efforts to do that I think I think at this point there's enough awareness of the good and bad potential of this that
217. 31:54people are now aware of it so I it's it's looking promising how would you use AI to scale up Productions and help with onboarding I
218. 32:03don't know what kind of Productions like like film Productions um and help with onboarding um you can have it help you put together
219. 32:11project plans it's really good at taking other documents and understanding them if you're doing Film Production uh if you use the
220. 32:20um here let me let me show you something that's kind of cool this this may not be at all what you do but I think this is
221. 32:26something worth worth sharing um oh that's funny that dude our little fashion dude I'm a fashionista look at me and my
222. 32:38yellow jacket all right uh that's pretty funny well it was funny to me okay so I'm gonna go to post so if you go to poe.com
223. 32:49when you first get here it will look like this where it's just it's just like uh like a Chachi PT window down below you can type in a
224. 32:58prompt and it'll it'll give you some stuff up at the top there's explore you can go explore Bots and there's create a bot if
225. 33:05you want to create a bot you can do that there but these first six choices here sorry there's explore there's creative
226. 33:12bot these first six choices here um are are different chat CPT models basically Sage is Poe's model gpt4 Claude plus and Claude instant 100K
227. 33:24those are all subscription only it's like 20 bucks a month gives you access to those three and then you get Claude
228. 33:29instant and chat GPT so the clawed ones are from anthropic Sage is from Poe which is quora and then Chachi BT and gbt4 are from openai
229. 33:40this one right here Claude instant 100K what is that 100K mean with normal chat gbt you have what's called a a 4 000 token context window
230. 33:51it's increasing and for some people it's 8 000 now but for the most part it's four thousand what that means is three
231. 33:57thousand words you can you can have a context window essentially a chat history of about three thousand words
232. 34:05this hundred thousand is a hundred thousand tokens that's about 75 000 words so I I yesterday I was trying to find a
233. 34:12screenplay in PDF format and I couldn't but I found one today so this is the Matrix this is the screenplay for the
234. 34:19Matrix okay so screenplays are what 120 pages I I assume this one's longer than that 124 126 okay so slightly longer than normal
235. 34:34can I copy please tell me I can copy here yeah okay so fade in so I am going to copy this entire screenplay so I'm going
236. 34:45to hold down my shift key did that why did that not eh what's going on uh this is very weird
237. 35:00um okay this is the shooting script the numbered
238. 35:15shooting script okay there it is okay so now I've copied all of it I didn't get all the fade in but I got it all
239. 35:24I'm gonna go to Claude instant 100 000. and I'm just going to come into this little chat box I'm going to say based on this screenplay
240. 35:35and then I hit shift return to do some line breaks and I'm going to paste in all of The Matrix so let's say you're you're uh in a
241. 35:46production thing uh or you're a scriptwriter and someone says hey go read the new version of the script let's say um describe the main
242. 35:56characters and what happens to them question mark and so now it's gone off and it's red the whole of the screenplay
243. 36:12and it's we're waiting it's doing some processing still waiting the main characters in The Matrix are
244. 36:22Neo he's a protagonist the story follows this journey from being an ordinary person living in the make Matrix to
245. 36:28realize he is the one Morpheus Trinity agent Smith agents sentient computer programs sent by machines in summary
246. 36:37these are the key events the key events are Neo realizing he's the one undergoing combat so now if I'm in production again I don't know what
247. 36:44production you were talking about I could say please provide me a detailed prop list for Neo in the First Act
248. 37:00of the film we'll see what it does see if it gives us that I didn't train it with a prop list was I don't even know if it knows what a
249. 37:12propolis is but I assume it does because these models have betrayed have been trained on essentially all of the internet
250. 37:20here's a detailed prop list for Neo in the First Act of the Matrix home and work computer equipment Neo works as a
251. 37:27programmer and a hacker so he has numerous computers monitors Towers keyboards clothes at the beginning he
252. 37:32dresses dresses like a typical late 90s clothes for a hacker baggy jeans t-shirts hoodies so there you go um what are the first
253. 37:47uh 10 locations and could any of them be re -dressed versions of the same location I don't
254. 38:03know if that's a thing again but it will probably figure some [ __ ] out for us um
255. 38:17here are the first 10 locations subway station office virtual reality space as you can see several of the locations
256. 38:29like hotels Office Buildings and subway stations could utilize redressed versions to represent similar locations
257. 38:35within the confines of the Matrix so that is from Poe using the Claude 100K thing you could pop in any sort of large
258. 38:46document like that these models are incredibly good at analyzing big documents and helping you do that kind of stuff so hope that helps
259. 39:05um by the way if you're new here welcome my name is Kyle Shannon this is the AI learning lab we'll sing Hamilton songs later we do
260. 39:12sing-alongs every night and uh these URLs behind me that's not true by the way I'm a smart ass that is true
261. 39:21uh if you're new to generative AI if you've not played with chat GPT that chat.openai.com that's the official
262. 39:28chatgpt website go there you owe it to yourself to go there um bing.com this is Microsoft search engine but it's also chat GPT if you
263. 39:37click on the chat button that's gpt4 for free po I just showed you it's got a bunch of things there including that
264. 39:44hundred thousand token version and then prompts.chat is um a document that will teach you about prompting and it'll give you a bunch of
265. 39:52ideas for personas that you can give chat GPT to act like act like a screenwriter act like an accountant act
266. 39:59like a counselor and it'll do that and then futurepedia.io is um just a big directory of like thousands
267. 40:08of different AI tools most of which are horrific pieces of [ __ ] that you shouldn't waste your time on
268. 40:14but many of which are so mind-blowing that you won't be able to get your head around what in the hell is coming good luck
269. 40:26how about Boomer points Boomer points are okay you know you got your Boomers you got your gen xers we get ignored in
270. 40:33most of the lists we're cool with that we're bitter we're better at you Boomers you [ __ ] us up all right but you made us strong
271. 40:48by the way if you haven't seen the bear turn me off and go watch that [ __ ] um episode six and seven of season two
272. 41:03holy [ __ ] some of the best television I've seen ever if you haven't seen the bear at all shame on you all right kind of toward
273. 41:13the bottom I'm new to Ai and an elder Millennial welcome welcome you're on the cusp huh um do you know any free AIS that can
274. 41:22create and edit vids not really edit okay so if you want to do editing there are probably some AI tools that do some
275. 41:29you know chopping and snipping and things like that so I would say go to futurepedia.io and um just look for the video tools
276. 41:40um like I said there's going to be a bunch of them in there that are going to want to charge you a lot of money and
277. 41:45they're big pieces of [ __ ] so just you know go in knowing that if you want to generate stuff there's a couple of uh
278. 41:54things out there right now probably the one that's worth experimenting with is Runway ml so go to run just Google
279. 42:01runwayml as in machine learning and then they have an app but they also have a Discord server and they've got this
280. 42:07thing called gen 1 and they've got one called gen 2. so gen 1 you can upload a short video clip and then give it a
281. 42:14prompt to basically use the action from that video clip but give it a completely different visual style that's pretty
282. 42:20cool and then Gen 2 is just you type um a video scene and it will make a video um that's relatively new technology if you
283. 42:32use those tools and you're and and your thought is um these are disturbing and creepy so were all the image generation tools a
284. 42:42year ago so a year ago these like mid-journey and stable diffusion the images that they created were kind of
285. 42:48they weren't laughably bad but they were disturbingly not good that's where the video is today so just sort of imagine it a year from now
286. 42:58um and it will it will get really good so hopefully hopefully that helps all right can you zoom out from an image
287. 43:03that's not AI generated yes you can let me go show you that we'll do it in uh Photoshop [Music] all right I am gonna go
288. 43:18so here's Photoshop uh this is by the way to get to this this was the thing I did this is an AI generated pendant and it was
289. 43:30it was a still life with some plants behind it and I did some generative fill to replace the plants with a with a person
290. 43:38um it kind of messed up the face the face is weird but but uh it was interesting so let's see let me flip over to finder
