aiLL 9/26/23 - The Implications of AI in Everyday Life

Transcript

1. 0:14foreign happy Tuesday night happy Tuesday night Luna stick what's happening how you doing Tobias long time no see
2. 0:36after a long night of saloning you come to the lab welcome everybody to the AI learning lab happy Tuesday night let's see is this
3. 0:49shirt gonna be useless damn I'm more I know right that was a lot that's a lot I thought it'd be a good idea to put
4. 1:00that together and then I was like whoo it's been a long day I know me too me too long day I think I'm gonna go
5. 1:06change my shirt let people know I'll be back uh tell them to share tap the screen do that [ __ ] let me change it to a black shirt you know
6. 1:16thank you
7. 2:02all right all right all right all right kind of slipping into something more appropriate my little house dress this
8. 2:10whole house dress I do declare hmm very nice a lot of activity going on outside [Applause] not to be distracted I do not want to be
9. 2:22distracted tonight if you know what I mean all right now he has to dress for us too exactly it's exhausting exhausting
10. 2:32all these people people people listen this is the AI learning lab my name is akala Shannon to that we're gonna learn about AI
11. 2:44it is artificial intelligence and you may ask yourself a cow what are your qualifications and I must say to you but I have not cheers
12. 2:57[Laughter] um uh I love my job if it wasn't for all the people exactly exactly what's the
13. 3:16the George Carlin routine yeah I know I want to start the people who hate people Club every time we get started I hate him
14. 3:27forget how that joke goes but it's pretty good the people who hate people Club sounding like South Park Canadian Prime Minister
15. 3:36hello AI fellas hello what's happening 84 folks in here welcome everyone my name is Kyle Shannon this is AI learning
16. 3:42lab that's Ali the AI learning lab um here's the deal here's what this is I'm not only a member Tom's the president very nice
17. 3:55that's why I'm in the woods this week nice oh escaping the people that's good great Salon tonight thank you been
18. 4:03trying to roll farmer everything nice Winston Kyla forgot it was Tuesday due to my irregular brain function welcome to my world
19. 4:14basically if [ __ ] shows up in my Google Calendar it happens if it doesn't I'm like I don't know did I say I'd do that
20. 4:22yeah you know the joys of add welcome to chat add um all right everyone so so let's let's jump over if you are new here and if you're new to
21. 4:35this world of AI and you're wondering like what the hell are these people talking about why is everybody talking about AI
22. 4:41the robot's gonna kill us if you're one of those um welcome and go there right now chat.openai.com that's that's the
23. 4:52official chat GPT website if you haven't played with it go there right now don't pass go do not connect collect 200 dollars
24. 5:01Monopoly is such a funny game like do not pass code do not collect 200 when I was a kid 200 was a lot now it's like two trips to McDonald's
25. 5:12good Lord the hell happened anyway do not pass go do not take two trips to McDonald's go right to chat.openai.com and start playing just
26. 5:21start asking it to write you a song ask it to write you a computer program in whatever language you want ask you to
27. 5:27write you a marketing plan ask it to write you a a meal plan for the family for the week and just watch what happens
28. 5:34and then these other ones you can play with as well Claude and Bard both of those these These are kind of the big three right now
29. 5:42and they're about to get bigger Claude just got a four billion dollar investment from Amazon it's pretty wild it's pretty wild what's
30. 5:53going on a lot of money flying into these things so these things are gonna get pretty serious pretty fast in fact chat Dot openai.com
31. 6:05is about to go multimodal right now you can just do words there pretty soon you're going to be able to upload an image
32. 6:13and have it talk and all sorts of weird crap so go there and play also if you have any questions about AI pop them in
33. 6:21the comments below I'll do my best to answer them I'm a little crispy at a big long uh what you call it Salon tonight
34. 6:29and uh I just got home a few minutes ago not not that long ago half an hour ago something like that watched a little TV
35. 6:36and now I'm getting on the lab so I gotta pep myself up let's get this going I got I gotta get on the internet I get
36. 6:50I gotta get on The Tick Tock is this this is The Tick Tock right is it on a tick tock this is that how do you do the tick tock
37. 6:59welcome to elderly teaching us teaching each other how to use computers [Laughter] are you happy with the choices you made in life generally
38. 7:14generally Ted looks at life what's happening um so if you have comments pop them pop them in the uh if you have questions pop
39. 7:25them in the comments below I'll do my best to answer them and then we'll we'll jump in here we'll do some stuff let me let me
40. 7:31show you a couple of things so these are things I told you just go play with so go there and give you a couple of things to
41. 7:39screenshot also prompts.chat will teach you how to uh how to talk to these things um you can screenshot that Bing
42. 7:49perplexity.ai pi and po are kind of the Kids Table of these large language models but they're still worth playing for so screenshot that
43. 8:02is great for writing short campfire stories for a book any of them any of them if you want really good
44. 8:09writing go to chatgpt4 right there but claude's good Bard's fine gpt4 is the best um pie not so much
45. 8:22you could play at labs.perplexity.ai but that's really just to be able to play with metas llama 2 models that's more
46. 8:28curiosity right now than anything um these are all image generation tools with the exception of hey Jen which is
47. 8:38kind of video and voice synthesis but they also have a translation thing there that's really powerful so mid Journey ideogram Leonardo
48. 8:46are three kind of competing tools that are all really good for their own reasons uh Adobe Firefly and Photoshop generator Phil is good
49. 8:55um promptbook openart.ai promptbook that'll teach you how to talk to these things future pedia is a big old list of products
50. 9:03um one thing I I want to share so um I've been invited by another Creator named Rhonda Phillips who's a creative
51. 9:10coach and a voice over actor and has a whole community of creatives so I'm doing a three part workshop on three
52. 9:18consecutive weeks starting October 4th I think it's a Wednesday October 4th 11th and 18th um three different you know one's a
53. 9:28beginner course one's applying AI to your business and one is uh Advanced AI and that's happening you go to rhondasvoice.com events
54. 9:40and then the salon the AI Salon that meeting I just came from there so we just had we have bi-weekly meetings at
55. 9:47the salon and we also have an online Discord server where we just launched guilds we launched five guilds tonight writing making art AI 101
56. 9:58entrepreneurship and AI in business so those five guilds launched tonight so if you're interested in finding Community
57. 10:07around ai go to there and storyvine.com that's my company so that's all that stuff um big news in the past couple of weeks or
58. 10:21no sorry past couple of days um Amazon investing four billion dollars in uh anthropic the company behind Claude dot AI is very interesting
59. 10:36and um chat GPT is about to go multimodal within two weeks so coming soon to to a chat gbt near you it's going to start
60. 10:48doing some crazy [ __ ] metaconnect conference tomorrow in California AR and AI announcements interesting
61. 10:56interesting interesting interesting I would not be surprised if meta didn't say here's a new large language model that's that's
62. 11:07GPT four like but I think what they're going to announce I I think what they're going to announce are um
63. 11:16AI chat Bots with personalities like there's going to be a young one and a funny one and a serious one and things like that so
64. 11:24all right what's going on over there someone's rsvp'ing cool all right let's see
65. 11:39beautiful all right let's see let's go to the questions who's got a question do miss the salon tonight hope it was great
66. 11:55uh it it was well I don't know I'm you know as the host of those I never know how they go um I talked about some of the big news
67. 12:03items and then we launched guilds with some good conversation and launch guilds and those look really exciting
68. 12:11I want to join a date science data engineering Guild any of that going on oh data science data engineering it's
69. 12:19there's not tired Runners so what I would say is go join go join the salon Discord and start participating in the
70. 12:27community and that could be something if you're interested enough in it you could end up running that Guild so what we're
71. 12:33trying to do is kind of slowly roll out guilds make sure there's no overlapping we have nothing right now in data
72. 12:39science and data engineering so that could be an interesting one so go to the salon.ai go there go there right there
73. 12:50and the third link down is the invite to the Discord all right
74. 13:08which is great for writing short oh writing great uh campfire stories for books honestly any of them short stories
75. 13:19and short campfire stories it could knock those things out like bang just start practicing just go just go right now like while I'm rambling
76. 13:30just go to go to chat chat.openai.com or go to claude.ai either one of those top two and just tell it you're you you're a
