aiLL 8/3/23 - The Democratization of AI is Here

Transcript

1. 0:00foreign
2. 0:21welcome to the AI learning lab my name is Kyle Shannon we'll be getting started here briefly it's all brand new
3. 0:34right now tonight what's tonight Thursday night I got a workshop I'm teaching at 8am tomorrow so probably can't stay on too late but
4. 0:46we'll get something going here we'll let some people join and jump in and say hello hello hello hello hello
5. 0:55we got Champa here Champ's ready to start the live you excited boy you excited buddy yeah come here yeah yeah oh that's a good doggy
6. 1:10good doggy doggy doggy welcome welcome welcome this is the AI learning lab if you're new here new to generative AI I ask no I
7. 1:21answer I ask a lot of questions too I answer as many questions as I can about AI if you pop your questions about AI in
8. 1:27the comments below I'd do my best to answer them if you're new to generative AI uh and you haven't played with chat GPT yet let
9. 1:36me let me flip this around this this big long list used to have only three things on it now it's got seven
10. 1:44things on it because that's how fast shit's moving um let me make it a bit smaller so it fits in a little space let me get my big fat
11. 1:54head out of the way there we go so if you're new to to the AI world and generative Ai and chat GPT and all that
12. 2:04happy horse [ __ ] take a screenshot of this and I'll quickly walk through what these things are welcome everybody welcome
13. 2:12welcome welcome welcome if you have questions about AI pop them in the comments below um my name is Kyle Shannon I'm an
14. 2:19entrepreneur I'm the founder and CEO of a company called storybine which is a an automated video platform that actually
15. 2:26uh we've just recently added AI it's it's based more on Aristotle than AI but but uh it is automated video editing
16. 2:35and uh for the past year and a half or so I've been deep deep deep diving into this generative AI stuff and when chat GPT came out
17. 2:43it hit me that the world has changed just not everyone knows it yet um so all of these things here what are they
18. 2:51the top one is the official chat GPT website if you've never been to it or done that go there right now well I ramble on about
19. 2:59how crazy all this [ __ ] is uh go there and start playing ask it to write you a song or a poem or explain quantum
20. 3:06physics to you it doesn't really matter just go there until you until you have what I call your your Kevin McAllister moment you're oh
21. 3:15home alone uh then you got Claude which is a major competitor to gpt4 that's the official one up here then you got Bing
22. 3:23which is Microsoft's version of gpt4 uh that's connected to the internet that's worth at least exploring because it's
23. 3:30connected to the internet and it can generate images and it can read images then you have Bard which is Google's
24. 3:37entry which until about two weeks ago was just utter steaming trash and now it's actually worth exploring it can
25. 3:45read uh images and it's connected to the internet so it's interesting in that way this one the labs.perplexity.ai that's metas large
26. 3:55language models that you can go play with for free and then you got pi and you got po all worth a uh a check uh
27. 4:02down in the mud says yay he's back I am back I was in uh Tick Tock purgatory
28. 4:16Tick Tock said I was a very bad boy and it kicked me off the platform they said we're gonna de-platform you Mr Shannon
29. 4:24we're gonna de-platform you you can't say your words anymore for exactly seven days to the minute and I appealed and they said no bad boy
30. 4:42welcome welcome everybody this is the AI learning lab we learn about Labs here and the AI stuff I got champ by my feet hey buddy
31. 4:52and uh I just got back at a lovely evening a little CEO networking event here in Denver called the Thunder view CEO something or
32. 5:04other we hang out hang out a bunch of Founders hang out and just [ __ ] about how awful it is starting and running
33. 5:11companies it's it's basically an AAA meeting for entrepreneurs that just can't stop um but I'm back from that I got a couple
34. 5:21of whiskeys in me so I'm loose I'm loose look my hair's looking good the glue held today I think we're gonna do okay tonight I
35. 5:30think we're gonna do all right there's the bear how are you this evening welcome welcome welcome welcome
36. 5:36all right let's let's get some questions uh hey glad you're out of tick-tock detention I know Lauren me too I'm back
37. 5:43I'm back I came back last night it was my first night back it was a little awkward it was like a you know like a
38. 5:49Junior High dance a little awkward but it was good we're okay we got there we got there I got a little feisty with
39. 5:55someone a couple of people um what's it telling me to do tick tock's telling me to do something swipe all right I'm swiping nothing's
40. 6:03happening ah good Lord whatever all right hey Kyle how do you generate images with AI oh there's good lots of ways here let
41. 6:17me I will give you some cool stuff all right so if you're new to AI the the list that I had before all of those
42. 6:29things were all chat gbt like things these things prompts.chat will teach you about how to talk to those large
43. 6:35language models how to talk to chat GPT and all the other ones and it basically just explains how you talk to them how
44. 6:41you provide them with context and then it also gives you a long long list of like personas where you can tell Chachi
45. 6:48PD to act like an accountant or a screenwriter or a therapist or a whatever a dog whisperer doesn't matter
46. 6:55but it'll give you lots and lots of ideas the next one down openart.ai promptbook is um explains to you how you talk to these
47. 7:05image generation tools that are on this page it's specifically designed for stable diffusion but it's good for all of them because it it
48. 7:13it's basically a 96 page PowerPoint presentation that explains all of the [ __ ] that these image generation tools know it's it's like a
49. 7:22little it's like a little art history lesson in a box and and it teaches you how to talk to these things and get really interesting results
50. 7:29um and then futurepedia.io that's just a big long list of AI tools okay so for images The Big Three are mid-journey Adobe Firefly and stable
51. 7:39diffusion I have leonardo.ai down there but that's built on top of stable diffusion it's just that's the friendly web front end
52. 7:48um that this company created and it's really really good right now mid-journey and Leonardo or mid-journey and stable diffusion are
53. 7:58close to neck and neck stable diffusion just launched their new SD XL model which is really good but in my opinion mid-journey still
54. 8:08the best and it's crazy good it's crazy crazy good it's crazy good I tell you I like it the mid journey and that's an
55. 8:14image I made on the mid journey I last a night I did it on the live look at that look at that stunning piece of piece of Madness
56. 8:23all right so how do you do it the first thing that you need to know about mid Journey is that not only do you have to learn
57. 8:31mid Journey you have to learn Discord so if you don't learn if you don't know Discord well you got to learn that so
58. 8:37when you go to midjourney.com it'll say join the beta and you click on that and it'll take you to this thing and you're
59. 8:43like wait where am I what am I what am I supposed to do okay so in Discord you have these things down the the side those are called
60. 8:50servers those are basically your different communities you'll go into one for mid-journey I'm in one uh called the
61. 8:56AI Salon which is a community that I started and then but we've got a mid-journey channel in here and then all of these things here are the
62. 9:04different channels so the different little conversations going on and then within mid-journey or within the salon
63. 9:10in this case in the mid Journey Channel this is where you make images so you you basically are giving a a command to a
64. 9:18bot and what you do is you do slash imagine and that'll pop open this thank you for the Roses bear it'll pop open this thing called prompt
65. 9:29appreciate that and you type in whatever you wanted to make so you could say um a bear in a Chicago restaurant making
66. 9:46uh beef sandwiches anyone else watch the bear God damn is that a good show all right so now off it goes there's all
67. 9:57sorts of fancier ways that you can talk to Mid Journey but the basic idea is you type slash imagine and then you type
68. 10:04in a prompt and then it's going to go off and do its Magic Madness and you'll notice the first thing that it generates
69. 10:12is like this blob and like and it's really blurry and you can't see any details and then you see a little more
70. 10:18detail and then you see a little more detail that's the way these these diffusion models work is they start from
71. 10:25fully noise images where it kind of amalgamates you know thousands of images into kind of a random pixel field and
72. 10:34then denoises them over 50 steps or so and they just get more and more um in Focus basically and every image that's generated is
73. 10:46actually a unique original image even though they the large language models and the image models have been trained on other
74. 10:55people's work the the piece that's created is actually not copy and pasted it's actually just original work
75. 11:04um based on a an insane mathematical formula so we ask for a bear in a Chicago restaurant making beef sandwiches the uh
76. 11:15I think the the two on the left here are really good so so one of the things you can do is you can just say okay the one
77. 11:21on the bottom I like that so that's one two three that's three so so you see there's u1234 and V one two three four u means upscale
78. 11:31and V means variation so I think three is good enough to just upscale but one I like but I want to see variations of it so I click V
79. 11:40and so there's my my large version of that image that's pretty good nice piece of beef there the Bear's cooking it up
80. 11:49wearing some lovely kitchen garb want to make sure that we're food safe there's a bear in the kitchen [Laughter]
81. 11:59uh and then the other thing you can do now is um mid-journey add these things called zoom out um where you can sort of zoom out and
82. 12:10it'll add to the image or there's these pan things so I can actually pan I can say let me let's say I want this to be a
83. 12:17horizontal image right now it's Square um I can hit the zoom button to the right or the Pan button to the right and
84. 12:25it's going to add to the right of this image and so it's going to create an entirely New Image there's the variations for number one
85. 12:35so let's see those we got um I think I think the second one there is is the best personality so we'll do upscale too
86. 12:50and that's going to give us see there there it's adding to the side of that image so that's going to do its thing
87. 12:56and then there's the upscale number two here so this one will do zoom out you can see what zoom out looks like
88. 13:04and and this is basically it this is you know you can get way more sophisticated with options and using different models
89. 13:11and doing image to image and doing all just there's all sorts of crap you can do um is this done yeah this is done so pendrite
90. 13:23so now we have a landscape image of these bears with different options in the backgrounds I think I like
91. 13:35this one's kind of got a nice row of jars here this one's got cool stuff on the table let's do the cool stuff on the table one
92. 13:46so we'll make that one which one is that number two we're gonna go upscale two so there's our upscale too all right there it is
93. 14:00go to that we open it in our browser so we can zoom in on that all nice and a little detail flipping amazing
94. 14:18and that was a shitty prompt like I didn't put it like I just you know I just put like bear in a Chicago kitchen
95. 14:23making a meat sandwich beef sandwich and we got that what did that take us two minutes two minutes two minutes get your [ __ ] together
96. 14:33it's a George Carlin joke how many times do I have to tell you six all right and then now we've zoomed out on this
97. 14:44image so we'll see which one of those looks good they're all pretty cool I like the one with the light here
98. 14:50these lights look fancy let's let's do that one number one we'll upscale number one and so this is all zoomed out version of