291. 43:46and go find a photo photo graph what's that that's a video okay here's okay here's a bad photo of me so let's pop this over to to photoshop
292. 44:02did it bring it in oh I see it brought it in as a um so I'm going to cut that I'm gonna go new all right let me do that
293. 44:16oh I see what it did hang on a sec new paste damn it oh
294. 44:37hang on ah I'm trying to move too fast here people bear with me all right there we go all right so so there's my own dumb fat head
295. 44:51hey there my name's Kyle um and let's say let's say we want to we want to make a video frame so so we'll move this I don't know kind
296. 45:05of there and then we'll grab the crop tool [Music] oh bottom purple that damn meow that's
297. 45:36um all right so let's say we want my my dang head in some sort of rule of thirds thing like that so what I can do is I can just grab a
298. 45:49little photograph of myself here I'm I'm selecting that let me make it a little bigger so you can see it
299. 45:55so you can't really see the selection tool but basically I've selected I've selected the photograph of me and now
300. 46:01I'm going to invert that selection so I'm now I'm selecting all the white stuff in the background and you'll
301. 46:07notice there's this little contextual menu thing here that's part of the new beta and that what that button says now is is generative fill
302. 46:16so I can either just hit return and it will sort of figure it out or I could type in modern um office with
303. 46:27uh impressive white board and windows and then I'll hit generate and oh I've got a quit Photoshop oh my God
304. 46:43why why all right let me say this let me save this image we'll do KS test where am I saving at desktop all right it's gonna be so good
305. 46:58what'd you do tonight I don't know I watched some guy try to launch Photoshop that sounds awful it was kind of bad
306. 47:07okay so I'm relaunching Photoshop beta which the code name was Strawberry Letter apparently okay so we're back
307. 47:19here we're gonna make it big again I'm gonna select this photo again and we lost almost everyone in the chat
308. 47:33this is the die hards now we're gonna invert the selection and so now I'm going to do generative fill I'm going to say modern
309. 47:43office with impressive whiteboard and windows and we're going to say generate and so now it will go generate some stuff
310. 47:58time to update oh I guess there's a new build of of this version of Photoshop so there you go so that is an office and
311. 48:13it'll give us three versions of that and so it gave me a big giant shoulder like the proportions are all messed up
312. 48:21there but you know it's making it out of nothing right so it it kind of got the glass windows right it got the the Whiteboard here um let me
313. 48:32let me delete that layer and then let's go back here select this we'll hit
314. 48:48inverse and I'm just going to hit generative fill but I'm not going to type anything in we'll just see what it does on its own
315. 48:55see if it does anything interesting if it's more or less better oh give me an office mate
316. 49:13he's got some sort of Lego toy or something there that's actually not bad some bikes although it's really janky back there
317. 49:21that's actually pretty good that you know again it's sort of sort of messed up the uh the shirt a little bit here but but
318. 49:27that's generative fill so yes you can do that with any image you can also completely replace the image and just generate it wholesale um
319. 49:40all right let me flip yarn camera back around hello everyone I'm back if you have questions about generative
320. 49:56AI pop them down below I will do my best Denver is boring I was just there Brooklyn is great that's why it costs way more
321. 50:03yes I agree David Horton I concur how do you answer the questions about AI from audience in live sir well Jesus Christ is my savior I read a
322. 50:18question like the one you wrote and I talk words out of my mouth I don't know and if I can't answer them I won't answer them
323. 50:27all right lol that's why I feel like I could never leave Miami yeah exactly I I know I need to get back to the east coast
324. 50:36that low-key energy happens in Alaska too people want to be polite instead of helpful or honest agree fun fact if you
325. 50:43Google Google you break the internet I like it fun fact if you delete your recycle bin you break your computer
326. 50:49that's good Eric that was that see that was good that was clever oh you did the other fun fact Eric's in
327. 50:56here Eric needs to be a stand-up Eric start a tick tock I think you could do this I've been working on SMS version but
328. 51:05needs some feedback okay cool I'm happy to play Gen X here welcome Gen X Gen X crowd it's crazy good I don't know what you
329. 51:15mean I assume you meant mid-journey I used AI today good good use it share it go to chat.openai um do me a favor since we lost a bunch of
330. 51:27people when I was trying to run Photoshop uh tap the screen give me a bunch of likes and maybe they'll let
331. 51:33some people know we're in here having a good time they'll want to come in and play and watch an old guy try to use tick tock in his computer
332. 51:43March how do you turn this thing on Marge the computer's Off Of You the screen went black do we gotta take it to the Apple Store mark
333. 51:56I love chats EBT I want to be able to create a digital package of top prompts uh you should be able to do that in
334. 52:02about 10 minutes have it write it for you um how do you answer questions I answered that I answer questions by answering
335. 52:10questions is it is Magic if you can't distinguish the difference between something and something Kevin McAllister
336. 52:17moment you realize Mom and Dad left for France without you but that's not that he realized they left him that's that he put on the dad's
337. 52:27aftershave but still good moment um I'm surprised by what chat EBT is capable of every single day you and me both
338. 52:37I like I've spent the past two days working on um a big proactive project for a client that I would not have done like
339. 52:48what I did for the client is way above and beyond what's reasonable and I put in like an hour or two but for
340. 52:56me to do what I did traditionally would have taken me two or three days and I wouldn't have done it and and it's
341. 53:04it's it's astounding it's astounding like you have a big long document and you're like you know we should kind of expand the focus and
342. 53:13add in this concept and that concept you can take that whole document paste it into chat GPT and go hey I wrote this
343. 53:21I want to add these two concepts Focus Less on a and more on B give me a draft out it comes and it's good It's [ __ ] amazing it's just
344. 53:36it's just amazing where do you find the best prompt I am I am if if you're a really organized person which I'm not I I have ADHD like like
345. 53:48the laminated card version like they're really the really good version the good kind um so my uh prefrontal cortex is just a
346. 53:56block of cheese so I have no executive function so if I were good at that [ __ ] I'd have like a whole spreadsheet or a
347. 54:02database or a notion set of notebooks organized in 16 categories but I don't um there's prompts.chat will give you
348. 54:12some there's there's a bunch of different prompt libraries there's one you can pay for them I wouldn't recommend that at all
349. 54:18um I would really just recommend start getting your head around here's the biggest thing that made a difference for me with prompting is
350. 54:27you don't have to have the perfect prompt what you actually have to have is the right mindset and I know this is going
351. 54:34to sound like weird and creepy when you use Google here's what Google has trained us for the past 20 years Google has said
352. 54:42write the most concise version of what you want to find as possible don't put in extra words don't don't confuse me or I'll give you wrong
353. 54:51[ __ ] right so so here's the just give me the [ __ ] you want to search for and then I'll give you the 40 choices
354. 54:58chat jpt kind of works the opposite where if you give it very little it will oblige and return very little it like
355. 55:05it'll just return generic [ __ ] but if you give it a lot of context like you say hey I want you to act like an expert
356. 55:12historian for Egyptian hieroglyphics and you know I'm a student of egyptology and and then you tell it kind of who you are
357. 55:22and what you're trying to do it will oblige you and it will get better and better and better and it's just the nature of how the tool Works
358. 55:30um so I I would strongly encourage you to just really start thinking of chat gbt more like a like a think of it like
359. 55:38this chat is the smartest intern you've ever hired um that needs an incredible amount of guidance oh and by the way it lies a lot
360. 55:49it will lie to you but you give it you give it lots of guidance and and sort of think of yourself as the pilot it is the
361. 55:56co-pilot and and just start playing with it like that and you'll kind of discover your own kind of prompts um
362. 56:03but there are lots of tools out there but but I just copying and pasting prompts isn't going to teach you as much
363. 56:10about how the stuff actually works I've Incorporated chat gbt into my daily workflows so much so that I subscribe to
364. 56:16the service that's awesome Albert yeah it just speeds things up chat is amazing there's nothing natural about natural
365. 56:23language processing learn prompt engineering yeah it's I think learning learning to prompt right now is is a