77. 13:44master of suspense and at telling spooky campfire stories and and then give it a topic and tell it to write you
78. 13:52a story and if it does if you don't like it tell it to make it scarier or funnier or shorter or longer whatever it might
79. 13:59be and just start experimenting with it just think of it less like you have to write the perfect campfire prompt and
80. 14:06more like you're going to have a conversation with a writing buddy and as that writing buddy does their thing you
81. 14:13go yes no yes no and it will start to narrow what it does and get better and better and better you kind of run it down this funnel from
82. 14:21generic crap to more specific good stuff all right papaya is my boyfriend just have found out he has six personalities variety
83. 14:33exactly which which personality are you dating tonight um if we only had these in a list that we could share
84. 14:43um we do but I haven't so uh not Thelonious not retropunk foreign Thelonious made a
85. 14:58Google list I just need to get it out there so it does exist I just haven't published it anywhere I'll get to it down in the mud I promise
86. 15:09what's the difference between chat gbt normal and chat GPT Plus or Pro it's it's subtle but really important and
87. 15:19there's some things coming that I think absolutely make chat GPT worth paying for I wouldn't have necessarily said that two days ago
88. 15:31but there is some stuff coming that I think is worth paying so let me get logged in here and I would tell you
89. 15:40I will tell you what we got going on I'll do it I'll do it right now I will do it right now you're wrong Dr Brewster
90. 15:48I'm very proud to be a woman all right this is chat so you got your this is your prompt history down here
91. 15:56when you pay for chat GPT plus not only do you have GPT 3.5 but you also have gpt4 the upgrade from 3.5 to 4 is worth
92. 16:05the 20 bucks in my opinion just because gpt4 is that much better but wait there's more and by the way I'm not affiliated with
93. 16:14any of these sites so so don't go getting all [ __ ] and many ideas like oh he's he's just trying to get us there to
94. 16:21get a sign up because he gets some sort of bonus no I don't I should because all I do all night is drive people to sign up for their stupid
95. 16:30[ __ ] service by the way yeah that's why you're all here how do I make money with their [ __ ] come on man tell
96. 16:42me how do I do it how do I make money I look like Jonathan yeah wait what's his name Jonathan Winters ah Alan the
97. 16:48winters ah I don't have any money with this ah okay so you pay money and you get gpt4 but you also get Advanced Data analysis and plugins
98. 17:02plugins um are mostly useless but there's like a thousand of them now and So within that thousand there are probably 25 to 50
99. 17:12that are actually really interesting and good they're little extensions to chat EPT that kind of connects it to
100. 17:18different services and the internet and things like that then you got this bad boy Advanced Data analysis this is worth
101. 17:25the price of admission in and of itself you can upload a CSV file to that and ask it to analyze the data okay so all
102. 17:32of this is cool all of this in my opinion is worth paying for but oh wait it gets better three days ago open AI announced this
103. 17:48Dolly three which is their new text to image generation model and it looks actually quite remarkable um for for a lot a lot of reasons like
104. 17:59that was Dolly two that's Dolly three and it does it does hedgehogs better than you can imagine and you can
105. 18:09name that Hedgehog Larry and it'll remember Larry throughout your prompts so so they announced that three days ago and then yesterday
106. 18:18they announced this in the next two weeks chat GPT is going multimodal you can talk to it it'll talk back to
107. 18:28you and you can and it can see you can upload images to it and analyze it so then you combine that with Dolly three
108. 18:34all of those new additions in addition to Advanced Data analysis and plugins are going to be available first to
109. 18:43people that pay their 20 bucks so you pay your 20 bucks you're going to get first access to something that's going to effectively
110. 18:52change how we use this thing how is it going to change it I don't know what does it mean that it can now see hear draw and talk and speak
111. 19:04and write I don't know but it's Gonna Change so that's the difference the difference is you get a really quick decent large
112. 19:15language model or you got a [ __ ] brilliant one plus you get all this other stuff that's coming with it so
113. 19:21that's it that's it that's all I got to say on that it's all right how you doing brother you're going down
114. 19:32today right right right right right right right right right right right right carry on Boys carry on right why
115. 19:46does he act like this oh digital guards reached level 19 much like Scientology this channel is about leveling up
116. 19:58it's not but it's a fun competition I know we could game of five people engaging with content creators no no
117. 20:09this could be good then we take our our game theory from our game companies where we made 83 trillion dollars and
118. 20:15then we put it in a in an engagement Channel you know it's good it's solid dragonfly thank you very much loves me a good gummy
119. 20:24all right let's see thanks for clueing us novices in uh oh tanker a Kush by the way I love your name I've said that before um
120. 20:35you're welcome and we're all novices no one is expert at this anyone who tells you they're an expert at this [ __ ]
121. 20:46because the thing I just showed you the thing I just showed you what'd you just show us Kyle I wasn't paying attention
122. 20:52chat GPT can now see hear speak and draw it isn't launched yet but when you go from something that can make words to
123. 21:04something that can do all that stuff's Gonna Change how's it going to change I don't know because I'm a clueless newbie just like
124. 21:11you are with it I mean I'm paying attention to it but I don't know what that [ __ ] means does that mean all the prompts that I
125. 21:20wrote are now all like useless probably all those use cases do I flush those down the toilet probably
126. 21:32why am I laughing because otherwise I'd be crying oh man all right Steve-O hey Kyle hey Irregulars was shaking Stevo metaconnect conference
127. 21:48tomorrow that's cool question are you teaching prompts with multi-layers um oh when I do the workshop yeah and what I'm going to try to do
128. 22:01what I'm going to try to do when I do the workshop is you know it's funny it's like I keep thinking about okay you know
129. 22:07what do I want to put in the workshop what do I not and all of this all of this new capability is probably going to
130. 22:14launch right around the time when that Workshop kicks off so I'm I don't know what to teach so so what I'm
131. 22:23likely going to teach are are kind of fundamental ideas rather than specific prompts because specific prompts are going to change so much
132. 22:32that it really is thank you Clint it really is about getting curious about this stuff and it really is about
133. 22:39um yeah multi-layered multi-layered you know having a conversation with chat GPT versus writing the perfect prompt I
134. 22:46think people are obsessed with this idea of how do I write a prompt for this or how do I write a prompt for that it's
135. 22:52it's not really how they work you can write a prompt that'll get you something and then you're like yeah
136. 22:59that's not quite right let me try this oh it's better but still kind of sucks like this oh that's better
137. 23:07and then you start to figure out just kind of these different Pathways to have the conversation and uh and that's it
138. 23:15mods and Irregulars present Turtles uh you deserve a turtle for that you deserve a turtle here we go this is this is not just for the
139. 23:28Irregulars this is for everyone you're welcome in advance [Music] but here live at a front village with my friend the
140. 23:43zombie Jonathan you're looking good Jonathan just got an awesome face paint job what do you think I like turtles
141. 23:58oh it gets me every flipping time totals totals totals totals okay have you commented on the wgas AI demands wins with the am PTP I
142. 24:14have not looked at their um demands but let's go see if we can find it w g a a i um contract
143. 24:29do new wa g a contract explained AI is not a writer
144. 24:44minimum TV writers room Staffing okay that's good so you need a certain amount of writers in a room for a TV show
145. 24:53where's the rest of this oh continue reading makes an exception for solo writer shows uh uh improved streaming residuals
146. 25:08that's good artificial intelligence the wga wrangled guard rails around the use of generative AI per the guild AI can't
147. 25:17write or rewrite literary material AI generated material will not be considered Source material under the contract meaning that AI generated
148. 25:28material can't be used to undermine a writer's credit or separated rights oh I see what they're saying so so in writing
149. 25:35if you if you write a script and then another writer comes in and rewrites more than 50 percent of your script then
150. 25:43that writer gets co-credit and then if I forget the rules might be different now this is what I was writing
151. 25:49and then there's another percentage where they get so well oh I think it's up to a certain point they get co-credit