99. 14:59number one so there you go Bang so one we sort of pan to the right this one we pulled back and it added these
100. 15:07lights it added all this the the little bar stool here all this crap to the side here all this stuff is all added
101. 15:15all new without having to think about it just zoom out zoom in pan left pin right make a bear in the kitchen here's a bear in the kitchen
102. 15:23[Laughter] can I use this crap it makes me [ __ ] giddy because it's like you know like a year ago none of this was possible
103. 15:35nine months ago Chachi PT didn't exist you could use you could play with gpt3 3.5 or gbt3 whatever it was at the time
104. 15:44but it wasn't it wasn't sort of put in a little package that made it just like this magical little tool stick in some
105. 15:51words out comes a novel not quite that good but pretty [ __ ] close stick on some words out comes an illustration of a bear or a photograph
106. 16:03I mean you know what would it what if if you've ever tried to do anything in the creative realm like to get something
107. 16:11even remotely close to this concept you know it would be a week or two of Googling illustrators or photographers
108. 16:21two or three weeks of interviewing them because they always flake out because they're creatives I used I used to be
109. 16:27the chief creative officer of an agency so I know of what I speak um and then it would have been you know two or three weeks worth of
110. 16:38pre-production and a shoot especially for something this complicated or an illustration and you'd have spent a [ __ ] ton of money
111. 16:47and now it's like Bang now the downside of these things is to get things really precise and predictable to do things like text it's
112. 16:57really really hard um but the state of the art right now is pretty remarkable knowing that these tools are going to get much much better
113. 17:07um so that's that one the other things I play with if you work in a corporate environment you should use Adobe Firefly
114. 17:12or at least explore it because all of their image Generation stuff was was trained on basically non-copy written images
115. 17:21um leonardo.ai is just a really really good web interface if you don't want to deal with Discord which is a pain in the
116. 17:28balls go to leonardo.ai I think their free version gives you like 150 credits a day which sounds like a lot but it's
117. 17:37like I don't know 15 credits per image or something like that or eight credits an image but you can do a decent amount
118. 17:43of images per day for free here but it's really good you can train the model in your own images there's all sorts of
119. 17:51it's just it's just good it's just good so anyway that's that uh if any of you are kind of trolley like oh he's just making
120. 18:00money off this no I'm not I don't make money off any of these guys now should they reach out to me and go hey
121. 18:07can we give you money yeah sure I'll talk about your [ __ ] for money uh but they're only up here because I actually think they're worth you
122. 18:16spending your time on all right so dang that was a good first one we dove in day what's your what's your name
123. 18:23Jelly Bean jelly bean pup Mom I'm gonna go nine out of ten on the name solid name it's a little bit of a mouthful but
124. 18:32all wholesome awesome stuffs Jelly Bean pup mom there's a dog in there Mom jelly beans can't go wrong with any of that all right why why not
125. 18:45how to how to make money with AI well Gene as we often say on this channel you can make money watch out every day
126. 18:56here's my thoughts on how you make money with chat gbt it doesn't matter that it's chat GPT it doesn't matter that it's ai ai is just a tool
127. 19:06um the first step in making money with chat is start to get AI literate start to play with these tools a lot don't don't
128. 19:15start with I need to make money as the goal just start understanding what they are because in learning what they are
129. 19:22what you'll quickly realize is you might be good enough at something to make money that you never thought you were
130. 19:29good enough to make money at before like making images or writing words a lot or coming up with marketing plans I don't
131. 19:36know a lot about marketing well this [ __ ] does so Step One is like get curious get EX just start exploring start working on
132. 19:45personal projects start working on projects for other people learn these tools then you don't need to learn all
133. 19:50of them like pick one pick the top one that's that's the original that's why it's at the top it's why it's bold
134. 19:57chat openingi.com you can just play with the free um 3.5 version so if you don't pay anything there it's GPT 3.5
135. 20:08um and start learning about prompting and then here's my philosophy on how you make money with any tool so this is if
136. 20:16you want to make money with Photoshop if you want to make money with Microsoft Word like like the the thing about you giving money
137. 20:24with chat GPT they could be making those videos you can make money with Microsoft Word how do you do that well you write good
138. 20:31and you use Microsoft Word how do you make money with AI you learn how to solve a problem better faster or cheaper
139. 20:38then you could solve it manually right so the prerequisite here is just start learning what these things make possible
140. 20:46and then do things like go play with mid-journey or leonardo.ai or adobe Firefly start to understand the image
141. 20:54generation thing start to get your head around these tools are actually generating original content and if you can get
142. 21:02astute at writing prompts to get these things to give you good results you'll be able to make money you'll be
143. 21:10able to teach courses you'll be able to do Services you'll be able to build software or projects that you never thought were
144. 21:19possible you'll be able to support local businesses you'll be able to do [ __ ] for your family all of that stuff's possible um
145. 21:30yeah lots and lots of it we had a guy at the the salon the AI salon so the other thing that I started is a group called
146. 21:36the AI salon so if you go to the salon.ai there's a link tree there again this isn't for pay all you bitter trolley
147. 21:46people not making any money off this jerks uh if you go to the salon.ai um it's a link tree we've we've got to
148. 21:55meet up there so we meet every other week and then we've got an online Discord Channel and there was a guy a
149. 22:02good friend of mine we used to work with 30 we used to work together 30 years ago his kid had a disease and was
150. 22:10misdiagnosed for a long long time and then after a while they finally got diagnosed and they found out it was not good and he took the
151. 22:20blood work they originally had where he was misdiagnosed and popped it into chat GPT and the first thing and he said I want you
152. 22:32to stack rank what diseases you think these symptoms in this blood work might indicate and the first thing it returned at the
153. 22:41top of the stack ranked list was the disease that his son had so like the depth of what these tools can do is is kind of beyond what you can
154. 22:53imagine now one of the freaky things about them is they do what's called hallucinating so they lie a lot I like to call it mansplaining as a
155. 23:02service and it's just the nature of how the things work so over time that'll get better but right now just know that they might lie but
156. 23:12they are incredibly like this these are the most sophisticated tools ever put in the hands of mere mortals
157. 23:21like in the history of humanity it's Insanity what they did is they basically trained all these models on all of the public internet
158. 23:30so these things they're like an idiot savant where they're really dumb at knowing basic [ __ ] but they're [ __ ]
159. 23:38brilliant at knowing all the stuff and so if you can guide them well and say I want you to act like this kind
160. 23:45of mathematician that specializes in this branch of mathematics and I want you to help me think about how I would
161. 23:51solve this problem it'll be [ __ ] brilliant um and it'll just know it you want to know about six different kinds of you
162. 24:01know psychology and how the different theories compare tell it to make a table comparing six different theories it will
163. 24:09all right what did they say you did they said it was hate speech so so I I we we listen to or we were doing a joke about
164. 24:19I like turtles which is this funny meme from like the early 2000s I think where this kid at a carnival was interviewed and
165. 24:27he's like in zombie makeup at Halloween and the newscaster goes you know you know how's your day going he goes I like
166. 24:35turtles just random so then I showed a clip of another comedian that said something similarly funny but it had a
167. 24:44not a bad word but a a a uh a word that's on the list of words you can't say anymore and so I played the clip and
168. 24:53then I sort of just repeated what the comedian said in kind of a joking way and then the scream went boom whatever
169. 25:07I'm back I'm back baby all right where how can I train llm on my own data okay so all right now the hair's starting to
170. 25:17lose it's lose its mind which is that's what it does this time of night this is just this is one of the things we deal
171. 25:23with at least I got [ __ ] hair right okay um so you have a couple of options so what's your name Crum uh as in our crumb
172. 25:32that's cool name like the name uh 8.5 out of 10. um a couple of options you have one of them is claude.ai well you got two
173. 25:42options here oh so start with Claude claude.ai is from a company called anthropic it's a bunch of X openai people and claude.ai is free
174. 25:53and it's got a hundred thousand token context window which basically means you can either upload I think you can upload
175. 26:01up to five documents or you can just paste into your chat up to 75 000 words and start interacting with it it's got a
176. 26:0875 000 word memory basically so so if you just want to train it on like a document you've got or you know
177. 26:18some sort of thing like that just go to Claude paste it in there or upload it and start interacting with it it's
178. 26:23[ __ ] remarkable you have a two hour conference call and and you turned on transcriptions on on Google meet or whatever it is download that
179. 26:31transcription copy it paste it into Claude and tell it to summarize your meeting and give you all the you know to
180. 26:38do's and who was in the meeting and who said what and who didn't speak so you can passive aggressively humiliate them
181. 26:45on email it'll do that for you so one option is you just kind of do an ad hoc in something like Claude
182. 26:52chat EPT if you pay for the plus thing has this tool called code interpreter which code interpreter you can upload a single
183. 27:02100 megabyte file but here's a trick that file can be a zip file so as long as it's less than 100 megabytes you could upload a zip file
184. 27:12full of CSV files or full of PDFs or full of whatever and you can start interacting with code interpreter and what code interpreter
185. 27:22can do is it actually it can actually analyze that data so you can upload 10 csvs and you could say hey code
186. 27:28interpreter what's interesting about this data and it will start writing and executing Python scripts to analyze the
187. 27:35data and tell you what's interesting about it and then you can say um why don't you identify the most significant Trends in this data and
188. 27:44visualize them and then it will and then it will write code to generate charts and graphs to visualize that data
189. 27:52[ __ ] astounding so the those two tools right there code interpreter at openai and Claude dot AI um are just staggering okay so so that's
190. 28:04that now if you're geeky if you're good and geeky and you don't mind getting your feet dirty with little Python
191. 28:11Programming and maybe spinning up a little Google collab notebook then I would explore a project called Lang chain as in language chain
192. 28:21and Lang chain lets you embed your own data your own documents so you can take a directory full of PDFs and embed them
193. 28:31and then you basically store what are called vectors in a vector database vectors are basically how this [ __ ]
194. 28:37works so they basically turn words into plots in mathematical thousand Dimension space basically it's