366. 56:30valuable tool I think in two years it's all this stuff's gonna drift into the background it's going to be much more
367. 56:36automated and if you want to see an example of that there's a plug-in for chat apt called prompt perfect where you
368. 56:42type the word perfect before a prompt and it will rewrite it for you kind of like the mid Journey stuff I showed earlier
369. 56:49um so so yeah chat GPT is amazing and currently dictating my waking hours it dictates mine it sometimes wakes me up thinking about it too
370. 57:00I'm like oh I could do this with it cyber security of the Future Good robots and bad robots fighting on the internet
371. 57:09I think that's right I was talking to um a a group of uh people that work for like a foundation you know they write grants
372. 57:18and read grants and give out money and [ __ ] like that and I was explaining chat GPT and this woman goes well wait a minute if
373. 57:26if it does All That Couldn't someone writing a grant proposal just write it much more quickly and like yeah they
374. 57:31could and then she goes well then couldn't the people reading those Grant proposals read them much more quickly
375. 57:38I'm like yeah they could and I said you can imagine a future where the robots do the grant writing and the grant reading
376. 57:45and then I said and then what's left what's left is you get to talk to the human beings right so kind of there's a
377. 57:51very good chance that these Technologies automate out all of the soul-crushing time-consuming repetitive
378. 58:01just you know exhausting drudgery work they automate all that [ __ ] and we get to return to connecting as human beings
379. 58:12it might return us some empathy and some human connection and things like that that's where I'm hoping it goes so even
380. 58:19though cyber security of the future was meant as a joke Eric which as a stand-up you got some work to do on the writing
381. 58:26but I like the spirit um I think that's actually not too far from where things could go um don't trust it because it learned
382. 58:34good info and bad info more bad than good well sure possibly but if you understand how these things work there's a there's it's
383. 58:43worth Googling there's a YouTube video of Andre carpathi from openai speaking at the Microsoft developer conference
384. 58:49talking about prompting and how the prompts work and how the tools work and one of the things that he talks
385. 58:56about is he uses an example of a math equation and if you go to chat GPT and you're like here's a math equation go
386. 59:06solve that math equation for me what it will do is it will go out and look at everything that's been trained
387. 59:12on which is essentially all of the internet right all the public internet and what he talks about there is in all of that data there's
388. 59:22this professor that solved that equation correctly and that information's in there and then there's this student that
389. 59:30didn't solve that equation correctly and then 30 fellow students you know 15 of which didn't solve it correctly so if
390. 59:38you just say go solve that math equation it's going to look at all of what it's learned related to that math equation
391. 59:44kind of amalgamate what it's learned and give you some sort of answer it's just spitting out individual words so it
392. 59:52looks like it's solving the equation but that equation might be wrong because you didn't narrow the search
393. 59:58because you just said go solve this math equation but if you tell it I want you to act like a Stanford math professor
394. 1:00:04that specializes in this branch of mathematics and here's this equation and I want you to not only solve this
395. 1:00:11problem but I want you to show me the order of operations of how you did it what you're doing is you're increasingly
396. 1:00:17narrowing the What's called the latent space of the information that it's looking at and you have a much higher likelihood of
397. 1:00:25getting the right answer so don't just don't diminish what it does and what it makes possible without really
398. 1:00:35understanding how it works because otherwise you'll you'll you're on the risk of getting left behind because this
399. 1:00:40stuff's coming really really fast yo make it rewrite the Matrix with Batman and the Joker we can do that um all right let us go
400. 1:00:57let me flip the camera around so if you're new here what I was showing earlier was um if you go to poe.com and if you're a subscriber
401. 1:01:10there's this thing called Claude instant 100K which is a large um context window chat gbt clone basically where I can put in up to 75
402. 1:01:21000 words so the 75 000 words I chose to put in here are the the screenplay of the Matrix right so so there's The Matrix
403. 1:01:30and then I was asking it about plot and characters and things like that so what did you say rewrite with Batman wait where did it go
404. 1:01:46Tick Tock just blew me out of here damn it ah oh here we go yo this is from Eric Eric you're very you're very peppy tonight
405. 1:02:02make it rewrite the Matrix with Batman and the Joker okay so I'm gonna say re write the oak the opening scene
406. 1:02:17um of a version of this screenplay that also features Batman and the Joker
407. 1:02:33ain't better waiting waiting we're waiting we're waiting I'm sorry Dave I'm afraid I can't do that
408. 1:02:49here's my attempt computer screen night computer pulses on dark computer screen a beating heart a phone begins to ring the cursor as the
409. 1:03:01trace program activates yeah data flashes on screen caltration initiated we recognize the voices as Batman and
410. 1:03:07the Joker The Joker good evening Batman I thought I'd give you a call Batman where are you Joker the trace Narrows down the phone Joker
411. 1:03:16you know we're around and on and on and on um let's see um what uh what would be the log line of this new
412. 1:03:32movie so the log line is the little like the Hollywood pitch [Laughter]
413. 1:03:49here's a possible log line after discovering he's actually living in virtual reality created by artificial
414. 1:03:53intelligence a hacker must harness his newly humanized super human abilities to rescue his mentor from the clutches of
415. 1:03:59Agents while also confronting a deadly chaos Unleashed in Gotham's most notorious super villain the key plot elements Neo discovering he
416. 1:04:08actually lives in the virtual reality known as The Matrix gaining superhuman abilities in The Matrix his mission is to rescue Morpheus from
417. 1:04:16the agents the inclusion of Batman and his Pursuit Of The Joker who's unleashing the log line hits the major plot points so I'm going to say
418. 1:04:25um it seems let's see it seems like um you just duct taped the Batman characters
419. 1:04:42onto the existing plot can you give me an original idea where it is more fully integrated
420. 1:04:59question mark so so one of the tricks with these tools is if it gives you shitty work tell it to give you shitty
421. 1:05:06work and tell it what you want better I mean that's the kind of the role of humans in these is you know you're the
422. 1:05:11director you're the conductor you're the pilot it's there to serve you neo Discovery lives in a virtual reality known as The
423. 1:05:21Matrix however instead of facing mundane agents Neo encounters some of Gotham City's most infamous villains within the
424. 1:05:28Matrix who are working for the AI so let's see rescue his mentor Morpheus who's been captured by the Joker
425. 1:05:38the Joker's testing dangerous Joy buzzer devices on Morpheus that threaten to drive him insane if activated takes the role of Nightwing
426. 1:05:50so yeah so there's there's a there's a version of the plot that's actually decently integrated so yeah you can even though you may have been
427. 1:05:58joking there Eric see we can go on the internets with that chat GPT and we can make some [ __ ] happen we can rewrite some screenplays
428. 1:06:09all right uh welcome everybody if you got questions about AI popping below I'll do my best to answer them could I connect
429. 1:06:17this via API to my bubble app yes you can make all sorts of cool [ __ ] with bubble and the open AI apis
430. 1:06:25they are in there what the heck and I am having issues finding a job wait what the heck and I am having well
431. 1:06:33I start playing with this stuff and talking about it everyone that I know that is actively playing with AI and
432. 1:06:41putting it out there that they're playing with it is having um good luck getting jobs in fact a good buddy of mine
433. 1:06:50this is crazy the guy that I work on a newsletter with he just got a major new job um largely because of the stuff the
434. 1:06:58newsletter that we're doing on AI he's like he like a massive increase for a big public company is this answer a copy
435. 1:07:07and paste of someone else's answer I don't know what you're talking about possibly is what answer I don't know what you're talking about
436. 1:07:16um do you pay for the Claude instant no well Claude instant 100K you have to pay for Claude instant you don't and Cloud instance really fast
437. 1:07:25Claude plus gpt4 and Claude 100K you have to pay for on Poe um Sage Claude instant and chat gbt you don't have to pay for it and the Bots
438. 1:07:36you don't have to pay for what production work production for employees I think is what he meant yeah I don't if I don't quite know what his
439. 1:07:45question was the Bing AI is better in your opinion that's interesting I I the the gpt4 for for me here is not as
440. 1:07:56good as gpt4 and open AI but if you find it's better that that's the thing about that's why I have these different models