152. 25:56that's what it is and then above 50 they they get full credit or something like that with Story by the other one right
153. 26:02so there's there's Ampersand credits and there's there's and credits and then there's Story by credits when someone
154. 26:10gets basically written out so they're basically saying you can't take something we start with and then rewrite
155. 26:17it with AI and and say that that we're out so that's that's interesting a writer can choose to use AI when
156. 26:25performing writing services great if the company consents provided the writer follows applicable company policies okay
157. 26:32the company can't require the writer to use AI software great the company must disclose to the writer if any materials
158. 26:40given to the writer have been generated by AI that seems reasonable wga reserves the right to assert that exploitation of
159. 26:48writers materials to train AI is prohibited okay so they don't they don't want to take all the famous screenwriters that are alive Jam all
160. 26:56these things in a model and just say we're going to make the thing based on all your work explicitly that seems that seems reasonable
161. 27:03it I mean it doesn't seem it doesn't seem completely out of touch it says they can use it they're going to use it that seems pretty good
162. 27:12so good on them I like it solid solid is a rock because their starting position was ai's evil we'll never use AI [ __ ] Ai
163. 27:23and this actually says you know don't use AI to basically take our screen credit away from us that's pretty smart that's good and then we as
164. 27:33writers get to use it but Kyle how will they know how will they know they won't listen November 1st this is what what are we
165. 27:43like we're almost in October right November 1st Chachi BT is going to be in all the Microsoft products right Microsoft Word
166. 27:53Microsoft Excel Microsoft PowerPoint this stuff's gonna be everywhere the it's it's in notion it's in all the
167. 28:02writing tools it's going to be in every one of the screenwriting pieces of software Guerra won't Tee It
168. 28:08they'll resist for a month or two and then Final Cut final draft or one of them will go oh now there's an AI
169. 28:16assistant my AI co-writer whatever the [ __ ] they call it and then they'll all do it so it's it'll come
170. 28:23holy acronyms Batman producers can use scripts for AI models wait no it said they can't use scripts for AI models didn't it
171. 28:33hold the wga reserves the right oh they do oh wait to assert that reserves the right to assert that exploitation of writers materials to
172. 28:44train AI is prohibited the [ __ ] does that mean they reserve the right to assert that the exploitation is prohibited
173. 28:59oh the wga okay so the writers reserve the right to assert oh oh I see what they're saying so what they're saying what they're
174. 29:11saying right now is they can't assert it right now because it's too nebulous like how all this [ __ ] Works basically
175. 29:21but they're basically saying once there is sound law they reserve the right to be able to assert that it's
176. 29:29prohibited by law okay yeah I get it okay send it to Claude yeah in fact let's do that copy well actually you know what maybe we can
177. 29:40get the is there a link to the actual um wga contract 2023 d
178. 29:58tentative agreement summary of the deal terms all right copy oh wait not that select all let's see summary of the name the meat
179. 30:15copy oh here let me go do this all right so I just went I got this is a summary it's not the actual contract but it's a summary from the
180. 30:24oh no hang on hang on this is not the WG hang on let me see if I can actually go find it [Music]
181. 30:42foreign tentative agreement 2023
182. 30:58is the set this looks like it so what is this is this a PDF yes okay
183. 31:21copy all right so here's what we're going to do we're going to hop over to chat jbtay we're gonna go Cha-Cha by Ty
184. 31:30and then we're going to go to plugins you all were asking about plugins what do you do with a plug-in Kyle I don't
185. 31:36know nothing about breaking what do you do plug in how do you use the plug-in hey Kyle Kyle Kyle listen how you plug in hey Kyle Kyle
186. 31:46so we're going to use the plug-in web pilot and webpilot has the ability to go out on the web and read [ __ ] so we're gonna
187. 31:54say based on this agreement please help me understand oh wait this agreement and then I'm going to paste in the agreement
188. 32:09please help me understand the section where the wga reserves the right
189. 32:25um to deem use of AI unlawful I forget what it actually said
190. 32:40but let's let's see how this does
191. 32:58based on the provided agreement it appears that the writer scale of America has addressed the use of artificial
192. 33:03intelligence in the context of MBA covered projects the here is a relevant section AI regulation the use of AI is regulated on
193. 33:16MBA covered projects AI cannot write or rewrite we got that one AI definition AI written material is not considered
194. 33:22literary material Source material AI usage of writers disclosure Guild's rights okay Guild's rights so um
195. 33:37explain the guilds rights I'm sure there's someone on here that knows laws that just is [ __ ] laughing right now but
196. 33:49but okay it's actually a good point so if you know law the thing that I'm reading you're like well it's obviously this
197. 33:57is one of the things that Chachi PT is is good at like if you're ignorant about something you can get onion you're
198. 34:05ignorant about it really quick without having to get a [ __ ] law degree so explain the guilds rights section uh since it sounds um
199. 34:17nuanced and convo luded to me see there's a fancy word with multiple syllables convoluted um I remember I used that once and and my
200. 34:30mom and my uncle are like what the [ __ ] is that word I was like I gotta get out of this town all right all right convoluted uh to me uh
201. 34:40talk to me like I am in fifth grade all right so imagine you have a big box of crayons that wasn't quite far back I
202. 34:56wanted to go and you make some really cool drawings with them you don't want the other kids to use your drawings to
203. 35:01say you made okay so the Guild's rights section is kind of like saying that the writer's Guild oh can say no if someone tries to use
204. 35:11AI models to teach a robot how to do it so the writer's Guild can say no in simple words this section means the
205. 35:21writer's Guild can stop people using what writers have written to teach computers oh I see okay oh okay we
206. 35:29reserve the right to sue your ass if you train if you train your models on our work we reserve the right to sue your ass all
207. 35:38right right makes sense perfect you we have laid down the law go ahead and train it on our [ __ ] we reserve the right to take you down for
208. 35:47it I like it all right what would be the best practice for Real Estate lead generation using chat GPT
209. 35:53um for lead generation probably not chat GPT although let's go look at while we're here let's go look and see I'm pretty sure
210. 36:03that there's it depends what you're trying to if are you trying to find customers are you trying to find properties
211. 36:11because I know that there are plugins um their real estate plugins but I don't know that there's real estate
212. 36:20customer plugins uh real real estate Top Hat enhanced real estate data and location-based tools why are there so
213. 36:36few of these real estate huh there's very few there's very few of them huh strange
214. 36:52I think there used to be more plugins that were real estate related this is this is real estate agents
215. 36:58discover the best fit real estate agents tailored to optimize your home and then this one's called top hat
216. 37:07um I would probably go to something like Bing or Bard that are connected to the internet and then basically just start asking it
217. 37:16like here's the deal it's it's it's not so much um you know that you want to use chat GPT to find what did you say you want to you
218. 37:27want to use them to find prospects or whatever it is it's like what are the steps that you normally do to find those
219. 37:33prospects and then have chat GPT or have Bard or Bing say to you here's how I would approach that problem like like
220. 37:43put chat GPT in the expert seat and ask it to guide you on how you should Prospect for customers that might be the
221. 37:52way to do it and then the ones that are connected to the internet can actually find sources to go do that
222. 37:59that's probably how I'd do it but chat GPT as it stands is not connected to the internet so I would use Bard or I would use bing
223. 38:08bing bing bing bing bing bing bing why is it grayed out Kyle because it was a bad girl he's a bad girl Sydney was [ __ ] to me
224. 38:18so I made her Gray [Laughter] fat girl she's a very bad girl um it's awesome Kyle exactly get on ignorant get on ignorant people you know
225. 38:35you're sitting in here all ignorant and you need to get unignorant people customers oh customers selling properties yeah again I would go to
226. 38:50um see Bing's probably got um some [ __ ] safety things I can't look for properties for sale because you might stalk someone
227. 39:02Bing's really on my [ __ ] list right now so let's try bard I'm gonna say
228. 39:20oh I think Siri took over my microphone again there she is I knew it
229. 39:37Siri go away I didn't get that could you try again go away okay bye uh I I can't oh no not crashed my
230. 39:53browser God dang it oh Siri dang featuring Anderson Puck by Mac Miller now playing [Music]