195. 28:49it's like incredibly complicated but it doesn't matter it just basically reads all your documents learns all the words
196. 28:54in them and how they're semantically related clusters them and makes them searchable that's what it does so once you've got Lang chain once
197. 29:02you've got all your documents embedded you basically have your company GPT and you can do whatever the [ __ ] you want
198. 29:09with it and you can use all sorts of different models and it's crazy so uh if you're geeky go check out langtay if
199. 29:15you're pseudo geeky like I am where like I'm just geeky enough to get [ __ ] sort of working and then when it breaks I get
200. 29:21frustrated and want to smack things um it might be a little frustrating if you're just a tad more geeky where you
201. 29:28don't mind sitting with code for four hours and that makes you put you in your happy place then you'll love Lang Ching
202. 29:34it's not that hard it just requires a little bit of tech chops and then if you're not geeky then just go use those two tools that I talked
203. 29:42about so hope that helps all right that's holy crap there's 150 people in here hello everybody my name is Kyle
204. 29:49Shannon it is a Thursday night it's 9 42 I'm here in Denver Colorado yes I'm live if you're thinking like but he hasn't
205. 29:56answered my question yet why isn't he answered my question why is he just talking to crumb this whole time who's
206. 30:02Chrome I don't see crumbs little speeches little typed in words cuz I answer all the questions and Crum
207. 30:13asked earlier than you did and so Crum got the ants crumb got the crumbs crumb got the crumbs that's how the cookie crumbles
208. 30:22at the learning lab you can ask people that have been here before I got a lot of integrity I definitely I like turtles all right
209. 30:38what'd you do last night I don't want to talk about it all right oh Tick Tock just jumped me to the bottom
210. 30:45bug somebody fixed that [ __ ] bug if anybody here works for tick tock put it in the [ __ ] instead of kicking me
211. 30:56off for seven days for repeating a word a comedian said that apparently it was fine for him to say but not for me to repeat
212. 31:03fix this [ __ ] bug if you're wondering why I'm so wound up I've had a few drinks um all right
213. 31:19I am John screaming in the wilderness repent and turn away all right good day good day good day uh need some AI for my real estate
214. 31:32business well I would start with uh what would be fun to explore is go to oh I don't know why that's a new story
215. 31:43all right a new story just popped up that was stupid okay there's a surprise um if you've got access to chat GPT plugins
216. 31:52so if you pay for chat GPT Plus go searching the plugins there's 800 plugins now there's a bunch of real
217. 31:59estate plugins in there I don't know if any of them are any good because I'm not in that world I haven't I haven't played
218. 32:04in that world in a while but you can if if you've got a plug-in that works like I'm pretty sure there's a
219. 32:11Zillow plug-in and there's local ones in Canada and [ __ ] like that um you can start chat gpting with local
220. 32:18real estate listings so you can do things like you know find me houses in this range in this neighborhood
221. 32:25um now describe that neighborhood now um write up marketing blurbs for each of those properties that I can put in an
222. 32:34email for this kind of client out it'll come so you could just start in an ad hoc way just [ __ ] around with plugins
223. 32:41at chat GPT and and just see see what it makes possible I'm happy to show you how they work I mean basically you just go
224. 32:49find a real estate plug-in enable it and then start talking and say show me listings and it will um so you can do that you can I these
225. 33:02here's the thing to do rather than saying I need AI for my business I would flip it and do this what problems do you need to solve in
226. 33:11your business start with the problem and then and then start to say well uh okay I need to learn how to do social
227. 33:20media okay great chat CPT can teach you how to do that it can write all the copy it can teach you how to do it it can
228. 33:26give you a whole strategy if you're like well I need better images of these properties or I want images of that town well maybe then
229. 33:35you play with mid-journey and start generating images that are kind of sexy images you know if you need to help
230. 33:42writing letters or analyzing lease documents you know all this stuff's really good so start with the problem figure out the
231. 33:49problem is and then start to investigate which tool do I need to solve that and it's not always going to be AI
232. 33:58sometimes the right tool is like Salesforce or probably not Salesforce the uh real estate equivalent of
233. 34:07Salesforce Salesforce is a [ __ ] train wreck but what Salesforce needs is AI to put a [ __ ] Simplicity layer on top of that
234. 34:15complicated nightmare [ __ ] show of a piece of software ask me what I've been using for the last decade I [ __ ] hate that thing
235. 34:30hey you able to access beta plugins for chat GPT I just purchased the monthly thing yes so here's what you need to do
236. 34:39this is non-obvious and really [ __ ] frustrating and why they buried it like this is beyond me but let's get you up to speed Okay so
237. 34:50if you pay for chat GPT plus it's 20 bucks a month then you get access to plugins and code interpreter that I just do I hit print instead of
238. 35:02Plus oh they've added some new things here actual copy um if you go to gpt4 you should see default code interpreter and plugins if
239. 35:14you don't see these that's because they buried that [ __ ] in settings because they're dum-dums I don't know why they
240. 35:22did it and that's beta we don't want everyone using it okay so you go to your settings and then in your settings
241. 35:29there's an option called beta features so you click on beta features you turn them all on why they would not just default this to
242. 35:36being on is flipping Beyond me but whatever there's probably some lawyer involved well we've got less liability
243. 35:44if you don't default them to Beyond and make them explicitly log into our I don't know why it's not focusing
244. 35:52this is another thing Apple needs to fix their their focus they need to fix that ah normally the little hand trick works well well whatever
245. 36:04and then I'm gonna show you these there we go okay that's good enough all right and then the way it works is
246. 36:13you go to plugins you click on plugins let's go find the real estate one we were just talking about the real estate
247. 36:18one we'll turn off web pilot and we will go to the bottom here we'll go to the plugin store we'll go real estate there's only two
248. 36:32real estate agents real estate data and location-based tools maybe zillow's the only one let me go back to huh did I spell that wrong no
249. 36:46let me close it do I have Zillow in here I think I do I think I have a Zillow hmm maybe there's not real estate plugins
250. 37:03real real time photorealistic date enriched real estate data and
251. 37:17location-based tools huh strange that there's not real estate I thought there would be I thought there
252. 37:25was well this is this this is the state of the the industry right now so we've got we've got this thing so let me see um find me recent
253. 37:36listings in 80211 let's see now it should fire off that plug-in so once you have yeah so fired off the plug-in so once you've got these
254. 37:48plugins installed you basically enable them for whatever chat session you want to do it in then you ask your questions
255. 37:53and it goes off and does [ __ ] it seems I need more specific informations could you provide more specific details oh such as a full
256. 38:01address oh I see what it's doing um so it's like um let's see what is the average home price in uh let's say
257. 38:16Aurora Colorado see if it if it can do any of that it might not even be able to do that no it doesn't like it I don't know
258. 38:31what's going on with real estate I know I've looked up real estate before I wonder if they I wonder if they got rid
259. 38:37of something let me try another thing we'll go to plugins we're going to turn off this piece of [ __ ] that didn't work
260. 38:45Top Hat didn't work for me but if you're into real estate you might know how to use it better than I do I'm going to go
261. 38:50to web pilot and so web pilot will go search uh list search websites let me see if it'll search Zillow um use Zillow to find
262. 39:05recent listings in 80211 let's see if it see if it does it it may not because you may have to log into that site
263. 39:19using web pilot so it's going off and doing its thing it might come back and say I apologize I apologize this is a this is We're Not
264. 39:29Gonna help you in that we're not gonna do it dude um look how cool it is he's surfing on the internet using AI it's super exciting
265. 39:45I'm sorry but I was able to directly directly access Zillow however I did find several several other Real Estate
266. 39:54websites that may have information for you Ah that's good that's that's doing it it found realtor.com Redfin Trulia Remax
267. 40:05there's Re Max 80211 Denver Colorado so now it'll pop me out yeah there you go there's the hood there you have it so
268. 40:19with stuff where you want to get really specific like this you just kind of have to experiment and dick around a little
269. 40:24bit like like I just did and and that's hang on um part of the joy of playing with technologies that are this new is that
270. 40:41you get to play with stuff that is truly [ __ ] remarkable the downside of it is this is all spitting duct tape right now everything
271. 40:49like plug-ins are just a [ __ ] train wreck code interpreter is brilliant and mostly works but sometimes it just doesn't um
272. 40:58you know sometimes these things just say yeah I can't do that I'm not gonna do that go [ __ ] yourself
273. 41:04and it's just because everything right now is new and raw right so so the re this channel will not be necessary a
274. 41:14year and a half from now maybe two years from now that's when I'll get into twerking right and I think that's what everyone's
275. 41:20waiting for quite frankly no one really cares about AI they're like when's he gonna twerk I mean with hair like that
276. 41:26why wouldn't he be twerking isn't that what people want to see I think it is twerk while I'm npcing I'll be [ __ ] loaded I'll be rich I'll
277. 41:39be rich I'll be famous wait aren't you the guy that twerks and does that stupid NPC [ __ ] yeah that's me you want a signature no
278. 41:49until then we'll talk about this AI [ __ ] because it's also new and raw and changing all the time like I said that
279. 41:55list right there used to be three sites now it's seven that was like two weeks ago that was three now it's seven
280. 42:02and probably within about a month there's going to be two more added to that that I know of apple is going to be
281. 42:09on here probably around September is my guess right isn't that when they do their fall thing they do their developer
282. 42:14conference conference in June and their Hardware release in September I would think that there's going to be some AI
283. 42:21announcements from Apple in September so right now there's a lot to talk about because there's a lot to explore so just know that
284. 42:33um this shit's a little janky and it's not going to be perfect and don't dismiss it because of that understand that these things are the
285. 42:42worst they're ever going to be and if you watch that thing write you a [ __ ] marketing plan or write you a
286. 42:48Country Song complete with chords even if it sucks and and do it like that if you've ever had writer's block
287. 42:58which if you're a writer is about six times a day that's gone for me it's gone if I'm feeling insecure about my writing
288. 43:07skills I'm like I don't really want anything I mean oh you know what I could do let me just stick a little shitty idea and chat GPT
289. 43:15oh yeah that is a pretty good idea there's more there's more writer's block is gone shit's remarkable like [ __ ] crack
290. 43:30all right I used mid-journey to create a Snowman on a beach for a snowbird dance flyer for a client he loved it that's it