441. 1:08:03here different people have different ways of approaching this and if Bing works for you like listen that's gpt4
442. 1:08:09connected to the internet and it's free so if that's better for you [ __ ] more power to you the whole point of the
443. 1:08:14channel is just if you haven't played with these things go play with them go play with them until you have your Kevin McAllister moment
444. 1:08:23then you can decide it's all a bunch of [ __ ] and you're not going to deal with it but I don't think you will
445. 1:08:28because everyone that I know that has gotten their head around this thing says holy [ __ ] it can do that
446. 1:08:37and then they're Off to the Races even Snapchat is a pretty good AI everything's gonna have a pretty good AI
447. 1:08:42pretty soon I use mid-journey to copy and paste same stuff you wrote down will AI show me the picture and person wait if I use no so so
448. 1:08:55there's a couple of things if you use the exact same prompt it will give you four different images and then
449. 1:09:03if you use this exact same prompt again it will give you the same four images you can find this thing called the seed
450. 1:09:10which is a long number basically that is the seed of that particular image so if you use the same seed as I used it'll
451. 1:09:20get close but it'll still be kind of new and unique the nature of how these things work is they're not copying and
452. 1:09:27pasting elements from other people's images what happens is when they when they learn all these images they they put
453. 1:09:37them into what's called a diffusion model where they they turned every image in the database into random noise they
454. 1:09:43use a noise generating algorithm and then they combine a large language model which understands your words and it sort
455. 1:09:51of points to all the images that have been tagged with those words some it's basically some version of that and then
456. 1:09:57it kind of averages all these random noise images together into a single random noise image and then over 30 to
457. 1:10:0570 steps it denoises that image into some new original piece of work so every time you render it unless you use the
458. 1:10:13same seed it's going to give you a new seed a new randomized thing and a new image um and that's that's one of the challenges of it
459. 1:10:23um is that it can be hard like if you got something that you love it can be hard to replicate that same thing in
460. 1:10:30different areas using things like um zoom out and generative fill within Photoshop you can improve that there's a
461. 1:10:38thing with stable diffusion called control net that can control things a bit better but that's one of the things
462. 1:10:45it's not great at is is consistency and replicability but in terms of generating new stuff it's [ __ ] brilliant brilliant
463. 1:10:55brilliant Brown yeah all right uh you don't need to ask questions three or four times I'm very far behind on the questions so uh
464. 1:11:06remind me where I find prompt perfect please okay so let's go show let's go show some prompts and I will show you
465. 1:11:17prompt perfect okay so we're gonna go to chat GPT you can make money okay so if you've not used chat GPT this is your prompt history
466. 1:11:28um if you pay the 20 bucks you get this little yellow icon and that's um chat EBT plus which then gives you
467. 1:11:36access to gpt4 with like the normal version the Internet connected version and plugins and I'll show you these in a second
468. 1:11:45if you pay your 20 bucks and you don't see these options you actually have to turn them on because I don't know why
469. 1:11:50they're doing that because they're [ __ ] I I don't know I think it's a bad decision right now but it's probably just a processor thing
470. 1:11:58um you go to your settings and then in your settings there's a option called beta features and then you turn on
471. 1:12:05browse with Bing and plugins and so that's important you got to turn that stuff on and then once it's turned on we're going to go to plugins
472. 1:12:13and now you have this little drop down menu and the drop down menu is here's all the plugins that I've got installed
473. 1:12:20which basically these are the ones I pulled out of the the plug-in store that I said these are these are the ones that
474. 1:12:27are interesting in any given chat session I can enable up to three of them so let's say I want to do something
475. 1:12:33about travel or food you know I would turn on my travel or my food plug-ins and then I would have kind of
476. 1:12:41like a little customized application that can do chat GPT stuff but also can use the plugins and do whatever they do
477. 1:12:48so the one that's called prompt perfect is a plug-in so you just turn that on and then the way it works is you you
478. 1:12:56type the word perfect as your first word of your prompt and then you type a crappy prompt and it will rewrite it for you so let's say um
479. 1:13:04uh let's say create a meal plan for me for the week that isn't boring uh but
480. 1:13:19is keto friendly all right that's kind of a crappy prompt because it's generic and whatever and so because
481. 1:13:28I typed the word perfect in front of it it fires off prompt perfect and now it took my crappy prompt and it
482. 1:13:34rewrote it please design a weekly meal plan that adheres to the keto diet ensuring it is both interesting and
483. 1:13:40diverse with a focus on providing a detailed and comprehensive response cool and then so it now gives me that
484. 1:13:53and actually let me let me show you another thing I'm going to stop generating here I'm going to go copy that initial prompt
485. 1:14:02and I'm gonna go turn on two plugins because this this kind of cool to to watch this I'm going back to plugins so
486. 1:14:09I've got prompt perfect there and I'm going to turn on instacart because we were doing a meal plan so now I'm gonna so now I've got
487. 1:14:21prompt perfect and instacart so I'm going to take that same prompt I start with the word perfect create a create a meal plan
488. 1:14:28so it should fire off prompt perfect it did well there it rewrote The Prompt it's not the same as it did before this is
489. 1:14:43specifically designed to follow the ketogenic diet for an entire week the plan should prioritize variety and avoid monotony
490. 1:14:52so there's day one day two day three day four day five day six day seven and notice what it did here at the end
491. 1:15:03would you like me to create a shopping list for these meals using the instacart plug-in so all I have to do is say yes
492. 1:15:11it knows the plug-ins in there it knows it can create me a shopping list because of how that plug-in was described internally
493. 1:15:18and now it will probably fire off the instacart plug-in it should boom there it goes and now it's going to give me a link to
494. 1:15:33that shopping list and I could have those groceries delivered to my door tonight champ
495. 1:15:50come here stop [Applause] did it do it all right there's your instacart shopping list so if I go click on that Bang
496. 1:16:08and there's all my stuff right so from a single prompt it made me a whole meal plan there's all the ingredients
497. 1:16:19pop those in my cart have them delivered Bing Bang Boom Off to the Races isn't that nuts nuts freaking nuts
498. 1:16:31and and you know the future of these plugins the plugins right now are really kind of crappy and janky and the interface is horrific um but
499. 1:16:42here's how I like to think of all this generative AI stuff right now whatever you play with right now all
500. 1:16:49this stuff is as bad as it's ever going to be today so if you think about the amount of time it would have taken you
501. 1:16:57if you wanted to go Google a meal plan planner and find a website and sign up for it and most of them charge you money and then
502. 1:17:05go make yourself a list and that whatever however long that took what we just did in three minutes with those plugins this stuff's gonna
503. 1:17:14get way more transparent and way more automated where you'll literally be able to just say to your agent
504. 1:17:20um make me a meal plan that's different than what we had last week it'll remember what you had last week and have
505. 1:17:27the stuff delivered and have it be there within a half hour of when I get home and and it'll know that kind of stuff so
506. 1:17:35I'm here to tell you how stinking adorable you are well thank you very much uh I appreciate that oh it's a little warm and woman is it me
507. 1:17:50glare ah okay first time you save yourself countless hours of work you'll fall in love with AI That's exactly it man it's like
508. 1:18:01the the best way I can describe the impact of Chachi BT and and mid-journey like all of these things is that is that
509. 1:18:10they're just they're Humanity amplifiers they're like Talent amplifiers like whatever you're good at it will make you
510. 1:18:17way better at that and then all the stuff that you are not good at that you always wanted to be good at like you
511. 1:18:24have a desire to be good at something but you're really shitty at it like me withdrawing now you can be good at it or maybe
512. 1:18:31you're not great at writing okay it doesn't matter now you can be um that's going to be remarkable now you want to hear a downside to it
513. 1:18:41you know how when you get one of those scam emails one of the ways you can tell they're a scam email is because they're
514. 1:18:46English is really bad that's going away people with really bad English are going to get really good at those letters that's coming
515. 1:18:58uh can't you use AI to create it for you I don't know what that was in reference to but probably yes