231. 40:10laughs [Music] yeah you can keep on losing me you lost
232. 40:26me you lost me at dang what I think my machine is is really crashing at this point anyway I would go to Bard ask Bard to ask you ask Bard to
233. 40:39tell you how to source the best way to find people selling their properties have it give you strategies and then see if you can get
234. 40:48it to execute those strategies that's how I would do it sorry I would share it but my machine is just crap in the bed
235. 40:53the crowd chants send it to Cloud send it to Claude Kyle's like I am shut up Siri exactly oh man Siri and I know listen listen I am
236. 41:05well aware that I could go into my settings if I choose chose to search them deeply enough and turn that stupid thing off
237. 41:14but that wouldn't be funny would it wouldn't be funny to watch me get exasperated I I feel like [ __ ] Bill
238. 41:21Burr every time Siri comes on I feel like Bill Burr is losing his mind I guess on the basis of plagiarism they
239. 41:31would have to compare similarities of the writing um that's just such a [ __ ] weird slippery slope I don't I here here's the
240. 41:43deal Clint these large language models are not copy pasting you could get a large language model even that wasn't
241. 41:49trained on someone's writing to sound a lot like that person's writing um so I don't I don't quite know how they
242. 41:57do it I think that's probably why they're saying we reserve the right because there's probably listen here's the deal these tools
243. 42:09these tools and ones like them are the most profoundly powerful writing tool in the history of humanity it
244. 42:20sounds to me like the writer's Guild kind of figured out we don't want to hobble Innovation on a tool that could ultimately help us a lot
245. 42:30but if if you if you train these things responsibly and they turn into tools that can really help us speed up our process and get
246. 42:41things into production faster and make us all money faster great but if you cross the line and train them too
247. 42:48explicitly on our specific writing styles and try to replace us with our synthesized writing then we'll come
248. 42:55after you that feels reasonable to me if like it doesn't feel as short-sighted as their initial position was I assume
249. 43:04that sag that this also to me feels like this is a good foundational model that sag could use for the actors
250. 43:12synthesizing the likeness of an actor I think it means if you trained a bot using all JK Rowling's and ask the AI
251. 43:19uh AI to write a children's wizard book yeah that that's essentially it it's it's basically the wga you know if
252. 43:30the producers are gonna have to be smart about how aggressively they train these models kind of what it says
253. 43:40it's it's it's it's kind of a don't be a dick Clause if you're a dick we'll sue you as long as you're not a dick as long as
254. 43:49as long as you're recognizing that we are an important part of this process and you don't cross whatever that
255. 43:56imaginary line is then we'll we'll then we'll play nice too they would probably use AI to find out if it was written by AI the problem with
256. 44:04that joker is like like open AI removed their their AI detection tool off the market the AI detection tools um provide
257. 44:13many many false positives they unfairly Target um non-native uh English speakers um and and the fact of the matter is
258. 44:28almost all writing that I do today is neither a hundred percent me or a hundred percent AI it's all this hybrid kind of work right
259. 44:39because I'll go have it help me give me an outline and then I'll write some [ __ ] and then I'll have it fix some [ __ ] and
260. 44:44then I'll have it do this and I'll have it do that like that's it's just gonna turn into this muddy [ __ ] mess
261. 44:49so so then you you have an AI analyze which parts of this are Ai and it it doesn't and then you're gonna be in a
262. 44:55[ __ ] lawsuit for months ago and that line I wrote I remember I wrote that I have notes on an index card and it's
263. 45:03just that's not that's not tenable I I don't think for either side I don't think either side wants that either so so the the whole the whole
264. 45:12AI detector thing is is going to be a very short window from 2023 where we're like that was cute that we had that idea
265. 45:22that's going to go away all right um You would tell AI to write as if it was not written by AI yeah exactly you
266. 45:30can there's there's eight billion ways to circumvent those things train at AI to find out if it was written by AI that was ordered not to
267. 45:38write like AI exactly Manta like it it's it's a ridiculous circular circle jerk all right is good I like Cowboys um
268. 45:57Cowboys is tasty all right let's see carry on all of you welcome welcome to the AI learning lab where we talk law even though we know
269. 46:12nothing about it I can't even parse a basic sentence if it was written in the triple negative [Laughter]
270. 46:24yeah yeah see yeah yeah you palooka yeah see hey Kyle cool shirt Cassius Clay yeah yeah see yeah or maybe that was the
271. 46:34other shirt before I left I don't know I don't know what's going on here anymore discussion Engine versus answer
272. 46:40generation machine absolutely absolutely totals toodles is tasty the video was on Facebook today that I was notified about
273. 46:52which video this video this video watch video holy acronym Batman what would be the best practice okay oh
274. 47:04sorry I'm I'm way back on Old questions ah if there's any mods in there toss up an interesting question it's awesome Kyle
275. 47:15exactly I read that one already get an ignorant get on ignorant yeah that was like 15 minutes ago what the hell am I doing here
276. 47:26Bard gives me things like most cash sales in given zip codes yeah I don't know you know what might be worth worth trying Diego Diego Fuego
277. 47:38Flame nine out of 10 on the name Diego Fuego flame I like it you know what could be interesting to try wait I just had an idea and it went
278. 47:49out of my head uh uh um I don't know I lost it sounds like spy versus spy from Mad Magazine
279. 48:03very funny gotta love Siri that's hilarious hilarious and terrible oh
280. 48:18yeah yeah so Ted looks at life has the answer there Diego Fuego yeah download MLS data that's that's it download MLS
281. 48:26data as a CSV file upload that into code interpreter do you know how to do code interpreter Diego Fuego Flame it's a great great name
282. 48:41you should also turn Alexa on and Google assistant and others I really should that would be a fun thing well it's it's
283. 48:48that that one woman who's got the tick tock that's really funny she sort of really dry and nerdy and scientific
284. 48:54she's got that thing where she acts like Alexa and Siri and they're sort of talking and like chat gbt comes in and she's all she's all like
285. 49:03bragging about how awesome she is and they're all like she's great she'd be great at that um all right
286. 49:17triple negative is still negative hey that's like math Joker we try not to do math in this channel how was the salon I was out eating
287. 49:28guacamole you win uh the salon was good the Slime was good we talked about news so four billion dollars into anthropic was news item one
288. 49:41um Dolly three was news item two and then uh Chachi PT going multimodal in two weeks was news item three and all of
289. 49:51those things kind of played into each other in fact I have a a slide that I did I used our the beginning is here
290. 50:03slide that we made in here one night hang on get out of my way see the beginning is here
291. 50:18um this is this is kind of wild so AWS Amazon invests four billion dollars well they invested 1.5 billion with the right
292. 50:34to invest up to four right so it's essentially a four billion dollar investment so that's Amazon anthropic here at the
293. 50:43bottom you've got Microsoft and openai Google and deepmind Tesla which is effectively an AI company an xai
294. 50:51you got meta doing whatever the [ __ ] they're doing so you got those five kind of major entities that are you know these really
295. 50:59interesting pairings then you've got Apple they increase their r d budget to 22.6 billion dollars in 2023 mostly on AI you got inflection
296. 51:10that raised 1.3 billion dollars you got Oracle and Co here and there's probably two to five that we don't know about or
297. 51:20or that I've just missed so what is that that's five six seven eight so let's say there's ten major like billions
298. 51:30billions of dollars investments in competing Technologies and competing efforts and and so what this says to me is AI becomes
299. 51:44like infrastructure level capability just like the internet like right now in the early days of the internet there
300. 51:51were like internet companies and non-internet companies right now the Internet is just it was just everywhere
301. 51:56right it's in all your business like we don't talk about the internet you know being here and not being there
302. 52:02that's how we're talking about AI right now oh well over in this corner you can use AI in this corner you can't what
303. 52:08this says to me is this [ __ ] is going to be just like water if you're online AI will be there and
304. 52:16like this kind of magical crazy ass [ __ ] that we're using you know today in chat GPT it's just going to be in everything so
305. 52:26that's wild so that we talked about and then we talked about you know the dolly three and the multimodal so so OB AI in
306. 52:34the next two weeks is going to fundamentally transform from text only to some crazy ass new kind of thing and then Gemini is