291. 43:38and and what's what's good about that Lauren like why what I loved about what you wrote there is it sounds like what
292. 43:45you did was you know just a Snowman on a beach that sounds like a hilarious kind of parlor trick you'd do in mid-journey
293. 43:50oh let me see a slow man on a beach but you did it for a client right he had a snowbird dance and you're like ah here's
294. 43:56an idea I got a problem to solve I need an image that represents that let me pop into mid-journey do that show it to the
295. 44:03client maybe you'll like it oh I love it how'd you do that how'd you do that so fast that's the magic and that's that's that
296. 44:11trick about don't lead with the AI lead with the problem and then figure out which tool do I need to use to solve that problem and what
297. 44:19you'll find is that these things can solve problems way better way better than you think they can and then sometimes it'll suck
298. 44:31and then you tell it it sucks and it's like oh I'm sorry here's another attempt and you're like no no that sucks too oh
299. 44:37I'm terribly sorry about that let me try again now and sometimes you do that 10 times and it just doesn't get better you're like [ __ ] it
300. 44:47you just go write it yourself but I would say 90 of the time it gives you something that is 90 of what you
301. 44:56need eighty percent of what you need as you get better at prompting you you can you can sort of focus these things
302. 45:02better and better I can consistently get to 80 quality level within 10 minutes and then I spend the next hour or two
303. 45:10that I would have spent getting to 80 refining it thank you whiskey Kyle this is awesome my first time seeing you live fun and informative
304. 45:25the AI learning lab it's fun and informative you can make money Jelly Bean pup mom love the name love the name
305. 45:36all right let's see this is amazing thank you you're welcome I appreciate that where the hoes at that's that's
306. 45:45just disrespectful there the middle man funny but disrespectful all right the detail is crazy good oh on the on the on
307. 45:53the bear in the kitchen making the beef sandwich it was kind of a big long log of beef I guess it was a filet mignon he
308. 46:00was making beef out of it's like a wagyu filet it was pretty funny which AI is this uh the one I was using
309. 46:10with the images was uh mid-journey Carlin was King Carlin was King he went through an unfunny period but toward the end he found it again
310. 46:22he was [ __ ] brilliant in the 70s holy [ __ ] what software free or paid the mid-journey is paid you can do like 15
311. 46:32or 20 images although I think they've cut that out because every it's so busy all the time I think uh you have to pay
312. 46:38for it it's 10 bucks a month for mid-journey leonardo.ai which is built on top of stable diffusion has a free
313. 46:45you get like 150 credits a day that recycles every day and it that equals like maybe 15 images something like that
314. 46:52or slightly less or slightly more than that um if you've got Adobe Creative Cloud you can use a generative fill within
315. 47:03Photoshop you can use Adobe Firefly and if you're geeky um just go dick around with stable diffusion it's all free it's all open
316. 47:14source you can do anything you want with the developer Community for stable diffusion is insane the tools are insane
317. 47:21but you gotta be a little bit of a yeah I like the code because if not it can just be really frustrating and listen it's not that
318. 47:33hard right if you can you you like go to chat GPT and ask it what Google collab is and ask it to explain to you you know
319. 47:41what it's about and then go get yourself a Google collab account and then go to Reddit and see all the crap they're
320. 47:49talking about in the stable diffusion subreddit and they'll say here's a Google collab notebook for this tool
321. 47:54automatic 11 11 whatever it might be and then you can go play with them um so there's lots of free options out
322. 48:00there but you at this stage of the industry you got to be a WeChat geeky the other thing you can do if you want
323. 48:06to generate images for free go to bing.com thank you the bear that's my favorite the dangler lights I like the dangly
324. 48:13lights they look like a pretty little necklace it's like a pretty little necklace I like looking pretty you know
325. 48:18put a little necklace on pretty little necklace um foreign idiot I don't know he seems smart but then every now and then he's just like
326. 48:30he's just like a dum-dum he talks about the pretty little dangly necklace I don't know why he does that it seems unnecessary anyway bing.com
327. 48:40is if you go there and click on the chat button well let's go do it oh wow that's barred that's barred not Bing go to bing.com
328. 48:53so you go to bing.com it's just like Google used to be except no one used it but if you click on chat well now this is gpt4 so this is gpt4
329. 49:03which at open AI costs you 20 bucks a month and here's free and it's connected to the internet and you can dial in sort
330. 49:09of these different modes creative mode precise mode balanced mode and you can go um make me an ill illustration let me just do it with the
331. 49:25microphone so I don't have to type weird [Music] illustration of a vacuum cleaner causing a tornado
332. 49:44and so you just tell Bing to make you a pretty picture and then it shows you this stupid icon of a computer making
333. 49:53a picture that's such a Microsoft thing to do like they take up the entire screen to tell you they're making an
334. 49:59image but there you go there's a vacuum cleaner making tornadoes now this isn't as good as mid Journey it's not as good
335. 50:06as stable diffusion it's not as good as Adobe Firefly but it's free and from what I heard the so this is based on
336. 50:13Dolly which is open ai's image generator from what I've heard the new version of Dolly I don't know if they're
337. 50:22going to call it Dali but the new open AI image generation tool apparently does fully coherent text like you could say
338. 50:30make a billboard that says hot dog and it'll say hot dog and it's apparently [ __ ] brilliant so I don't think this
339. 50:37is gonna suck for very long but it's free so if you just want to do a quick little image for a comp for a little
340. 50:43pitch deck you're doing or or just a someone's birthday card or you you want to do a kids book and you want to make a
341. 50:51thing of them being uh so let's let's do that um make me a children's book illustration of a vacuum cleaner sucking up a tornado
342. 51:04to save the farm so let's say you had an idea I would like to make Jimmy a book and then you're like hey I wonder if chat GPT
343. 51:18could help me write a book oh it can that's a swell little book I wonder if Chachi BT could help me make pretty
344. 51:24pictures of the book oh yeah there they are there you go it is I I don't know like I've been I've been fairly deep in this world for
345. 51:40more than a year now year and a half maybe two and every time it does this it blows my mind now are those illustrations good or
346. 51:49usable not really they're pieces of crap but it's at least in the neighborhood of what I was talking about and if I took
347. 51:55some time to actually work on that prompt are you kidding me you know I mean
348. 52:09it's nuts it's just nuts it's not it's not that that's that's so I hope that helped you Royal what is that obot user
349. 52:17or gbot user cubot user Royal Cube bot user um and the other thing I would do Royal Cube bot user is go to this document
350. 52:29right here wait hang on gotta make the letters smaller the letters are too big I can't see them Kyle they're right behind your big fat head
351. 52:38all right screenshot it now wait hey it there you go screenshot it now this one openart.ai prompt book if you want to
352. 52:47learn how to use these image generation tools this is a godsend you can go to sites like Lexica lexica.a I think it's lexicon.art
353. 52:59lexica.art you can just scroll through a bunch of images and then if you click on the image you can see the prompt that it
354. 53:06was made with and then you can pop that in an image generation tool and make pretty pictures but if you want to
355. 53:12understand how these things work and what they actually make possible this prompt book is just it's great education
356. 53:17and it's really inspiring it'll give you like that that prompt book for me was like three weeks of just Rabbit Hole of
357. 53:25like exploring all different ways you can make images it's it's it's miraculous it's like it's like it's like
358. 53:32an art history degree in a in a box and then you know instead of the art history being the static thing you
359. 53:41learned about old people that made these you know crappy old paintings now you can go make your own artwork in
360. 53:49their style and just your mind will go squid and then there'll be a little drip it'll go squid and drip drip drip you're like
361. 53:58why is my brains dripping down my face you must be using AI just use gigapixel to upscale it if needed there's also what's that other
362. 54:11one topaz labs what's that one called gigapixel oh that's what you said gigapixel from topaz Labs yeah gigapixel is good you
363. 54:20can also upscale and stable diffusion but again you got to be a geek do you have an AI contract I don't know what that means
364. 54:31like am I getting paid by any of these [ __ ] no why do you work for one of them you want to give me an AI contract
365. 54:40I'm doing a workshop for a big company tomorrow um I'm doing another workshop for company at the end of the month but I don't I
366. 54:49don't know what you mean by what I don't know what an AI contract is if you mean am I getting paid by one of these guys no
367. 54:56I just these are there is more of a public service announcement so that people don't come in here and go well
368. 55:03where do I start you start here start at the top one why are there seven can't you just pick one no because they all do something
369. 55:11slightly different enough that they're all worth [ __ ] around with but if you just are overwhelmed and
370. 55:17you're like I don't even know what to do at all go to the top one and just spend the next three weeks there seriously wait oh
371. 55:35not joking not being hyperbolic chat GPT is so powerful I feel like I could easily spend the next decade if I did nothing but
372. 55:50exploring what that thing made possible the 3.5 version not even four not even a fancy new shiny version the dusty old tired 3.5 that's like
373. 56:01wicked fast I feel like I could spend the next decade trying to understand what that even makes possible so even though these tools are not
374. 56:12perfect and even though you feel like you're way behind you're not just go start playing ask it to make a picture of dogs playing
375. 56:20Billiards oh we can do that but that's right it's actually going to be curious to see if it how close it gets to um
376. 56:33make a classic painting of dogs playing billiards in a smoky Billiards Hall let's see what it does we'll see we'll
377. 56:44see how close it gets to dogs playing poker I assume it's gonna I assume it's gonna pick up that style yeah I picked up that style
378. 57:01oops [Laughter] he's got to pull key on his collar
379. 57:15they're all [ __ ] climbing all over the table it's totally on dogs who would play pool [Laughter] never a dull moment at the AI learning
380. 57:28lab yeah oh good lordy lordy lordy okay where how do I train my llm on my own data Decades of legal writing crumb you
381. 57:43need Lang chain Lang chain or wait for three months is my guess Microsoft I assume if you're a lawyer
382. 57:54you probably use the Microsoft Suite of products Microsoft 365 and all um they're gonna roll gpt4 into all of
383. 58:03those products I think it's going to cost 30 bucks per user per month uh for the for the AI tools to be rolled in um
384. 58:11and then Google is going to do the same with bard both of them have cloud storage tools right Google Drive
385. 58:22I don't I don't remember what Microsoft is called but I don't really care um they're gonna those tools are gonna let
386. 58:33you train on your own data I think Apple's gonna come out with something that'll let you train on your own data
387. 58:38if you want to do it right now and you want to be in complete control of it if you if you're not geeky enough to do
388. 58:45Lang chain get someone who is um and put Lang chain together because basically that it's what the Decades of