516. 1:19:04yes use Reddit for prompt libraries yeah that's a good idea yeah that's that's another thing thank you for the little hearty head um
517. 1:19:11reddit's a really good source if you want to kind of geek out on stuff or you know dig into open source large language
518. 1:19:16models or prompt libraries or what are the best prompt tools or who are the best prompt Engineers Reddit all the
519. 1:19:23subreddits they're they're geeking out on this [ __ ] so an important thing to know about chat gbt it went from zero to
520. 1:19:32100 million users in six weeks it was the fastest adoption of technology and history so because of that
521. 1:19:39the entire developer Community is just kind of laser focused on all this stuff right now so the the subreddits within
522. 1:19:46within Reddit around all this stuff are on fire so you want to learn about the latest you know features of stable
523. 1:19:55diffusion that the community is putting together and launching it's all in there you want to figure out about the video
524. 1:20:00Generation stuff and who's doing interesting stuff it's all there so that's that's a really good call
525. 1:20:06run through where to find prompt perfect again oh I just did that cool um what app do you use for chat GPT I
526. 1:20:15don't I just use Safari or Chrome on my mobile device and I just point it there and make it a bookmark on my mobile device although
527. 1:20:24I I heard from Rachel Woods uh he was a tick tocker here um today that the the official chat GPT app is now connected to the internet uh
528. 1:20:35but but it's for plus users only so if you pay the 20 bucks the iOS version of the official openai app for chat GPT is now connected to the
529. 1:20:47internet but I just use the website because it's it's really good it's a very simple website what's a prompt okay
530. 1:20:53so the prompt is just basically the query that you type into chat GPT to prompt it to do something so um explain quantum mechanics to me
531. 1:21:08um as if you were Shakespeare that's a prompt right and you can get really clever with them um you can put really basic things in
532. 1:21:17if you put in things that are too basic or too concise it doesn't have enough context to kind of narrow
533. 1:21:24what you really want and it'll tend to give you generic kind of bullshitty kind of answers um so the whole idea of prompt
534. 1:21:32engineering is really understanding you know how do I kind of logically put together sort of language Lego blocks to tell
535. 1:21:42this thing what I actually want and and the trick for me is don't think of it like you've got to write the
536. 1:21:48perfect prompt at once think of it more like a conversation you're going back and forth with this tool saying I want
537. 1:21:55you to act like this kind of expert and then I'm going to ask you questions about this and then I want you to help
538. 1:22:00me with that can you do that yeah I can do that Ah that's awesome you're the best no you're the best oh Bubba I love
539. 1:22:07you Bubba I love you Bubba think of it like your Bubba and it'll get better so don't I think of it more like prompt
540. 1:22:14crafting than prompt engineering but depending on how geeky you are it is definitely thinking like a programmer
541. 1:22:21but using your native language Googling on chat EBT has been the biggest challenge for me having to switch that mindset yeah exactly that
542. 1:22:29that that idea of it it's just not Google it looks like a Google search engine but that's not what it is and and
543. 1:22:37that that has taken me a while to get my head around as well which prompt formatter do you suggest for mid-journey
544. 1:22:44I don't know I there's there's a bunch of them out there those those um those Bots that I used on Poe poe.com
545. 1:22:52mid-journey and po.com slash MJ captain um are both good um the other thing that you could do if you if you want to practice your prompt
546. 1:23:04crafting is go find there's websites out there that will sort of give you the the um the syntax of a mid-journey prompt it'll
547. 1:23:14like here's all the options you can put in right here's the modifiers and the illustration Styles and the the
548. 1:23:21photography whatever bits and you know all sorts of different elements you could take that documentation paste it
549. 1:23:29into chat GPT say you know I want you to read this documentation on mid-journey and then I'm going to put in a prompt
550. 1:23:37and I want you to give me back a well-crafted mid-journey prompt so you can actually create your own bot to do that
551. 1:23:45um which which would be good practice to just practice your prompting thank you I'm Gen X I try give it as much
552. 1:23:52descriptive info for specific requests I'm getting better yeah Michelle Donald that's that's that's kind of the trick is just keep
553. 1:24:01experimenting with it just keep you know when it gives you five shitty answers in a row don't give up on it just keep
554. 1:24:10telling it why it's shitty like that's the thing again like I equate this very much to like working with an
555. 1:24:17intern you know you get an intern and you're all excited and you're like you're like okay I want you to come up
556. 1:24:22with a marketing plan and they go what's marketing and you're like oh God oh God okay okay here's what marketing is so
557. 1:24:31what's the plan huh you know and you spend that first you know four hours or four days or four weeks with the intern
558. 1:24:38trying to just get them to understand the base Concepts so you can then have them go off and do some research
559. 1:24:43um it's very much like that it's much more compressed time-wise but but giving it that context is really the key to to making it better
560. 1:24:52you just gave me my next job good I don't know what I did but oh is that the stand-up comedy yeah start a tick tock
561. 1:24:58Channel Eric I think you'll be good AI writes grants for you yes um that means you could you get to spend the money build your
562. 1:25:06future exactly you can make money with chat GPT just have it write you a grant submit that Grant get paid for that
563. 1:25:13Grant and go spend the money exactly let's hope um a cyber SEC joke was tongue-in-cheek I figured it was but it's still not far
564. 1:25:22off from where I think it's headed I'll get chat GPT to write my jokes one of the in at imprompts.chat one of the
565. 1:25:28early things is act as a stand-up comic and I've had it act as a stand-up comic and it's pretty shitty at writing jokes it understands
566. 1:25:37it understands the mechanics of a joke but it doesn't quite understand jokes so it will occasionally stumble on something that's funny
567. 1:25:47I get Chachi PD to write my jokes yep uh so a question go on um fake news Okay biggest social challenge to me with AI is that people
568. 1:26:00anthropomorphize AI That's gonna get worse Eric so I I think that um so right now Chachi EBT just kind of looks like a Google search form
569. 1:26:11um I think what it's going to start to look like is like your little buddy like an agent like an agent that is either your little
570. 1:26:19friend it's I people joke about you know well maybe they should bring back clippy remember clippy from Microsoft if you're if
571. 1:26:27you're a Gen X or or Millennial you know you know clipping clippy was this failed AI sort of recommendation engine living
572. 1:26:35within Windows it looks like you're trying to write a business letter pring want me to help you with that it was just [ __ ] awful um
573. 1:26:43I think there's going to be effectively a more sophisticated version of that that I'm sure Apple's gonna have an
574. 1:26:50amazing one I'm sure there's going to be something within Windows that's like that I think that where chat GPT goes or
575. 1:26:57we're pogos where some of these things go is that right now like Poe has all these Bots and all the Bots are just
576. 1:27:04kind of these type in the thing I could imagine in the future those Bots become like 3D avatars and they actually
577. 1:27:11talk to you and I can have one that talks like Shakespeare and I can make my Shakespeare bot look like Shakespeare
578. 1:27:18talk like Shakespeare and you can actually interact with it like you're having a conversation so I actually I I think the answer
579. 1:27:25anthropomorphization is gonna get um is gonna continue um and I think virtual companionship is you know they're sort of the
580. 1:27:37you know the sort of uncomfortable you know creepy version of that um but I think there's also a really
581. 1:27:46healthy version of that where you could have virtual physical therapists that are available 24 7 or you know virtual like assistant
582. 1:27:55um uh you know therapist where you have your real therapist but you have kind of a virtual version of them that you can
583. 1:28:01talk through things and if there's something serious that needs to get escalated to the real person the the bot
584. 1:28:08will be good enough to do that um Harvard today announced that they're putting um next fall their their computer science 50
585. 1:28:19they're like seminal computer programming class at Harvard the teaching assistant's gonna be an AI bot so it's going to check your code it's
586. 1:28:29going to help you debug it's going to help you document it's available 24 7. so yeah so it's like like this stuff is coming where
587. 1:28:38people are going to be interacting with these things very much like you know companions and they'll say please and thank you
588. 1:28:46like I say please and thank you when I'm using chat GPT all the time um I need to stop dominating your comments with snarky remarks and pay