307. 52:45going to come out probably sometime in October or early November from Google and it's going to be multimodal so all
308. 52:52of a sudden we go from you know my my silly little my silly little charts here where I've got these three text models and I've got
309. 53:01these four text models as my top seven like this probably drops to two or three multimodal models and we drop all these
310. 53:14text only things to the bin so I don't know I don't know what's gonna happen it's crazy so can you run that past me again upload
311. 53:25a CSV file of the MLS to where yes Diego Fuego flame because you have an awesome name I will go show you how to do it
312. 53:35and there's there's a lesson people if Diego Fuego flame didn't have a good name and said hey could you show me that
313. 53:44I'd be like no not with a name like that not with a 3.2 out of 10. no never um so we go to chat JPay Tay chat.openai.com
314. 53:56and you're like Kyle maybe you could watch your screen a little because because it looks like a dog took a poop on it Okay so
315. 54:04in order to use what what they now call Advanced Data analysis um you need to pay for chat GPT plus once you
316. 54:15pay for chat GPT plus it doesn't just automatically show up why because it's stupid it's almost as stupid as Siri so you get
317. 54:26a settings in beta and then you go to Beta features and then you turn on plugins and Advanced Data analysis okay
318. 54:34if you don't do that you won't see it you'll be like but I paid about twenty dollars with my Advanced Data analysis
319. 54:41Kyle why do you call it code interpreter all right so then you click on your gpt4 and you go to Advanced Data analysis
320. 54:51all it does all it does is add this little plus button right there so here's what we're going to do I'm going to hop over
321. 55:00to a site called kaggle k-a-g-g-l-e oh I spelled it wrong but here we go and I'm going to sign in sign in with the Google
322. 55:14that's what you call your single sign-on the oauth action action and we're going to go to data sets and then I'm going to
323. 55:21type in MLS MLS there should be some MLS data sets in here MLS data examples here we go how much I don't want notebooks I want CSV files
324. 55:36let's see there's 28 of them oh let's see multi-listing service multi-listing 2018 is this
325. 56:01no that's soccer that's soccer we don't want soccer we want real estate listings all right USA real estate data set how
326. 56:16big is it 11 megabytes okay updated a month ago so this isn't up to date but we're going to download the file it's an 11 megabyte file
327. 56:28bang it's in archive12.zip and then we're going to go back to that Dr chat JPay tag okay so once you've chosen
328. 56:41um Advanced Data analysis or what was code interpreter it now adds this little plus button here so you're going to hit
329. 56:47that plus button and you're going to go find your data file so we've got this archive12.zip and then we're going to say to it I
330. 56:54don't know what's in there because I didn't I don't read manuals I didn't read what's on that page so I'm just
331. 57:00going to say I'm gonna see if Siri lets us do this this time please look at the enclosed CSV about real estate
332. 57:14and tell me what's in it and how we might analyze it and so what this tool does is it it does exactly what it sounds like
333. 57:27like I can just talk to this thing like I'm an idiot and know nothing about pulling data out of a real estate CSV file
334. 57:36because that's the that's truth and it will understand it and so what it does is the first thing it does is is it
335. 57:43unzips the file how does it do that it writes its own python code so it Imports a library to to unzip it it then writes
336. 57:55the code to unzip it unzips it and then and then it writes code to go look at the data it says let me open
337. 58:01pandas and and go see what's in the uh in the thing there it is so there's what's in there and then it writes us a little analysis
338. 58:10there's status of the real estate number of bedrooms number of baths size of Lawton Acres potential analysis strategies
339. 58:22um data descriptives okay whatever so I'm going to say I'm a real estate investor and I would like you to coach me on how to
340. 58:36find the properties that are the best investments period give me three strategies for how I can game the system
341. 58:49saying something like how can I game the system it's safety uh guardrails May kick in and go I don't recommend that
342. 58:55it's important to invest investing with it's a it's important to approach investing with integrity and within
343. 59:02legal and ethical bounds therefore it's not appropriate to game the system but I can certainly provide you some
344. 59:09strategies you just can't joke with these things and my God people all right so identify under price properties
345. 59:17analyze market trends focus on class cash flow and return on investments would it be possible for you to combine
346. 59:26those three approaches into a single strategy period what might that look like um what might that look like yeah absolutely combining these
347. 59:39approaches this is called the univ unified real estate investment strategy or the Yuri's Yuri's Yuri's it's it's the urine strategy urines
348. 59:55u-r-e-i-n-s urines you heard it here first folks how did you make his millions uh we don't really talk about it
349. 1:00:10um identify Emerging Markets Okay so would you like any specific um I'd like to see you identify
350. 1:00:26three possible properties for me to invest in using this strategy and provide an overview of why each property
351. 1:00:37is a good investment in your opinion and I'm getting holy moly and Diego Fuego flame has got Flames there this is
352. 1:00:51insane isn't it like listen if you know something about real estate you're going to be able to do this way better than I
353. 1:00:58am right like I don't know [ __ ] about this so let's see what's it gonna do we're going to filter the properties that are
354. 1:01:07for sale we're going to group them by location oh now let's group them first so let's group them first period
355. 1:01:17find me properties in Clarksville Tennessee that we can do this with I had investment properties in Clarksville Tennessee that I got totally
356. 1:01:29ripped off on so I'm a little bitter about it I I was I was part of the Carnage of 2008 with a shitty investment partner that was a disaster
357. 1:01:46um um sure how about coffee properties in Colorado
358. 1:02:01that's where I'm living right now I'm in Denver Colorado where does he live he lives in Denver Colorado really
359. 1:02:09he doesn't seem like he's nice he seems like an East coaster oh he is he used to be he's he's soft now though
360. 1:02:18because he moved to Denver and he got softer really yeah that's what I heard that's what he says I don't know if I
361. 1:02:25believe him I think that apartment could actually be in Manhattan because he seems he seems too bitter to be west of
362. 1:02:32the Mississippi I mean sure if he was in LA it would make sense anyway uh let's see oh my God okay there's nothing in Colorado
363. 1:02:43please tell me a city that does have some properties for sale well you get the basic idea here Diego like you can you can figure this out
364. 1:02:53from here all right yeah I'm done with you I'm done with you let's try one more time regenerate come on you can do it
365. 1:03:00you can do it you big damn dumb dumb it bombed out all right whatever sometimes it just pick up some and
366. 1:03:11jumps the shark it just jumped the shark so but that gives you the idea that gives you the real that you know that
367. 1:03:18gives you some idea what we're going to do in here you know what I mean you know you know what I'm saying you know it's right like
368. 1:03:26when I invest in property you know I'm thinking first do I go single wide or do I go double wide and I'll tell you what
369. 1:03:34you don't go triple wide I'll tell you why that's just a house okay I mean I know you can move them in Parts but it's
370. 1:03:41a big pain in a hoodoo anyway so start with that and then I and then I look you know is is it is it rusted out
371. 1:03:51and if it's rusted out I'm like no I'm not touching it and you know sometimes I put paint on it make it look like it's
372. 1:03:56not rusted out but I know how to spot that you see the you see the bubbles you know you see bubbles in the nice paint it ain't so nice Okay
373. 1:04:05so anyway can any of these AIS make me a logo yes yes they can gonna go make one where am I gonna take you I'm gonna take
374. 1:04:18you to the AI Salon this is where we were tonight and look we just opened guilds we've got making Art Guild a
375. 1:04:26writer's Guild AI 101 entrepreneurs we got guilds yay we also have in AI salon and by the way if you're sitting there going why are
376. 1:04:35you talking about this hair salon I don't know where that is where's that ass along Kyle that's not fair why are
377. 1:04:39you talking about something I can't get to well there it is thessalon.ai go there and there's a link to the Discord
378. 1:04:47there's also a link to our Meetup you can come to our Meetup all right anyway folks so now we're in here we got this thing
379. 1:04:54called mid-journey play so we're going to go in here if you if you ain't ever used mid-journey it is miserable
380. 1:05:03but it makes the best images so we're gonna go slash imagine uh and then it says prompt and so then what
381. 1:05:14we're going to do is we're going to go logo of a hedgehog
382. 1:05:29for a lawn mowing service where the Hedgehog has wheels instead of legs and is bearing