389. 58:53legal writing you basically just point Lang chain at directories have it embed it it'll cost you some amount of money
390. 59:00for the embedding that's it's a decently expensive process but I think they just cut the price of that like a tenth of
391. 59:07what it used to be it's it's cheaper than a lot cheaper than it used to be and then it will embed all of your legal
392. 59:13writing into a vector database and you can basically have your legal writing GPT so yeah go do that Lang chain get thee to a lang chain
393. 59:25henceforth this Fortnight yon Sparrow see and they said an acting degree would be worthless
394. 59:39cards [ __ ] cards do you have an AI course um I thought about one I have wait where is it oh it's it's over
395. 59:51I have a Graphic for an AI course see Hands-On Workshop topic learning Chachi BT um I actually just put together this
396. 1:00:05workshop for a corporate client so I might spin that off the reason I've been hesitant is um I mean this channel isn't really about
397. 1:00:14doing that I don't mind just teaching it but I you know this channel is obviously a little random and silly sometimes so I
398. 1:00:24think it's probably not a bad idea to put something together so I think I'll do that I mean I've got the graphic for
399. 1:00:28it now so the first step to anything is start with the marketing and and if people are like I'll buy it you're like okay
400. 1:00:38and I'm being facetious there that is not the first step to anything but if you got the Photoshop skills you might as well do it uh
401. 1:00:50I don't know I was on there for two hours and all I did was try to sell us this Workshop he just tried to sell us
402. 1:00:56this Workshop over and over and over and then we went to buy it and it doesn't even exist I don't understand how he's gonna make
403. 1:01:02money with chat GPT he doesn't have a workshop I don't know what he's thinking and he certainly needs some hair product all right
404. 1:01:17how to train your own AI model how do you do it how do you train your own air model well you start with the Lang chain
405. 1:01:26why do they call it Lang chain well these are large language models Lang chains short for language chain chaining language together
406. 1:01:37you see it's rather obvious isn't it get the to Lang chain I've begun the digital marketing and my mind is going wild thinking how I can
407. 1:01:48use this yeah the Okay so it sounds like jelly bean pup mom you're having you're at least having the beginnings of
408. 1:01:57your Kevin McAllister moment the I didn't know it could do that take whatever you're learning in digital
409. 1:02:03marketing and go it doesn't pick any one of these tools just go to Bing Bing's free or go to chatgpt 3.5
410. 1:02:11um the the non-paid version and just start whatever you just learned if they just taught you how to do marketing plans or
411. 1:02:17how to think about social media strategy or whatever it is take whatever you learn and just go in there tell it tell
412. 1:02:23Chachi BT to act like a a marketing manager or marketing strategist and then start asking it questions and start
413. 1:02:30having it teach you what you just learned in your mind like I'm telling you it's quit brain juice drip down your face
414. 1:02:40mama why are you crying oh it's it's not tears it's just brain fluid it's it's spinal fluid why is it coming out of
415. 1:02:47your ears well that happens that happens darling yes darling I think it's time for bed now mama needs to make another marketing
416. 1:02:55strategy oh Lord what do I do uh I use chat GPT from a real estate business how can the other apps help me um
417. 1:03:09I don't know they may not if you want to do if you wanted to do some sort of clever marketing campaign where like I don't know let's say you
418. 1:03:19live in Toronto and you want to you want to make references to local Toronto landmarks and you want to put something
419. 1:03:29humorous like a funny donkey walking down the street with whatever Toronto landmark in the background like you I
420. 1:03:36mean you could get creative with some of the image stuff I don't know that you need to go there um I mean you could you could use things
421. 1:03:46like generative fill within Photoshop where maybe you've got a vertical image of like a house and you want to sort of
422. 1:03:54extend the width you could I I wouldn't I wouldn't worry too hard about what tools what am I missing
423. 1:04:05like like I said the just using chat gbt for Real Estate it's so deep the amount of things that you could do just with
424. 1:04:12these tools especially if you start playing with code interpreter with chat GPT Ruth it's it's um you can dig very very deep
425. 1:04:24um with with code interpreter and Analysis and things like that I I wouldn't push too hard if you really want to geek Out On Tools go to
426. 1:04:32futurepedia.io that little site right there and there's like 4 000 AI tools there and just look at look up real estate and
427. 1:04:41see what's in there um most of that stuff's gonna be pretty crappy but there might be some interesting stuff in there right because
428. 1:04:47when you start putting these large language model tools as simplification layers on other complicated systems or
429. 1:04:54like analysis of things like real estate like these things are going to be really powerful so there's there might be tools
430. 1:04:59at Future pedia that you can play with I wouldn't I wouldn't worry too much about it though like I I would keep doing what
431. 1:05:05you're doing Find value solve problems that you have now using these tools and as you get better at that then you can start start to Branch
432. 1:05:14out when you have a specific problem to solve that these can't solve but I don't I'm not a believer in just manufacturing
433. 1:05:21reasons to use AI I I think it's I think it runs entire companies down rabbit holes they shouldn't be down in the first place
434. 1:05:31they think I'm making money sounded like wait oh no I'm not sounded like numbnut it was funny as [ __ ] they think I'm making
435. 1:05:42money no I'm not thank you thank you the inadvertent comedy is my is my favorite kind of Comedy the comedy where you just stumble
436. 1:05:51upon it and people like say that thing again that was funny and like I don't [ __ ] know what I said no I'm not so
437. 1:05:57thank you for documenting that that's pretty funny that's how the whole you're gonna make money with Chachi betai started
438. 1:06:04like when I first started this channel it's like everyone that came in how can I make money with chat gbt I'm like you
439. 1:06:10don't listen to the people that sound like this uh I think I'm gonna do t-shirts of that one there was another one I had a couple
440. 1:06:18of things I was gonna do t-shirts on I love that answer okay awesome uh can you give an example I don't know what I
441. 1:06:26was doing at the time so if you give me if you ask me more specifically what you want an example of I can
442. 1:06:33can you change your face into another face while live well yes I can frankly my dear I don't give a damn like one of these things
443. 1:06:48I can't apply this effect while live is that a groundhog over your shoulder you're just happy to see me [Laughter]
444. 1:07:00uh comedy it's comedy all night why is he chewing his nails all night I don't know I probably have a vitamin deficiency
445. 1:07:11how do I get text into scientific references wait how do I get text with scientific references ah okay so super important good question
446. 1:07:22did I just kill my slide no no that's the one okay so in this
447. 1:07:37all of these things hallucinate so if you ask chat GPT to explain a scientific concept for you it will tend
448. 1:07:46to explain that scientific concept for you quite accurately and then if you say explain that scientific concept for me
449. 1:07:53and give me references it will give you references and if you click on the links they likely don't exist
450. 1:07:59it's called hallucination these things just [ __ ] out and out lie so you want to use tools here that are
451. 1:08:07connected to the internet so chat GPT is not however there are plugins for chat gbt um that can go search like archive where
452. 1:08:21the scientific papers are so you can use a plug-in and say you know find me a paper on I don't know image recognition
453. 1:08:28in AI that was launched in the last week and it'll bring up a paper and then you can start interacting with the data from that paper
454. 1:08:36you could also use code interpreter here to give it scientific data and start interacting with the scientific data
455. 1:08:43code interpreter because it's writing and executing python code can actually do calculations so even though chat GPT
456. 1:08:50can't code interpreter can so code interpreter is kind of a Magic Bullet but it's not connected to the internet
457. 1:08:57so you have to go to the Internet find your data put it in code interpreter all right claude.ai not connected to the internet
458. 1:09:04so you don't need to worry about that one bing.com is connected to the internet and does pretty well at giving
459. 1:09:11you citations so go to bing.com put it in precise mode tell it you want scientific references and then go check them and they will
460. 1:09:20likely be at least in the neighborhood of appropriate even if it might not be you know completely accurate
461. 1:09:27um pretty good bard.com is connected to the internet Google is full of Geeks they're not good at the humanity thing
462. 1:09:35but they're really good at the technical thing so I'm thinking that bard.com is going to be really good at that
463. 1:09:42labs.perplexity.ai that's just uh that's meta's um llama 2 model it's not connected to the internet don't worry
464. 1:09:49about that one but perplexity.ai is connected to the internet and you can limit it searches to specific sources so
465. 1:09:57just perplexity.ai is at least worth exploring Poe and Pi dot AI or not um then I would go to futurepedia.io
466. 1:10:07and just look for anything that's like there are probably specific tools that are being built right now for doing
467. 1:10:14things like scientific research so it's it's a bit of a mess and again that's the reason I have all these here
468. 1:10:22is like some of them are better at creative writing some of them are connected to the internet some of them
469. 1:10:26you can use really long things with you know code interpreters just a beast in and of itself poe.com you can do bots
470. 1:10:33with pie is super conversational they've all got some interesting little twists like the the Llama 2 models are open
471. 1:10:40source commercially licensed full-on large language models like the 70 billion parameter model is as good or
472. 1:10:47better than GPT 3.5 and it's free and it's commercially like commercially licensed from meta so so they all have their advantages
473. 1:10:57right now and I you know I wish it were more simple than this but what do you want for the most powerful tool in the
474. 1:11:03history of humanity huh what do you want what do you want you people you just can't get enough all right bitter AI guys oh yeah
475. 1:11:19Gen X bitter you betcha all right what's generative AI well well Sally wait what's your name Frosty Anna uh
476. 1:11:339.5 out of 10 on the name like the name frostiana generative AI so chat GPT the G in GPT stands for generative generative
477. 1:11:43pre-trained Transformer those other two just basically means you don't need to bring your own data it's already been
478. 1:11:48trained thank you silver fox hello welcome um and Transformers just the technology that makes it all possible but the
479. 1:11:56generative part means that these tools wait they're all too small now let me make them bigger these tools chat GPT and the like
480. 1:12:05are generating original content so the way it works is you write your prompt one of the misconceptions about these
481. 1:12:14large language models and the image generation tools is oh They're copying and pasting other people's work they're not
482. 1:12:21they're not now the base Foundation model is trained on others people's work so there's some gray area here but it's not copying and pasting
483. 1:12:30the way it actually works is you put your prompt together whatever it is I want you to act like Shakespeare and I want you to do this
484. 1:12:37and I want you to write me a poem it's about a chubby dog you give it all your poems all your prompt and then when you hit the submit button