589. 1:28:55attention well then stop just then do it then Tick Tock it might have some advice all right watch I'm gonna be at work
590. 1:29:07tomorrow like oh it's like because I go uh does AI have biases based on information like religion or race yeah in fact so the
591. 1:29:21when I first started doing these lives I was explaining a lot about biases and how they work and how biases get in the
592. 1:29:28models and how to get them out of there and I got banned twice in a row when I started talking about bias saw I'm not
593. 1:29:35going to talk about it because they kicked me off for seven days for I like I don't think I said anything but it was
594. 1:29:40just I guess because I was talking about bias it's like oh this is political discourse he's saying bad things about
595. 1:29:47bad people or bad things about good people or whatever or someone reported I don't know what it was but it doesn't matter but but um
596. 1:29:56yeah I mean it's so the the way open AI trains fine-tunes their models so they train it on kind of all of the internet
597. 1:30:05so the internet's going to have whatever biases it has so it's been trained on all that [ __ ] and then the way they train it is with
598. 1:30:12human reinforcement learning so it's basically people going yeah that's a good answer that's a bad answer that's a
599. 1:30:16good answer that's that answer well people are biased right so that introduces bias so um there's there's a company called
600. 1:30:22anthropic so if you go to poe.com three of the models there are called Claude different variations of Claude that's
601. 1:30:29for a company called anthropic and they have this thing called the Constitutional learning model so instead of humans
602. 1:30:35going thumbs up thumbs down they've got a constitution like you know um you know Do no harm increased Prosperity
603. 1:30:44um whatever you know don't cause pain whatever whatever the Constitution might be and then the the way the system works in
604. 1:30:51a dramatically simplified version is when it generates an answer it Compares that answer against the Constitution and
605. 1:30:57it will iterate on the answer until it can pass the Constitution right so so systems like that are going to start to
606. 1:31:04even out the bias and I see here's the thing about bias is that everyone's got biases so the the model
607. 1:31:12that doesn't look biased to me is going to look completely biased to you so I think what we'll see in the future is is
608. 1:31:18you'll have multiple models that you can choose from and you can kind of opt into the value set of your choice that's
609. 1:31:25that's my guess I don't know that it's going there but there's not a single perfect model because human beings are
610. 1:31:31[ __ ] up and biased and that's just the way it's just the way it is um to access prompt perfect do I simply put
611. 1:31:39perfect before my prompt yeah you go to you go to gpt4 with plugins you enable prompt perfect plug-in and then just type the word
612. 1:31:50perfect in front of your prompt and that's all that's all all she takes all right most of the data is probably okay yeah
613. 1:32:07well I don't think that most of the data is probably one race or another like like if if you're if you're talking
614. 1:32:13about the training the people that do the training for chat GPT because they were in San Francisco the the that's
615. 1:32:21probably true of the people doing the human reinforcement learning for open AI but the um openai was trained on like all of the
616. 1:32:32public internet so not just America like all of the public internet so to the extent that all of the data on the
617. 1:32:38public internet is biased or skewed in you know toward One race or another that's going to be in there but but I
618. 1:32:46wouldn't it I wouldn't characterize it quite so broadly as is what you put there and I don't want to say it out
619. 1:32:52loud because I don't want to get the channel shut down again because I'm not saying anything I'm just
620. 1:32:57trying to explain what the [ __ ] is going on I've seen a lot of mid-journey images featuring people of color yet the it's
621. 1:33:05they're getting better and you but but you have to really prompt for it um the the people of color thing in the
622. 1:33:12in the stable diffusion and in mid journey and probably to ex to an extent with Adobe but I bet Adobe
623. 1:33:19has been adobe's been way more conscious of what's in their training data so I would imagine that adobe's worked really
624. 1:33:27hard to balance that so that would that would be an interesting experiment is compare mid-journey stable diffusion and Adobe Firefly
625. 1:33:35for diversity um I have plus and it's on I had forgotten where prompt perfect was so thank you for answering of course Silver
626. 1:33:43Fox that's why we're here um have fun with that also go to uh poe.com and go to the uh the mid Journey prompt writers actually just go go
627. 1:33:53explore all the prompt all the the PO Bots the the uh the chat bots in po there's there's some really good ones in there
628. 1:34:04great job you've been live for 90 minutes that's why you're shiny and your voice doesn't work anymore I can't form
629. 1:34:10words thank you for the roses that is a Rose Garden you've given me there I I appreciate the hell out of that wood
630. 1:34:20what is what is tricking me out with roses I they're everywhere thank you very much I appreciate that
631. 1:34:27sorry old question oh I got lost uh Tick Tock just jumped me to the bottom so I'm going to scroll up a little bit and grab
632. 1:34:34some questions here did you see the documentary on him I don't know what that was oh my God it's great worth a watch I love that you
633. 1:34:45guys have conversations with yourselves in the comments that's actually super cool wow some more roses thank you so
634. 1:34:50much I really appreciate it um it's a human amplifier well said yeah yeah it's yeah it's yeah I feel like for the last two days my um
635. 1:35:06My my what do you call it uh my my uh uh grasp of the English language has been suffering like I I just I words not
636. 1:35:18so good what's that Steve Martin joke some people have a way with words other people oh not halfway I guess you'd say
637. 1:35:27that's what I felt like for the past two days not halfway hey Kyle do you work for open Ai No if not what's the reason
638. 1:35:35that you talk about chat GPT every day because okay so a couple of things I'm not a machine learning guy I'm an entrepreneur I was there
639. 1:35:49I was I was working in New York City when the world wide web web came onto the scene and I started one of the first
640. 1:35:55digital agencies I had a big Epiphany about the internet and the World Wide Web once it came along and what Chachi PT represents is is
641. 1:36:06essentially The Identical um inflection point as as what the World Wide Web did for the internet so let me explain a little bit
642. 1:36:17the internet had been around for decades right it was you know invented whenever it was invented 50s 60s and you know uh
643. 1:36:25and then and then in the 70s they started connecting all this [ __ ] and um and then it was largely you know researchers and um
644. 1:36:37scientists and universities and research paper you know repositories for decades so if you weren't really a scientist or
645. 1:36:48a com you know an engineer of some sort you really didn't have much use for the internet you could use it but you had it
646. 1:36:53was all command line and then Tim berners-lee invents the world wide web and it allows us to create these web
647. 1:37:00pages with these blue underlined words on them you can click on that and that will jump you over to another page and
648. 1:37:07that changed everything right so all of a sudden the internet went from being for scientists and researchers to anyone
649. 1:37:13that could learn a couple of HTML tags could put up a page and Link that to some other page and that changed everything it transformed
650. 1:37:20you know it collapsed geography it it you know collapsed time you could just instantly go from one piece of information to the other
651. 1:37:30um you know instant Communications all that sort of stuff radically transformed how we do business how we communicate how we connect how we
652. 1:37:39date how we everything shop so I talk about chat GPT all the time because machine learning and AI has been around
653. 1:37:50for decades and and you know certainly in the past 10 to 20 years it's been increasingly powerful and we hear it
654. 1:37:56about a lot and it's it's AI is is in almost all the software we use like all of the flip and Tick Tock filters that
655. 1:38:04make you pretty and give you a head of hair and all that sort of [ __ ] that's all AI That's AI how that's positioned and tracked is all AI
656. 1:38:12chat GPT represents putting that magic in the hands of real people regular people I have a degree in acting I've started a
657. 1:38:24bunch of companies I love technology but I'm not a technologist I like understanding how technology works and then figuring out how that's
658. 1:38:33going to change the world that's what I'm passionate about so for me chat gbt is the seminal piece of technology that
659. 1:38:41broke the door open and and here let me put another thing in context for you the World Wide Web to get to 100 million
660. 1:38:49users took six years it was six years from the Tim bernersley inventing until we got to about 100 million users chat GPT six weeks
661. 1:39:00fastest adoption of technology in history the other thing about chat GPT is this and why I talk about it all the time