383. 1:05:44its teeth chomping at the grass it's very possible it's going to be a horrible logo [Laughter]
384. 1:05:58like possible I think it is highly likely this will be a horrible logo um one thing that I would strongly encourage you to do
385. 1:06:11is go to the Fast book go to fastbook fastbook um [Music]
386. 1:06:29[Music] and then search for mid-journey official and then you're going to go to mid-journey official and then within mid-journey official
387. 1:06:42search for Logos or just logo and then it will show you you know just piles of things that they've
388. 1:06:59curated out of their community that are good things so like what I would do is I would just scroll through here until you find the
389. 1:07:09style you like right like if you like that kind of you know 90s sort of vector style or if you want something more 3D
390. 1:07:15or something more retro um like that's got a kind of look to it so so that that would be a good thing to do
391. 1:07:25let's go see if our Hedgehog is awful or not um where is it there it is oh it's actually not bad oh those are really
392. 1:07:32good they're cuter they're cuter than a little hedgehog it didn't quite replace the legs this one is sort of did
393. 1:07:43but that's like a really good one that's like an aggressive thing so so now that you've made um now so in mid Journey once you make an image
394. 1:07:54you now have some options you've got these V options and these U options U stands for upscale and then v stands for
395. 1:08:01variation so I like this number too it goes one two three four so I'm going to do V2 which means give me variations of logo number two
396. 1:08:26and I'm going to show you one other tool other than mid Journey because mid journey is a pain in the [ __ ]
397. 1:08:33that's his brother Badger it is kind of a badger isn't it it's not it's not really a hedgehog Take the Last Train to Clarksville
398. 1:08:42that's good this is good all right these are great actually okay I love three so we're gonna upscale number three boom [Music]
399. 1:08:58look at that that is a logo that's a t-shirt man if I had Badger the Hedgehog's lawn cutting service that would that would sell me
400. 1:09:10some uh some I'd sell me some merch with that [ __ ] now notice in here there's there's this second bot that I've got there's the
401. 1:09:20mid-journey bot that makes the image there's this thing called Trace Journey which because they're jerks they don't give
402. 1:09:29you any free free any free tests but there's if you pay for the subscription I think it's 10 bucks is the minimum one
403. 1:09:38you can vectorize a mid-journey so so these things are relatively low resolution like I think this is 10 24 by
404. 1:09:4510 24. so if you wanted something to put on a t-shirt you you know you'd ideally turn that into a vector well there's a
405. 1:09:52tool that does that called Trace Journey so that's pretty cool there so now the downside of um whatever you call it
406. 1:10:03what is this mid-journey the downside of it is it can't do text so we're going to go over to a different tool now called
407. 1:10:10ideogram and this is on my little list of hit parades Ido wait i d o Graham dot a i am going to log in
408. 1:10:25an ideogram does text but ID grams shitty at images it's not shitty at images it's just not mid-journey good so I'm going to take
409. 1:10:35the exact prompt that I gave before logo of a hedgehog for a lawn mowing service we're going to give the lawn
410. 1:10:41Mong service a name the lawn mowing service called Larry's Larry's hedge Hogs
411. 1:10:55[Laughter] and that's pretty funny that's comedy right there that's kind of an inside joke because that was uh
412. 1:11:06where the Hedgehog has wheels instead of legs and okay now the other thing you can do here is you can you can pick some of these
413. 1:11:15modifiers that that will help like I'm going to put typography in there I'm gonna put um illustration in there and let's go vibrant
414. 1:11:32that's good all right and so now we're gonna generate Bang and so now this should make us something my prediction here is this won't be
415. 1:11:43anywhere near as good as mid-journey in terms of the image but it should have text that says Larry's hedgehogs that's
416. 1:11:51actually a good name for a lawn mowing service I'm sorry a Landscaping service yeah see the the uh just the the quality the
417. 1:12:04quality difference in the image between that which is just badass and like that I mean these are cute but these Larry's hedgehogs Larry's
418. 1:12:15hedgehogs right these have text some of them do some don't because it just it's it's the nature of this [ __ ] right now so let me
419. 1:12:24grab actually I like the layout of this one looks more logoy to me and so now I can go in here and I can remix any one
420. 1:12:32of these this is one of my um one of my uh one of the tricks with using this tool because this doesn't say Larry's
421. 1:12:39hedgehogs it says learn Jayhawks hedgehoggers so I'm gonna remix it and just see if it does any better on subsequent
422. 1:12:52I got 10 seconds left no 15 seconds left all right we'll wait for the 15 seconds there their servers are a little slammed I think
423. 1:13:03that was good brother oh good awesome Diego thank you I appreciate it all right let's see if we can get a Larry's Hedgehog that
424. 1:13:11actually has Larry's Hedgehog in it now this one you can't turn in that into that vectorized thing but you can take
425. 1:13:18any vectorizing piece of software that'll turn a bitmap into a vector and and you know logoize one of these things
426. 1:13:28oh that did something interesting it looks like it put a hog it put a potato between Larry and the S Larry's potatoes
427. 1:13:38there's hedgehogs there's hedge Hogs Larry's hedgehogs so we're getting close and that one's kind of mean and cute looking so anyway
428. 1:13:49so yes you can do logos with these now do you have the kind of control that you want if you're doing logo design no you
429. 1:13:56don't but if you want a concept up 50 different logos in an hour you can do it this way right and then like even even with a
430. 1:14:06modicum of you know illustrator skill you could take one of these things that's kind of half-assed and and make it good
431. 1:14:14so or you know hand it to someone on Fiverr and say go unfuck this up all right make me a Birmingham make me a Birmingham Kyle
432. 1:14:29oh you're a fine Lass I like this I you know what this copy is gonna go in a show and tell Channel
433. 1:14:43we're gonna go show and tell this Larry's hedge hugs Landscaping service logo
434. 1:15:03oops I guess I guess I should attach the damn picture in there shouldn't I where is it where are you oh Larry come on
435. 1:15:15copy image show and tell paste oh it's not let me do it uh um [Music]
436. 1:15:29but Kyle you're not paying attention to us I got 80d this is how it goes this is how it goes [Music] [Applause] [Music] save as
437. 1:15:45we're gonna save this we're gonna do it we're going to be so good you're gonna love this that's gonna be awesome
438. 1:15:52I'm good plus we're gonna go upload file we're gonna get Larry the Hedgehog and then we're gonna go bang look at that
439. 1:16:01Larry has now been immortalized that was worth it for me seems a little touched [Laughter]
440. 1:16:15using Dolly three props in mid-journey oh that's interesting Badger to the Bone lawn service that's pretty good too Share Bear join hey
441. 1:16:29share bear what's happening uh but up using Dolly three prompts in mid-journey I've used it to make awesome
442. 1:16:36synthwave Star Wars avatars oh yeah that's a pretty cool idea Larry's Lawn Service like it that's his brother badger
443. 1:16:44cousin Badger nice Badger Lawn Care awesome logo call it Buzz oh buzzes that's good
444. 1:16:59buzzes hedgehogs see that's something that you could use chat chibi T4 too is go take all those good names and Patch into the bone
445. 1:17:12uh it the dad jokes in here are funny time to sleep yeah I'm pretty tired I'm pretty tired Microsoft and chat GPT open AI
446. 1:17:26once wait once small nuclear reactor for electricity yeah that doesn't surprise me that doesn't surprise me at
447. 1:17:36all what's the advantages of using cross entropy loss function for doing multi-class problems here's the deal this that's the I don't
448. 1:17:45have that answer for you at all I could well what's that site let me Google that for you um the only thing I could do with that
449. 1:17:53question is let me GPT that for you so this channel is more about the after the inflection point that started on
450. 1:18:00November 30th 2022. I know very little about the the math and machine learning side of this other than to know enough
451. 1:18:08about it to understand how the [ __ ] works and why it does what it does um but when you start asking actual
452. 1:18:13technical questions about this stuff there's a guy in here that's pretty regular that that's also in the AI salon
453. 1:18:20named Pate who works for Google doing tensorflow stuff for them he can probably answer that for you I can't so sorry
454. 1:18:32what are your qualifications sir I would like to know about the weightings oh I got nothing all right
455. 1:18:39all right all right all right yeah okay okay okay um okay yeah yeah Peter yeah we're gonna need you to go on
456. 1:18:51ahead and come on in on Saturday yeah oh and uh one more thing Peter uh we're gonna need you to go on ahead and come on in on Sunday too
457. 1:19:03looks like somebody's got a case of the Mondays [Laughter] what's your age um too young and too old all at the same time I'm 58.