485. 1:12:47it basically turns that prompt into a mathematical weighting and it says what is the next most probable word that
486. 1:12:54would be the the next most probable word to respond to this based on all of these billions or trillions of words in these semantic
487. 1:13:04word clouds in this mathematical space and it sticks the word there and then it takes all of that plus that word and it
488. 1:13:11says what's the next most probable word what's the next most probable word it's just like when you're in Google and it
489. 1:13:16sort of finishes your sentence for you it's that technology on steroids So when you say write me a letter
490. 1:13:26we've done a fair amount of these in the AI Salon recently is um write me a letter that tells my current
491. 1:13:33employer to go [ __ ] themselves but in a nice professional way that doesn't burn a bridge and it does so when it writes you that letter it
492. 1:13:41seems coherent it seems like oh my God that's brilliant it seems like they must be copying and pasting that that's not
493. 1:13:46how it works it's literally generating word by word or in the case of images pixel by pixel it's it's sort of
494. 1:13:51denoising this randomly pixelated images that's an amalgamation of you know thousands of images that have all been
495. 1:13:59noised in this weird random way it's just crazy it's crazy technology but that's what it means it's it's
496. 1:14:05generating original stuff and that's so the difference when you look at chat GPT it looks like it looks kind of like a
497. 1:14:18Google Search right it looks like you know there's a there's a search box right so it looks like a Google search
498. 1:14:23well well what's Google give you well it gives you a list of links right so you're like it's an index and so you're
499. 1:14:28like okay so it's pulling from this database and just you know sort of popping the results on the screen that's
500. 1:14:34not what what the the generative AI stuff is doing at all it's generating original content based on the thing that
501. 1:14:42you gave it and that's also why um depending on the settings of the large language model of of the chat gbts or whatever
502. 1:14:50you can ask it the exact same prop prompt and hit you know submit four times in a row and it'll give you
503. 1:14:57completely different things each time same with mid Journey same prompt different results every time because it's generating in this sort of
504. 1:15:06controlled but sort of Random pseudo-magic way Arthur C Clarke says any sufficiently advanced technology is indistinguishable from Magic
505. 1:15:17[ __ ] Magic like the more I understand about this stuff technically the more magical it feels like first of all can't begin to
506. 1:15:27get my head around how someone figured this the [ __ ] out at all like holy [ __ ] is it amazing and they can't figure out why it's so good
507. 1:15:39like they know what it's doing technically but they they don't know why it's so good they they have what's
508. 1:15:45called emergent properties which is a it's a very kind technical term for we don't know why the [ __ ] it's doing that
509. 1:15:55Google barred at one point they trained it on some languages and then they had it um translate a language that it had
510. 1:16:06never been trained on and it translated it perfectly and they're like uh we don't know why it did that or how it
511. 1:16:13did that or why it knows that um should we give it to the public sure that's truly incredible all right how
512. 1:16:30can we make sure they are used for good not evil well that's part of what this channel is about what's your name I'm a librarian
513. 1:16:39um it's a good question so there's a thing called alignment um Sam Altman who's the CEO of open AI about three weeks ago announced that
514. 1:16:53moving forward they're going to commit 20 of their Computing resources of their processor time to a thing he's calling super alignment
515. 1:17:01so what alignment is in in machine learning is are these things are the goals and imperatives of these things aligned with the goals and
516. 1:17:13imperative of these things because if they're not and the imperative here says succeed at all costs and at some point these things
517. 1:17:25get in the way of that goal then they're gonna [ __ ] turn their guns on these things and make them go away so they can maintain their goal
518. 1:17:35so so the alignment issue is a really important one that that you've got human imperatives and then you you've got
519. 1:17:41these machine learning imperatives you want them aligned and so open AI is committing 20 of their compute power and they use a lot of
520. 1:17:49compute power so that's a significant commitment to making sure that these things are aligned the other thing is um
521. 1:17:56and this is on a macro level you know kind of the big company level seven or eight of these companies just got
522. 1:18:02together in Washington DC and they basically all agreed to as we're doing as as we as we companies that have tens of
523. 1:18:13thousands of gpus to throw at these things and make them stronger and stronger we're going to do that in a way
524. 1:18:18that that we you know we share information with each other and we we you know start aligning these things together
525. 1:18:25um so they made that commitment so that's happening um I think you'll start to see um if it's not government regulations
526. 1:18:35immediately you'll start to see government guidelines and then there's also things like what we're doing here why this channel
527. 1:18:44exists is I want everyone who comes here to go try this [ __ ] not because you can make money with chat gbt but because
528. 1:18:55the more individuals get literate with this stuff and start using it for small good right you help the pizza shop owner come up
529. 1:19:06with some marketing activation ideas to get more people to come to their shop every little thing that we do right now
530. 1:19:13individually to bend this stuff toward the good it's not going to change this but it's going to change how we use this and so every little
531. 1:19:25incremental step we make in the positive direction is going to be disproportionately impactful and so if we ignore these things or just
532. 1:19:35say well I'll get to it next year or well they're not perfect right now so I'm not going to deal with it
533. 1:19:42that allows a little more space for it to bend in the other direction and so this channel exists in in in the smallest way possible
534. 1:19:52can we just bend it a little toward the light because I think it's I think it's important I think it's important for us
535. 1:19:59individually I think it's important for us you know culturally and societally um and and just you know kind of future
536. 1:20:07of humanity stuff like these guys are playing at a really big level but we actually have a fair amount of deterministic um influence
537. 1:20:19on how these things get used more than more than I think anyone realizes that's that's where my head is with it
538. 1:20:28for anyone new here Kyle goes through all the questions one by one thank you Lauren I love it when my regulars defend my my Ridiculous Ways
539. 1:20:38well he's a Gen X so he's a little old he's a little old he hasn't twerked yet but he's promised that he's stretching
540. 1:20:50oh there will be twitching twitching twerking okay I can't even do a dumb voice right tonight good Lord can an AI go Rogue um yeah
541. 1:21:05so the the thing there so this one this one bing for sure bard pie and Poe all of those the keys to the kingdom sit behind the
542. 1:21:22corporate firewall and they've got safety guard rails all over them the Llama 2 model that meta just released to the world has no safety
543. 1:21:31guard rails on it so you can do whatever the [ __ ] you want with it so if you want it to start cussing and swearing and being
544. 1:21:39all sorts of nasty bias things it'll do that if you want to build a tool to do evil things it'll do that
545. 1:21:47and then you combine that with a project like Auto GPT or baby AGI which are called autonomous agents where you just
546. 1:21:54give the autonomous agent a goal and it starts operating on its own and and prompting itself and then using tools
547. 1:22:04all right let me go Google this and let me go see if I can find that information there let me merge these together now
548. 1:22:10let me see if I can send that person a harassing email see if I can blackmail them get their credit card info get them to pay me
549. 1:22:19twenty thousand dollars oh yep that was successful yeah so yes yes you can and again that goes back to the point of you know
550. 1:22:29are those people going to be out there yeah are they out there right now yes there there are people right now cloning
551. 1:22:35people's voices cloning you know daughters voices using that cloned voice to call her mother and say hey I'm in
552. 1:22:43trouble I need two thousand dollars and the mother sends it because her daughter called her and said she needed two
553. 1:22:50thousand dollars she knows her daughter's voice well it wasn't her daughter that's happening right now so it's not so much that the AI goes
554. 1:22:59Rogue it's that human beings be scummy you know human beings that don't know a better way to you know make money so they get get over on
555. 1:23:11other people in society um so it's really important that we get AI literate because here's the good news
556. 1:23:21the good guys also have access to these tools so as the bad guys get more sophisticated with their [ __ ] the good
557. 1:23:28guys are going to get more sophisticated at catching those [ __ ] and taking them down but yes it's there's gonna be probably
558. 1:23:37within a year or two a significant headline or two or ten that these things did something really bad and and what it will more than
559. 1:23:46likely be is some human being did something really bad Amplified by these tools uh I started a digital download store
560. 1:23:55I'm working on building it out with mid-journey and canva Etc oh that's awesome that's really cool coffee talk and Tech uh 9.5 out of 10 on
561. 1:24:03the name solid name like it is that word you can't say anymore no no I can't say it or it'll shut down again
562. 1:24:14yeah yeah is there a way to utilize AI to gain insight into maximum maximizing real estate marketing leads yeah I would
563. 1:24:23so let me go I'll go show you something because this will blow some mind morning um but we're gonna go use um code interpreter so
564. 1:24:35if you haven't seen code interpreter this is absolutely worth uh worth the price of admission so before I go to
565. 1:24:43code interpreter I'm going to run over to a slight a site called kaggle and what kaggle has is data and did I need to sign into this I don't
566. 1:24:55think so I think we just go get data so I'm going to go data and I'm going to go real estate what did you want to do
567. 1:25:03real estate marketing leads real estate leads real estate tenant conversion German real estate Airbnb house price prediction
568. 1:25:17no that's Leeds isn't right let's see USA real estate data set 10 megabytes let's grab this all right so I'm going to download this
569. 1:25:2919 megabyte file and I'm going to sign in with Google is someone there
570. 1:25:47um all right so we have that okay so now the way code interpreter works so so you have to have chat tpd plus you have to
571. 1:25:59enable it in your settings in the beta beta features or whatever that's called and and when you go to a code interpreter under gpt4
572. 1:26:09this middle one you have this little plus button next to your prompt and so I'm going to hit um plus and so I just downloaded
573. 1:26:19this real estate USA real estate data and so now I'm just going to say well let me let it upload what I'm going to say is um [Music]
574. 1:26:34please inspect the files in this ZIP archive and tell me what this data is
575. 1:26:50so if you're in real estate you've already got your data you've probably got these proprietary systems that you know
576. 1:26:59someone has programmed for the past 20 years to give you specific reports and you're wanting to find interesting ways
577. 1:27:05to look at that data well now you can do it code interpreter is a full-on data analyst so like some fascinating things
578. 1:27:12about it what it's doing technically is it's spinning up a virtual machine and then every time you ask it to do something it prompts itself