662. 1:39:08if they didn't improve it if it stopped today like you know Elon Musk signed that letter you know they should stop
663. 1:39:14innovating in AI for six months and let me catch up I think that's what that was about but whatever if they stop development of AI today I
664. 1:39:25feel like I could I could personally spend the next decade figuring out how to use chat GPT it is it is so deep it is so profoundly powerful
665. 1:39:37like I haven't scratched the surface of it so I just think it's that important and and it's also you don't need to have any computer
666. 1:39:48skills to use it you go to that web address and there's like a Google box there and you just type [ __ ] into it
667. 1:39:57tell me which plants like it's you know whatever it is it's the end of June in Denver Colorado what plants should I be
668. 1:40:06planting in my vegetable garden boom out comes your answer would I've taken the time to go Google that no because
669. 1:40:16it would have taken me too long and there'd have been too many ads and I don't really give all that much of a
670. 1:40:20[ __ ] about it oh but now I got this thing they'll just give me the answer boom there it is it's I I don't know because it's profound
671. 1:40:29no hate just curious I love your lives thank you yeah no no it's like it's a good question and it's like I've been
672. 1:40:35thinking about that a lot actually like at what when at what point am I going to change these URLs like like there's probably a better
673. 1:40:45prompting library to put up here than this but this is actually just a pretty good primer it's got a decent primer in it
674. 1:40:52and and the Persona thing where it says act like a screenwriter act like a stand-up comic that's pretty good
675. 1:40:59this is just a list of list of tools and I don't really follow all the tools anymore so there's probably five or six
676. 1:41:05of those at some point I will probably swap these out for something else I don't know when like it still feels
677. 1:41:12relevant to me and I've got ADHD so I'll get bored with this [ __ ] I'm an artist and I'm disappointed how
678. 1:41:19cartoony most images I've tried I want something painterly okay so RM Walter I'm gonna go show you something
679. 1:41:29because you can absolutely get there you can absolutely get to painterly um [Laughter] [Music] go to
680. 1:41:45open art Dot is it AI yes go to openart.ai promptbook
681. 1:42:00openart dot a I slash promptbook and it's going to take you to this document that's a stable diffusion prompt book
682. 1:42:11but I think this will help a lot because these concepts are transfer transferable to to other things like how you actually
683. 1:42:20you know put the the codes into the tools might be slightly different but the basic concepts are the same so so there's a bunch of different
684. 1:42:27sections in it it's a 96 page document so they talk in general about prompt engineering they talk about your prompt
685. 1:42:34format like do you want a photo or a painting what's the subject is it a person an animal perhaps a landscape
686. 1:42:41what details do you want to add special lighting environment color scheme point of view is there a specific art style so
687. 1:42:48to your point if you want it to be um you know claymation sculpture you can put that if you want it to be you know highly textured oil
688. 1:42:59paints you can put that is there a specific photo type macro telephoto so so it's starting to explain
689. 1:43:06to you that there are different elements different structural elements you can put into a prompt and then and then this
690. 1:43:12document just basically continues to unpack all of that stuff so this section on modifiers like here's photography modifiers
691. 1:43:21shot type style subject lighting context indoor outdoor at night in the park studio right um it shows you little examples of if you
692. 1:43:34use the word Polaroid it understands what that is if you put tilt shift or product shot like these these systems have an incredible depth of
693. 1:43:44understanding of art Concepts like the the this document is like a good art history um it's like it's like it's like art art school in a box
694. 1:43:55um lighting nostalgic Sunrise art medium so this gets to what what we're talking about with you right chalk graffiti watercolor oil painting
695. 1:44:05um artists so talking about specific artists so if you've got specific artists that have got a style that you
696. 1:44:12like you know you can invoke their names if they're modern artists I would encourage you to avoid using modern
697. 1:44:20artist names but certainly for the classics um you know certain Styles you can invoke them and it it is remarkably good at at
698. 1:44:31pulling that sort of stuff up so again this is at openart.ai promptbook and that should help I hope that helps
699. 1:44:47all right I'm gonna answer a few more questions I'm getting a little crispy I needs my money someone set me up
700. 1:44:56you can make money with Chachi day we all need money I need money give me the money money
701. 1:45:11I never like clippy clippy 2.0 clippy was annoying [Laughter] I remember clippy so annoying I know I read today clippy was back I think
702. 1:45:22someone I think someone created a clippy chat bot um I don't think it's back within Microsoft it would be a brilliant move I
703. 1:45:30mean quite honestly if Microsoft relaunched clippy as not sucking and quite honestly I think Apple's got a relaunch Siri as not sucking um
704. 1:45:42I think I like the Nissan The Nostalgia thing and and the sort of tongue-in-cheek marketing Buzz that
705. 1:45:50Microsoft would get by like leaning into clippy and saying clippy was you know was before it's time like that would be
706. 1:45:58pretty [ __ ] funny and then if clippy were gpt4 you know super master clippy that actually didn't suck cool uh yes yes
707. 1:46:12does AI have a code tester built in yes give it some code tell it to debug it for you it will tell it to document it
708. 1:46:20for you it will tell it to write you code it will tell it to take code from this language and refactor it into that language it
709. 1:46:28will now is it gonna be perfect every time nope because here's the thing about the way the large language models work
710. 1:46:34it doesn't actually understand what it's doing what it's doing is a probability exercise you write a prompt it comes up
711. 1:46:43with a mathematical weighting that then then says what's the next most probable word after this long prompt and then
712. 1:46:50what's the next most probable word and and so on and so when it writes all that stuff it's doing it word by word as a
713. 1:46:56probability engine it's got no concept of what the hell it just wrote so it may give you code that looks right but is
714. 1:47:02actually wrong so it's good if you know code but if you know code you can dramatically increase your productivity
715. 1:47:10with it it's it's remarkable I hope it doesn't get so good that it replaces my dog possible garbage in garbage out correct
716. 1:47:19um well AI should take that Harvard bias test it should sorry old questions and it looks like all right got a few more questions at
717. 1:47:28the bottom here and then I'm gonna take off thank you again for all those Roses woodsy appreciate that is chat GPT constantly upgrading
718. 1:47:41changing yes on the Fly no so when you use chatgpt if you just go to chat GPT 3.5 or gpt4 not connected to the internet
719. 1:47:53and you ask it for something current it'll basically say I'm sorry I only know things up to September 2021 and
720. 1:48:01then you're like well that's dumb if you use Bing uh which is connected to the internet or if you use gpt4
721. 1:48:11connected to the internet if you ask for something recent it will actually fire off a web search go search the web get
722. 1:48:17the answer and then incorporate that into your answer so that's how it gets around that limitation the limitation is
723. 1:48:23there because when they build the the core large language models it takes a tremendous amount of time to identify and tag the the core data
724. 1:48:32sources that get what's called embedded ingested into that base large language model so that takes months if not a year
725. 1:48:41or more and millions of dollars in many many gpus running 24 7 to do all that so why this thing works so good is because
726. 1:48:51they did all that work for us and then the fine-tuning the sort of more recent stuff is sort of sitting on
727. 1:48:57top of that and then like the internet searches and things like that so but yes like they're they're training gpt5 right now
728. 1:49:06um while we're using gpt4 but gpt4 hasn't even been fully rolled out like I you may not know this gpt4 is multimodal meaning
729. 1:49:16it can understand images so you could upload a picture of a dog and a fire hydrant and it would you'd be
730. 1:49:23able to ask chat GPT questions about that photograph describe this um when what year was that fire hydrant
731. 1:49:32made what if whatever it might be it it's it it's going to be able to understand image inputs that's not even
732. 1:49:39rolled out yet like that's not available to us there's a new thing coming called code interpreter where you can upload 100 megabyte
733. 1:49:48um spreadsheet of data it will read all the data and then act as a data analyst for you like that's coming so yes so
734. 1:49:56they're constantly upgrading it but it's not they're not dynamically updating that base model just because of the
735. 1:50:02nature of how they're built is it's a big expensive manual process um how long how much storage does GPT save I don't