458. 1:19:18what's the advantage of using cross entropy loss function actually you know what it's a good enough question I would like to know
459. 1:19:26let me chat gbt that for you and we're going gpt4 with something like this all right what's the advantages of using cross
460. 1:19:37entropy loss function for doing multi-class problems cross entropy lost known as log loss as we all know I mean that's obvious
461. 1:19:49measures the performance of a classification model whose output is a probability value between zero and one
462. 1:19:54cross entropy loss increases as the predicted probability diverges from the actual label duh being more punitive for predictions that
463. 1:20:06are confidently wrong hello I think everybody knows that I don't even know why I asked it seems obvious to me
464. 1:20:14probabilistic interpolation cross entropy loss okay I can't handle this anymore okay oh wait oh now it's doing math well good
465. 1:20:22Lord oh wait look [Laughter] okay we're gonna do this I'm sorry period you've mistaken me for someone who gives a damn period
466. 1:20:39um can you please explain that to me as if I had no idea what the hell you were talking about what's see if it even answers it
467. 1:20:52absolutely let's simplify it it didn't yell at me see Bing would have shut down the conversation if I talked to Bing
468. 1:20:59like that Sydney Sydney has gotten very prudish after she stalked that New York Times Reporter she's gotten very prudish
469. 1:21:06all right imagine you're training a dog when the dog does something right you give it a treat okay that I can understand
470. 1:21:12when it does something wrong you might say no you train your dog by rewarding good behavior and correcting bad behavior
471. 1:21:19perfect in machine learning we have a similar process where we are training a model think of it as a virtual dog and
472. 1:21:25you have to tell the model and it does something well and when it isn't this is where the loss function like cross
473. 1:21:31entropy loss comes in it's like a guide for the model to learn from mistakes here's a simple rundown of why cross-entropy loss is good
474. 1:21:41is a good trainer for our model it's clear it tells the model very clearly how wrong or right it is the more wrong
475. 1:21:48it is the higher the loss the more correct it is the lower the loss it's strict if the model is very confident
476. 1:21:53but wrong it gets a big no a high penalty it's a it's a bad model it's a bad model you confidently answered even
477. 1:22:03though you're wrong bad dog I like this this is good it's fair it works well even when the classes like dog breeds
478. 1:22:11are imbalanced some breeds are rare some are more common it's encouraging when the model is less confident but correct
479. 1:22:18the penalty is low encouraging the model to be more confident next time it works well with probabilities see if this is so listen people
480. 1:22:30look at what just happened there I didn't have to go to Stanford and get a degree in mathematics to understand
481. 1:22:36what the [ __ ] I was talking about if if I if I went to Google and looked up cross entropy loss functions of the
482. 1:22:44whatever the question was oh it's right here the advantages of cross-entropy loss function for doing multi-class
483. 1:22:49problems if I went to Google I I would spend if I had the attention span I would spend probably two hours reading
484. 1:22:56white papers and trying to parse what the [ __ ] it actually means and if I didn't have a math background I would
485. 1:23:02probably get lost and discouraged and give up give up I just said I'm a dum-dum can you talk to me like
486. 1:23:10I'm a dumb dumb and it's like these are it's a really good metaphor like it's really clear it works well with
487. 1:23:17probabilities it's good at handling situations where the model has to give probabilities like saying I'm 80 sure
488. 1:23:22it's a labrador 15 it's a Golden Retriever and five percent it's a poodle it's a good match so anyway yeah super cool so the advantages are
489. 1:23:32quite clear but now but so here's here's why this channel exists I know I did that as kind of a fun self-deprecating Razz of myself like I'm
490. 1:23:49a dumb dumb and blah blah but what are the implications of this you're you're working in a company you may not be trained up on some part of
491. 1:24:00what that company does and the boss says to you hey I would like you to give me an answer on why this is better than that
492. 1:24:08and as a good worker we would all likely go sure boss no problem I'll have that for you tomorrow great and if you know nothing about it you
493. 1:24:19panic and then you try to find someone who knows the answer to it maybe you do maybe you don't and maybe you're
494. 1:24:25terrified because it's a weird culture and if you show any weakness you get your head chopped off so then you spend the night
495. 1:24:34all over Google trying to solve that problem and you can't do it as opposed to hey let me get that to you by the end of
496. 1:24:41the day you go back to your desk you sit down with something like this you actually learn something about the
497. 1:24:46business you give the boss an actual answer right this levels up workers and it I don't know I I just the implications
498. 1:25:01of like this little use case is a real microcosm of the kind of shift we're gonna see in every piece of
499. 1:25:12knowledge work every piece of creative work knowledge domains where you have no right to be commenting in that knowledge
500. 1:25:23domain all of a sudden you've got enough knowledge and that knowledge don't domain to just to apply critical thinking to it
501. 1:25:30because now that I understand what this is I can talk about it at a minimum and if I spend another three days on
502. 1:25:39this with this tool I could probably be understanding the math and finding libraries that do this and testing it
503. 1:25:47and actually understanding what's happening in three days imagine if I spent three months digging deep on a topic like this
504. 1:25:55using these tools [ __ ] crazy it's [ __ ] crazy Sydney is very Persnickety Sydney is a she's persnickety
505. 1:26:12I'm gonna I'm gonna bite my tongue on that one uh number three has fangs exactly the fangs were the best
506. 1:26:20let's go look at that that's good oh wait I'm here like uh [Music] oh you know what I'm gonna do no Kyle what
507. 1:26:33I'm gonna talk to myself a bit more I gotta be driving people crazy tonight he's just off his rocker isn't he wait
508. 1:26:43why didn't that go over there the hell Apple why can I not drag that into my that photo library that doesn't make any sense to me um
509. 1:26:59[Music] let's see export next part [Music] is it jpeg desktop
510. 1:27:16um huh I got some I got some memory issues I think Microsoft Edge just hosed my machine again oh Microsoft could you fix your [ __ ]
511. 1:27:32software you know you're about to make a hundred trillion dollars when you launch co-pilot into your office 365. maybe you
512. 1:27:43can fix your damn software I don't know is it just me I'm not a robot goodbye that's what Bing said to me the other night I got yelled
513. 1:27:54at by Bing she hung up on me I told her she made a shitty image she goes she goes well I I'm sorry if I
514. 1:28:04disappointed you I'm not a robot I'm a large language model from Microsoft and um I hope you have a good evening
515. 1:28:11goodbye that's why I use person snickety Kyle oh Tick Tock safe wording yeah exactly oh man all right let me do a couple more
516. 1:28:23questions I'm I'm a crispy Critter as you can all tell this is not normal behavior well it is this is pretty this
517. 1:28:31pretty much how it goes here um I'm so infatuated with AI yes I wish I could pinpoint the best ways to use