579. 1:27:21and then it writes and executes python code to to do that thing so the first thing it did is it needed to look in
580. 1:27:29that zip file so it Imports some libraries it writes some python code and it looks in inside the data so then it
581. 1:27:36says it's a zip archive with with a single file named realtor data.csv and then it writes another python code to go look
582. 1:27:48at what's in there and it creates a data frame I think is what it's called to read the data so so now it's got in its memory
583. 1:27:56all of the data of this CSV and so it's told us their status price number of bedrooms number of bathrooms the size
584. 1:28:04the full address the street the city the state ZIP code the house size and the sold date okay um so I could say um please let me do this
585. 1:28:17please identify um the most relevant Trends in this data and visualize them for me so I live
586. 1:28:32I literally deal what's in this ZIP file I just downloaded it from some random site so it says to identify the most relevant
587. 1:28:39Trends in real estate data we typically want to look at factors like distribution of property prices number
588. 1:28:44of bedrooms and bathrooms size of properties Trends based on location the status of properties so I could potentially upload
589. 1:28:54Education data real estate data crime data economic data uh whatever else uh uh demographic data you know five different CSV files into a
590. 1:29:10single zip file and create all sorts of customized data analysis based on multiple data sets and it will just do it so so what did it say
591. 1:29:23s for the purpose of the analysis we need to assure ensure accurate visualizations we'll drop the rows with
592. 1:29:30missing data so it cleaned up the data for us so if you don't have perfect data it's like that's fine I'll go clean it up for you
593. 1:29:38now that we've cleaned the data let's proceed with the visualizations please note that due to potential presence of outliers
594. 1:29:45we might have to apply some Transformations blah blah blah blah and here's the charts and graphs these are
595. 1:29:51what do they say price distribution bedroom count um what is that something about listings price distribution um
596. 1:30:05you know can you make a heat map showing price versus uh bedrooms
597. 1:30:22I don't even know if that's possible I like like if you know anything about data analysis you're going to have way
598. 1:30:28more fun with this tool than I do because I don't know what the hell I'm doing but you can just ask a ridiculous kind
599. 1:30:34of questions like it would be swell to know this it'd be swell to know that and it just goes off and starts working for you so
600. 1:30:43code interpreters pretty special so just found a bunch of data and look there's a there's a heat map let's see
601. 1:30:55here's price ranges left to right and number of bedrooms up and down and obviously you know up to five or six bedrooms is you know
602. 1:31:08there's a fair amount of those and then once you get Once once you get into the 42 bedrooms you don't have as
603. 1:31:16much data you don't have as much data in the 42 bedroom range so it all kind of goes light yellow on you but you know
604. 1:31:24this is without me knowing anything about what you're trying to do or whatever but but this kind of tool and
605. 1:31:31then on top of this you could say um let me see um is there any city data in in this let's see what it says
606. 1:31:49okay so it's got a city column so um [Music] write me a marketing narrative for the top five cities that have the
607. 1:32:02best bang for the buck for number of bedrooms versus price so now it's gonna do because it's a large language model I can do the large
608. 1:32:13languagey stuff while I'm doing the data analysis stuff so it should be able to identify which of the top five cities and then it knows
609. 1:32:26enough about those five cities and about this is real estate that it should be able to write us five good blurbs for those cities
610. 1:32:36So based on the analysis Brookton known for its cozy community and peaceful environment Brookton offers
611. 1:32:44an impressive range of average of five bedrooms per property all at an average price of fifty thousand dollars good
612. 1:32:50Lord this makes it an extent outstanding choice for families looking for budget-friendly housing options Wyoming
613. 1:32:58don't let the name deceive you this isn't actually a cowboy state it's a city that boasts an average of four
614. 1:33:04bedrooms per property an average price of sixty two thousand five hundred right so yes you can do that now how do
615. 1:33:18um how do you get good at this I don't know you know about real estate you know about real estate marketing I don't know
616. 1:33:24[ __ ] all I know is these tools are this Alice in Wonderland Looking Glass into language and data the likes of
617. 1:33:35which we've never seen before so start exploring and you will discover some [ __ ] Carly Wells I'm in Denver two two LOL
618. 1:33:44welcome welcome oh no another Denver Nick Nick was Nick Nick syric um love our real estate situation right now
619. 1:33:55our real estate situation is a bit of a mess right now isn't it no one's buying anything they're calling it The Real Estate uh
620. 1:34:04winter nuclear winter I don't know no one's looking right now uh you're the best thank you also crumb
621. 1:34:12is in fact the best oh yeah I know crumb is the best it's Friday afternoon here oh no Tick Tock jumped me to the bottom
622. 1:34:20all right I can do a few more questions and I gotta go because you know I'm old I gotta get my beauty sleep how else
623. 1:34:27could I look like this if I didn't have beauty sleep hmm apparently I need some hair sleep got some [ __ ] Buster Poindexter going
624. 1:34:40on there now it started out kind of normal and it's just gone it's gone a little Haywire it's gone a little silly
625. 1:35:00how do you edit a picture in Skybox okay let's go do that because that's a fun one follow along if you'd like to play Skybox
626. 1:35:12.blockadelabs.com and oh we've got a new version beta 0.6.1 I wonder if it's doing anything cool now cooler okay so here we
627. 1:35:27are at this lovely Fantastical snow-covered desert outcropping of a world with a lovely sunset it's lovely look at the sun dancing off the
628. 1:35:41glistening snow it's fantastic oh Juniper you know what I'd rather be a fantasy Forest with fairies and fireflies
629. 1:35:58and daylilies and I'm going to generate that and so what this is is this is a World Builder so
630. 1:36:13if you're asking how you edit the image once it's generated you really don't so blockade Labs right now if you look at
631. 1:36:20the technology under underneath this thing it's surprisingly simple because I can download this image and if
632. 1:36:28you open it look here's our little fantasy forest with fireflies and daylilies look how pretty it's so pretty darling come in come in I
633. 1:36:39don't bite just a little no please come in come in oh do you hear the brook glistening um ooh what's up in the sky oh that's
634. 1:36:53freaking spooky come on you're not getting out of this you're not getting out of this lady's house no wait look at
635. 1:37:00the dude up here look at this that's terrifying look he's up there watching us this is not good this is not gonna go well all right
636. 1:37:11um so all this is doing it's making a flat image that is uh formatted in such a way that can it can be mapped to a 3D
637. 1:37:18cylinder basically a 3D sphere so that's what this is but I saw a YouTuber a tick tock video the other day
638. 1:37:27you can take these images bring them into Unreal Engine drop some 3D objects on them and actually start navigating through them
639. 1:37:36and then you can stitch multiple of these things together but this is this this for me really like if you want to you know if people are
640. 1:37:45like oh hey I can't do anything you know show them something like this where you can just describe a scene and then you
641. 1:37:51know 20 seconds later you can be twirling around in it this this blows my mind a bit because when you start taking this basic idea
642. 1:38:02and then and then rather than just generating a flat image that I map to a sphere I'm generating procedurally created 3D worlds
643. 1:38:11and then I'm also creating game mechanics and I'm also creating 3D objects and then I'm also creating characters and then I'm also creating
644. 1:38:19story lines and then I'm also creating synthesized voices and I'm also creating NPCs that you can actually talk to and have memory
645. 1:38:30we we are we are within spitting distance of being able to describe a triple A game and within a few minutes it's generated
646. 1:38:42and we're playing it so holy flipping crap so now there is a way I can remix this
647. 1:38:57keep this structure and refine the look oh so that's a way whoever who who has Fadi Fadi how do you edit a picture in
648. 1:39:06Skywalk there's a tool called remix where you can if you like the structure that you created here like it'll keep
649. 1:39:12that doorway and these stairs um it sounds like you could change the style of this out and keep the structure
650. 1:39:19so that might be a way to do do something that you're looking for but but just know this is a like this is
651. 1:39:25a 0.6 technology demo at this point it's not it's not you know this is not playable worlds Beer Hunter Google photos can do that
652. 1:39:38yep all right what does code interpreter do I just showed that can you make an app with AI you can make one with code interpreter um
653. 1:39:48you can also have code interpreter you can have code interpreter read data out of a CSV manipulate it and rewrite it as a new CSV
654. 1:39:59[ __ ] crazy it's crazy it's crazy crazy how does the Vista plug-in work um I don't know what's your name dead
655. 1:40:07knee 121 I don't know didn't he let's go try it out did they did that what's it called Vista let's go look for the Vista plug-in the
656. 1:40:23Vista plugin he wanted to go get the Vista plugin I don't say it I did not say it I did not set up as the plugin
657. 1:40:32let me see is it is it a chat EBT plug-in oh it's probably an extension Vista Chrome extension
658. 1:40:58vidsta oh maybe it's vidsta yeah hold vid stuff no I don't know
659. 1:41:13I do not know I do not know if I should speak them or Vizsla vizla did I say Vizsla maybe I did it wrong
660. 1:41:25oh here's Vizsla I got it installed turn your ID into captivating video AI Scripts select stop footage oh oh I did play
661. 1:41:34with this okay so here's how this works okay so this is just in general how plugins work so so when you install a plug-in oh wait
662. 1:41:44hurry hey hang on hang on hold your damn horses I know my face is too big right now and you need my face in the Teeny
663. 1:41:51box I understand what's happening here you don't need to get aggressive on me yeah yeah good plug-in it doesn't
664. 1:41:59necessarily mean that it's active for that session so in any given session you can activate up to three plugins like
665. 1:42:05right now I've got a web pilot turned on so now I'm going to turn on Vizsla Vizsla and so now I'm gonna say um let's see um
666. 1:42:20make me a promotional video doing outdoor activities in Santa Fe New Mexico during Festival season
667. 1:42:38I don't even know if Santa Fe New Mexico has Festival season but it might sure I can help with that so it says let's create a promotional
668. 1:42:52video showcasing outdoor activities video content video idea showcasing the vibrant culture did it fire off the
669. 1:43:06oh this is weird it's actually writing the code as if it fired off the plug-in but I don't think it fired off the plugin
670. 1:43:18although maybe it will maybe it's just writing all the code first oh yeah okay so it did it okay that's weird normally
671. 1:43:27it fires off a plug-in and it hides all this stuff that it does but maybe they've got this plug-in coded
672. 1:43:34in such a way it doesn't do that this is sort of a non -normal way it behaves but okay so video tone exciting and engaging video
673. 1:43:44type promotional video paced medium published platform YouTube 16x9 for the video metadata experience Santa
674. 1:43:52Fe Outdoor Adventures during Festival season description join us as we explore vibrant culture and thrilling outdoor
675. 1:43:59activities in Santa Fe New Mexico um and then it does a bunch of coding and and I assume it just sends that sends that over to to