736. 1:50:16oh saving history how much storage oh okay okay so there's there's a thing called a context window so any given session that you're doing
737. 1:50:27with chat GPT there's a essentially a 3000 word think of it like a sliding 3000 word window so if you start and say
738. 1:50:35write me a poem and then fix the poem and then do this and then do that and now turn it into a short story
739. 1:50:42at some point it will kind of forget about the poem you told it to write because it'll slip out of that context
740. 1:50:48window so that sliding window is about three thousand words um there are tools out there like there's a thing called Lang chain which is a
741. 1:50:57series of tools that are kind of glued together that give it longer term memory um and then there's the if you go to
742. 1:51:04poe.com one of the models there has a hundred thousand uh like a 75 000 word um context window so the the 75 000 word
743. 1:51:13context window at poe.com um has a much longer memory so so that's why sometimes it forgets what you're
744. 1:51:22talking about you're like well wait a minute you just I told you all that stuff and now you don't seem to
745. 1:51:26understand it that's why because it slipped out of that context window so um yeah that'll that'll help
746. 1:51:33hopefully explain what the [ __ ] going on there um again a year from now I I think these things will this effectively have infinite memory
747. 1:51:45um chat GPT has already announced that um you'll be able to create a profile within chat GPT where it will understand
748. 1:51:55who you are what you like what your background is like what whatever stuff is in there that you want to be kind of
749. 1:52:01your permanent memory you'll be able to put in there it's not there right now um so so just understand that we're
750. 1:52:09early and this stuff sucks for a while um I made an Elvis bot cool um in open chat to act as customers
751. 1:52:18customer support it responds with well thank you very much it's funny the last play I did I played Elvis in Picasso at the lapana's yield
752. 1:52:27in Ridgefield Connecticut that was the last play I did that was in 2004 it's a while ago how do you get money
753. 1:52:39you do stuff for people that are like I want to give you money I don't know what the question is how do
754. 1:52:47I get money I'm the CEO of a of a video technology company and we do a lot of work in the healthcare and Pharma business and
755. 1:52:56that's how I make money right now um if I keep talking about chat GPT this much open AI should start paying me a
756. 1:53:04sponsorship [ __ ] fee I mean this is like a nightly commercial for them um you need to create a commercial where
757. 1:53:12you can say you can make money I should do that that would be funny except I don't know how you make that without like I feel like all those
758. 1:53:21people are already Lampoons of themselves like they're just [ __ ] ridiculous is there a knowledge update coming in
759. 1:53:33gpt5 that's probably not gonna be for at least I'm guessing a year like like I I don't think open AI is interested in um
760. 1:53:45speeding gpt5 to the market I think they're probably Slow Rolling it because they're still trying to they haven't fully rolled out gpt4 yet
761. 1:53:54there's a lot there I mean the other thing is if you think about it when they created chat gbt um the team put it together in like two
762. 1:54:05weeks it was like a very short project like like the piece of software that they launched on November 30th was
763. 1:54:12really just a proof of concept that that kind of coincided with Microsoft saying hey we're investing investing a billion dollars in open AI
764. 1:54:20um and then in five days they had a million users and in six weeks they had a hundred million users so I think that
765. 1:54:27open AI right now is is just dealing with scaling you know getting the safety guard rails right like locking this thing down
766. 1:54:36because so many people are using it and it like this generative AI stuff went from like people talking about it a little bit to
767. 1:54:45all of a sudden it's like the most overhyped thing in the world and that happened overnight like with the World Wide Web
768. 1:54:51I started agency.com in 1994. it wasn't until like mid 1997 until people really started frothing at the mouth about the
769. 1:55:01internet and then it wasn't until 2000 that the the.com bubble burst so there was a good six seven year period where
770. 1:55:10the World Wide Web awareness of it was ramping up this thing was like overnight like it was literally overnight that that thing
771. 1:55:18just jumped onto the scene so crazy crazy crazy awesome show sir thank you very much I appreciate that
772. 1:55:30all right a few more questions it's 9 56 if you make the commercial you just demoed it'll be the next where's the beach
773. 1:55:45oh my God by the way if if if that if one of your questions is how do you make money with that gbt here's
774. 1:55:55the here's the the non-smart ass answer to that is chechi PT is just a tool so it's Photoshop so is Microsoft Word so is a shovel
775. 1:56:06so is a lawnmower so the answer to how do you make money with chat EBT is actually identical to how do you make money with any other
776. 1:56:13tool how do you make money with Photoshop well you get good enough at Photoshop that you can solve someone's
777. 1:56:19problem well enough that they're willing to pay you for it I can take your photographs and make them suck less will you pay me for that sure
778. 1:56:29here's 50 bucks great here's your picture if that's a good experience you got a business if you suck at Photoshop you
779. 1:56:36got 50 bucks time to move on same thing with this now here's the difference with chat gbt with Photoshop
780. 1:56:46you actually have to have some talents and skills with mid-journey and with chat GPT the depth and power of what it can create
781. 1:56:59is significantly beyond what any of us can do individually right so I could write in the next hour I could write a
782. 1:57:09plausible research paper on a subject I know nothing about on mathematics or on physics or on I don't know
783. 1:57:20um organic gardening techniques or on [ __ ] that I know marketing plans um um you know storytelling Theory like and
784. 1:57:30then I can go to Mid journey and I can or or stable diffusion and I can create sophisticated images even though I don't
785. 1:57:37have the talent to do that so so that's the difference with it is that again it's a Humanity amplifier so it can take
786. 1:57:43whatever you desire to do and it can do it really really quick um so that's that's how you make money with chat EBT is you figure out a
787. 1:57:52problem to solve you take something that you've got knowledge about or something that you've got passion about and then you figure
788. 1:57:59out which of these tools can help you do that faster better or cheaper than the next person down the line and then you
789. 1:58:06tell the world about it and if you do it right they're like here's 50 bucks and if you do it right they'll give you 50 more
790. 1:58:13and if you get 10 of those then you can charge 500 bucks and if you get 10 of those maybe it's 5 000 and then you have
791. 1:58:19a career or you have a job so I know nobody asked me how do you make money with chat GPT but if we did start with the commercial s [Music]
792. 1:58:31[Laughter] Amelia's wife hears it in her sleep that's funny um all right a couple more a couple more and then I'll go to go go go go go
793. 1:58:47Limitless ways to make money with chat GPT have you tried AI powered Google workspace I haven't gotten access to it
794. 1:58:55yet I'm a little cranky about it like I think Bard sucks Bard compared to gpt4 is just it's like not anywhere in the same neighborhood
795. 1:59:04um but even the suckiest Bard is better than what we had before in Google Docs right so if I've got a tool that can can
796. 1:59:16basically be my writing assistant as I'm writing things in Google Docs and when it comes out as co-pilot and in Office
797. 1:59:24365 that's going to be remarkable I mean I but no I haven't tried it when it when it gets truly integrated into the tools
798. 1:59:31we use like like photoshop's generative fill is really just a natural extension of if you've ever used Photoshop and you know
799. 1:59:38how to make a selection and invert it and you paint [ __ ] outside of the thing that you were doing before like anyone
800. 1:59:46who's used Photoshop has done that it's like that except there's [ __ ] magic pixie dust in it that just fills in your
801. 1:59:53photo for you and it looks world class you don't have to photograph out the you know the power line that went across the
802. 2:00:02tree and you had to like rubber stamp all the leaves and no matter how hard you tried it looked like [ __ ] and you'd
803. 2:00:07spend three hours trying to photoshop out a [ __ ] wire well now you just like Circle the wire and go remove wire give me more background
804. 2:00:20Everything Changes Everything Changes Everything another great AI date night in the bag thanks Kyle thank you Silver Fox AI date
805. 2:00:32night that's what we're gonna call these hi I'm Kyle Shannon this is AI date night [Laughter] all right everybody I am gonna go I I
806. 2:00:41hope you've had a good time tonight I appreciate all the hearts and all the roses and all the sweet sweet uh words
807. 2:00:48and didn't have any Power Trolls tonight we did have Eric working on a stand-up routine so Eric thank you for your comedy
808. 2:00:56thank you frumple for the rose and uh I'm gonna go take a nap or as as we like to call it in the in uh in Denver sleep
809. 2:01:07all right peace out everybody I'll see you tomorrow