518. 1:28:37them more effectively so listen baby baby Izzy baby Izzy I like it I'm down with the name I'm gonna go nine out of 10 on the name [Applause]
519. 1:28:50nobody right now really knows anything about these things the generative AI stuff the people that know the math part of this the engineers
520. 1:28:59they know a bunch of [ __ ] but kind of post chat gbt like how do we use these tools in our work how do we use these tools to
521. 1:29:06increase our productivity and increase our creativity and increase what we can do and expand and that did all that sort
522. 1:29:13of stuff nobody right now nobody knows what they're doing the whole reason I started the AI Salon was to create a community of people that
523. 1:29:24were all trying to find other people that knew more than them and what you learn in there is that yeah there's
524. 1:29:30people that might know some more about this thing or that thing but then there's other people that know nothing
525. 1:29:35about that and like you know something that you think's really obvious but they think is revelatory that's the stage of the industry that
526. 1:29:43we're in it's so new that nobody's an expert so the the the the single most important um skill right now is be curious
527. 1:29:57just be curious and explore explore explore just play with this [ __ ] just like these three go ask the same prompt to all three of those
528. 1:30:12and all four of those Bing labs.perplexity pie po Bard Claude and chat GPT why I don't know see which one's better at something
529. 1:30:24I did it one night where I asked it I gave it a I gave all seven of these a joke premise and I said write a joke for that premise and
530. 1:30:35um one out of seven wrote an actual joke that was gpt4 um two out of seven were so bad they would it was not it two
531. 1:30:47two and seven didn't even have joke structure and then like the middle ones had joke structure it was crazy
532. 1:30:53but we learned a lot right so just play just play like do stupid [ __ ] do fun [ __ ] do amazing [ __ ] volunteer to solve
533. 1:31:03someone else's problem anything that you can do in any domain right now to get your head around how these things work
534. 1:31:09are going to serve you moving forward because in two weeks within two weeks it's going to be something completely different
535. 1:31:19you'll be able to talk to it you'll be able to send it pictures you'll be able to generate pictures with it all within
536. 1:31:24the same conversation what does that mean I don't [ __ ] know what can you do with that I don't know am I gonna like throw all the pasta at
537. 1:31:34all the wall yeah all the walls all the walls yes at some point I'll figure out oh you can do this cool thing with it and someone
538. 1:31:43else will go be like oh my God that's amazing and then they'll show me the thing that they thought was just kind of
539. 1:31:48cool and I'll be like oh my God that's amazing that's the stage we're at right now so um any ideas about Chooch
540. 1:32:01I don't know Chooch I assume Chooch sounds like one of those product names it's like hey I got an idea we can build
541. 1:32:08something on top of chat GPT we need a domain what's available I like Cheech and Chong okay let's see if we can get
542. 1:32:13cheats no that's taken Chong's taken how about uh Hooch I like Turner and Hooch how about who's no oh how about Chooch
543. 1:32:21uh chooch.ai is available all right let's get that all right so my guess is it's going to be a wrapper app
544. 1:32:30with a with a uh a domain that was available I would like to be proven wrong oh I am proven wrong the chuchai vision
545. 1:32:43platform is used today in a white so I don't know it let me see let me see what it is this looks actually pretty cool oh
546. 1:32:49this looks this looks like it's been around for a while detect understand and act instantly okay this is object object detection
547. 1:33:00so here's the thing about this AI stuff things like this like Chooch and AI detection like this stuff's been around
548. 1:33:06for a decade or more right and and two Chooch when was this when were they founded about Chooch [Applause] [Music]
549. 1:33:212023 Chooch they definitely huge partners yeah they're partnering with big big technology companies they've been around for a while
550. 1:33:35um this kind of stuff's cool um I'm gonna use this stuff I'm gonna use this stuff in a project that I'm pitching right now for a big
551. 1:33:43Pharma company to do some video analysis stuff so um I like this stuff it you know there's very specific uses for it and it's getting
552. 1:33:54um like like Microsoft just announced or just released a an object recognition thing that that was just crazy and how
553. 1:34:03instantly and detailed it did object recognition so this stuff's gonna get it's it's already crazy but when you combine Tech
554. 1:34:12like this with the the generative AI stuff where it makes it really accessible to lots of people you're
555. 1:34:19going to be able to create crazy crazy tools so anyway all right all right Sydney rapper nickity I like taters um
556. 1:34:33what's the use case I don't know got distracted got distracted welcome to chat add [Laughter] oh man all right listen listen people
557. 1:34:47I'm gonna do one more question and I'm gonna get up on out of here I'm gonna get up on out of here all right great
558. 1:34:53job you've been live for 90 minutes oh I know my ass is numb you you don't think I know I've been doing this for 90 minutes
559. 1:35:00nobody's an expert and everyone's an expert it's so cool yeah exactly this is a magical time here's here's the thing okay
560. 1:35:07all this stuff no one's an expert Everyone's an expert exactly you could you can you can get up to speed in a week and sort of figure out
561. 1:35:17how to do something no one else has figured out how to do how you take what you're good at and start playing with
562. 1:35:22these tools and just just go start playing with it um this is a very very very rare time in history right so these moments when
563. 1:35:39major pieces of technology that are going to change everything they don't come along that often think printing press think steam engine
564. 1:35:51tractor computer internet World Wide Web AI right they don't come along that often and it's even more rare
565. 1:36:06to be actively involved when in the primordial soup of the beginning like we're gonna look back five years from now
566. 1:36:17we're going to look back to November 30th 2022 and go that was the moment when all of this [ __ ] had started
567. 1:36:25just [ __ ] around here on Tick Tock talking about this stuff and showing how to upload [ __ ] and making images and all that
568. 1:36:33this is the good stuff you don't get to be part of the primordial soup of revolutionary Technologies you know
569. 1:36:44very often in a lifetime like I've gotten two right I was there at the early days of the world wide web and I'm
570. 1:36:51here for this and I'm here for this um I missed the PC the the personal computer I missed that by about a decade
571. 1:37:00and I swore if I ever got one of these in my life I would [ __ ] pounce on it and I did in 1994 with the web and now
572. 1:37:08I'm pouncing again because this is this is some [ __ ] so anyway I love that I love you all I love that
573. 1:37:16you all hang out in here so let me see Silver Fox Ted looks at life Clint y'all are awesome share bear amazing Brian and major gift funding
574. 1:37:26thank you Joker so cool to see you madcap thanks for coming all right everybody uh I'm Crispy the eyes are bloodshot I'm tired on Fridays
575. 1:37:41where Hawaiian shirts okay Gen X rules Gen X rules all right Sherry Banks Daydream Fisher love the name Irregulars presents Turtles
576. 1:37:58all right everyone peace out have a good one