676. 1:44:12vsla.com I assume and here's a claim code let me copy that because we might need it so now this is going to fire us off to the site
677. 1:44:35and it looks like it's like looks like it's a bad site something was weird with that that that that was weird the the plug-in didn't
678. 1:44:44fire something's weird something's very weird though that just took me to a dead site let me go try to find the vsla platform
679. 1:44:58Vizsla so so here here again is what I tell you about these Technologies these days they're just not exactly where we need to be
680. 1:45:13but you shouldn't be sad in my Google sign in with the Google
681. 1:45:29um yeah it's not there all right so something something on awry um please use the this La plug-in
682. 1:45:45[Music] I did use it uh
683. 1:46:02yeah that's not that that's a weird site I wonder if if vessel is down like it might be possible that this site's actually down
684. 1:46:10I don't know something's weird there I don't know so so what's your name deadney dead knee 121 how does the Vizsla
685. 1:46:19plug-in work not very well oh it's comedy it's comedy you look being nuts using this AI tool
686. 1:46:38hmm all right are you Tom Arnold there's the first d-list celebrity sighting of the evening I got Tom Arnold before I got Rosie
687. 1:46:54O'Donnell tonight at least that [Laughter] oh there's a dog somebody let the dog back into the live hey Tippy hey buddy
688. 1:47:09all right I ask for variable I ask for verifiable sources in the prompt yes but Rebel the the even if you ask for verifiable
689. 1:47:21sources in the prompt chat CPT will give you what looked like verifiable sources but they won't be so that's why you have to use things
690. 1:47:34actually connected to the internet and go check them on your own there listen there there are whole projects right now
691. 1:47:40being built around solving this problem the hallucination problem is a big problem everybody knows it and just just know that where we are
692. 1:47:49with the technology right now if you want things like scientific sources you gotta you gotta verify them and and use tools like bard Bing um
693. 1:48:04claude's not connected to the internet barter Bing right now are connected to the internet all right oh that's meta's version yeah the
694. 1:48:16labs.perplexity.ai the reason I just have it there like they're just doing us all a favor by letting us go play with the The Meta
695. 1:48:24models without having to install anything that's why I have it there hey Tom Arnold hello welcome where's Roseanne West little Dan
696. 1:48:35give me money oh I like how you put that thanks Maria I don't know what I put
697. 1:48:50government guidelines AI is going to be a moral police [Music] well I kind of has that right now I mean I think
698. 1:49:03AI is the moral police in China I think that the the capital industrial complex in America
699. 1:49:17has made it so that we've just given all that [ __ ] away we've voluntarily given everyone all of our data all of our
700. 1:49:24location data our face data our voice data everything we've given it all away um and so they're already tracking all that [ __ ]
701. 1:49:35like this new generative AI stuff all this what this really does is it democratizes this power so it's not just
702. 1:49:42in the hands of the big corporations so it gets a bit scarier because now your [ __ ] neighbor can do that
703. 1:49:48surveillance on all your data because all of our data is already out there these tools are going to make it trivial
704. 1:49:53to be able to [ __ ] with people so um I I mean what I've learned over the past 20 years is that big brother isn't the government
705. 1:50:02Big Brother is like your [ __ ] neighbor and the companies that you voluntarily give all your data to because they give you free [ __ ]
706. 1:50:13oh Tick Tock it's fun videos here's my data oh Google Docs search engine oh it's all free Facebook there's all my friends I don't
707. 1:50:24have to pay for this this is awesome dude you can have it click click agree agree agree this is that was gone 25 years ago
708. 1:50:35all right I gotta go to bed soon I gotta go to bed I don't know what you people are doing keep me up what are those uh baby AGI not baby IO
709. 1:50:45baby AGI it's not a website baby AGI is a project that's not a complete thought due to edits AI Wars a common goal will never
710. 1:50:57be achieved due to AI Improvement AI Wars did you see the last latest mission impossible not no I haven't actually
711. 1:51:05lots of AI references all right I'll go check it out Kyle I hate arguing with chat GPT Claude being they get data wrong yeah I know
712. 1:51:14they do they do they do they do they do how can I summarize my articles that chat GPT Claude Bard says is offensive
713. 1:51:25views llama llama doesn't have any safety guard rails on it even if I say please Claude so okay so here's an important thing
714. 1:51:36chat GPT Claude Bing in particular bard pi and all the [ __ ] at Poe are all owned by companies that have
715. 1:51:49safety guard rails around their large language models all of them will refuse to do nasty [ __ ] unless you jailbreak
716. 1:51:57them you can look at there's just a paper that came out two days ago that they figured out some magic code you can drop at the end of a
717. 1:52:06prompt that jailbreaks all these things they didn't say what the code is but you can Jailbreak all of them but if you're
718. 1:52:13not going to jailbreak them they're going to refuse to do that [ __ ] for you llama llama two doesn't have safety guard rails so you
719. 1:52:21can just go do whatever you want there so try that one labs.perplexity.ai and if your articles are really nasty
720. 1:52:30and you don't you know want to send the [ __ ] out on the internet install it locally with something like GPT for all
721. 1:52:36and then install you know whatever of the open source large language models you can run on these tools and then you
722. 1:52:43can just go nuts give it all your nasty articles and tell it to rewrite all your nasty articles even nastier
723. 1:52:49in the world of AI evil will prevail Janet Jackson very very dark view of things huh I don't know that evil will prevail I think there is a
724. 1:53:04I don't I don't know what the percentage is but there's a non-zero chance that AI is as dramatically good as
725. 1:53:12everyone thinks it will be dramatically bad I think it may be profoundly empowering to humanity I think it may free us actually free us from
726. 1:53:23doing all of the shitty soul-crushing Cog in the machine work that we've been trained to accept since I've been around
727. 1:53:35I've never accepted that because I got a weird brain um and I just can't do that [ __ ] but a lot of people can and do and have
728. 1:53:51accepted that that's okay and that that's how life is what if all that shitty repetitive soul-crushing work what if all of it were automated
729. 1:54:04it was still generated just as much prosperity but all of those hours didn't have to be spent doing the shitty soul-crushing work
730. 1:54:14what if you could take all of that energy of all of those people and they could do something else like whatever
731. 1:54:20the [ __ ] they wanted to do what if that that's as distinct a possibility as the robots are going to kill us um
732. 1:54:36you're awesome Kyle at custom s data analysis yeah yeah I'm hiring me for data analysis that's good what's his
733. 1:54:43background pretty sure he's got a degree in acting perfect bring him on foreign he acts like a data analyst uh well yes
734. 1:54:53if if we uh if we if we do the pivot table on the uh we take the formula the the cell spreadsheet if you say take the
735. 1:55:00a worksheet number of 3.6.2 but I can talk like a data analyst all right I'm Gonna Leave totally agree
736. 1:55:09on the soul crushing work yeah screw the soul crushing work let's let the I mean here's the thing that like the big fear these things are
737. 1:55:17going to take our jobs no someone using these things is going to take your job but what if these things take all the shitty jobs
738. 1:55:24and just do them for us holy [ __ ] oh my god you're an actor not a tech guy I am that's true but I'm an entrepreneur I've been I've
739. 1:55:38started 12 or 15 companies I started one of the first digital agencies in the mid 90s I moved to New York out of school
740. 1:55:44started ran a theater company for four and a half years wrote seven screenplays wrote and performed a one-man show then I started
741. 1:55:52one of the first online magazines one of the first online agencies company called agency.com and then so it's all been about
742. 1:56:00storytelling for me my current company is called storyvine it's an automated video storytelling platform but I've
743. 1:56:06always had you know this this sort of left brain right brain thing where you know I love technology but I hate the
744. 1:56:15gratuitous use of it and I like it applied in creative ways so yeah so I'm just geeky enough to talk geeky but I'm an actor deep down
745. 1:56:30I'm leaving I swear to God I swear I'm leaving I swear seriously this time all right uh what the hallucination
746. 1:56:41problem what is the hallucination problem all right I'll just I'll close on this one the hallucination problem
747. 1:56:49um the hallucination problem is the way these things generate the documents they create the the answers they create
748. 1:56:58they're not actually aware of what they wrote they're writing one they're not writing they're doing a mathematical probability
749. 1:57:06of the next most probable word over and over and over again so it's literally like a calculator outputting a number
750. 1:57:15but they just happened to map that number to a word and then put the word there instead of the number
751. 1:57:21so it's literally just going like number sequence number sequence number sequence number sequence that's doing so when it
752. 1:57:27writes you a full page of text it's it's it's just doing a mathematical probability one word at a time for that
753. 1:57:33whole page of text we as humans read it and it looks like something sentient wrote it because it's like
754. 1:57:44coherent and comprehensive and then it's like you know here's your um here's the proof that what I just
755. 1:57:52said is is the case and then you click on the link and it doesn't go anywhere and you actually look at the link you're
756. 1:57:57like this isn't even real because it doesn't know what it did and it like it's literally just sort of mashing together
757. 1:58:05all of these words that are in these semantic mathematical clouds semantically grouped mathematical clouds and just sticking it on screen
758. 1:58:17so in the future what I assume will happen is some sort of pre-analysis where it'll write the thing it will then analyze it
759. 1:58:26to say hey is there any [ __ ] in here oh here's some [ __ ] why don't we actually go fix that um so what I would do is use Bing or
760. 1:58:34Bard that are connected to the internet and then that way the links that those provide it does a combination of a large
761. 1:58:42language model generation and a search on the internet and it combines them and so you'll have real links with the actual answer so
762. 1:58:53that's the hallucination problem and it's a problem these things are on acid they've taken the mushrooms they I I
763. 1:58:59really don't approve I you know they're very promiscuous they're very promiscuous they they they answer you
764. 1:59:07you know it's like my husband Bob it's like my husband Bob he's like he he answers me and he sounds so confident
765. 1:59:15and then I find out he didn't know what he was talking about they're just like Bob they're just like Bob I don't say anything it would be
766. 1:59:27rude it would be rude it's so sweet Bob but he's a pompous ass all right that's it I'm out I was just a Midwest housewife for a
767. 1:59:43while there married to Bob you know Bob I've met Bob we've all met Bob all right good night good night
768. 1:59:54um have you used prompts Royale I have used pronounced Royale that's pretty cool where you where the prompts battle
769. 1:59:58one another is that that's that one right uh that's right okay nothing is more believable more believable than pie yeah
770. 2:00:05pies pies flipping hilarious but I'll tell you lots of things Nana they aren't I love my mushrooms all right peace out
771. 2:00:13everybody peace peace peace peace peace see you tomorrow