Feb 29, 2024 -Creative Insights for Everyday Innovators

Transcript

1. 0:01let's do [Music]
2. 0:18this in a Westerly Direction car is my train driving I've been
3. 0:33wondering what it is I'm running from [Music] again feel like an 80y old man holding on to
4. 0:4729 and a on that horiz is a California line [Music]
5. 1:03couple H trucks carrying a wide L brief five house C in half cute little front door and two windows my Lord ain't sure that a cry should
6. 1:20laugh you see I broke a home up myself once as I Stumble to that door I read a note by the dawn's light said don't you come around
7. 1:39here anymore well [Music] I good
8. 1:55with hey there's no rapping there's no rapping we're singing Americana you can't just burst into rap it's just it's
9. 2:04not the genre I like good rap just like the next dog you know your self-awareness lately I don't know if you've got dogs
10. 2:16that are not self-aware but good Lord am I am I right we're in the middle of recording a song he just breaks into [Music]
11. 2:26rrap what what you trying to do a mashup is this like Fusion [Music]
12. 2:52stand between you and a hard plac is insane standing too near you and a fire makes it
13. 3:09clear your trouble to me real trouble can't you say hello [Music]
14. 3:23hello good day good people of the internet of the tick tock [Music]
15. 3:38[Music] Champ's trying to do crossover he might
16. 3:54be I got to talk to his manager listen if you want out of your contract you buy your way out of the contract
17. 4:05it's $3 million like every other dog in this house all right it's pretty straightforward oh what you can't work there's a [Music]
18. 4:22surprise oh now you're reluctantly singing huh all right the contracts you forgot you sign that didn't [Music]
19. 4:52you he knows he knows where his chicken sticks are bought [Music]
20. 5:10all righty everybody happy Thursday night is it Thursday night it [Music]
21. 5:28is [Music]
22. 5:44[Laughter] all right he's done he's done he's had it he's had it what are you going to do what is on my tongue all
23. 5:58right welcome everybody name is Kyle Shannon this the AI learning lab what are we going to do tonight we're going to talk
24. 6:06about stuff talk about stuff lovely image here uh someone got a behind the scenes picture of the AI learning lab
25. 6:13not a lot of people know this but this looks like a shitty home home office $20 million Production Studio you
26. 6:21can't see the crew the managers the handlers around me we try to keep it that we try to keep it so it appears authentic like these
27. 6:36t-shirts $350 a piece It's amazing And it like a new one every night these they don't get washed they get thrown out and
28. 6:45you're like but wait I've seen that same shirt with the same spots on the brand new every time amazing whole
29. 6:53crew puts the spots in it's just if you knew if you knew I start getting ready for these around 3:30 in the afternoon I
30. 7:03leave work come to the studio about 3:30 in the afternoon going to hair makeup I mean clearly a lot of makeup you know
31. 7:13you think I do this no it's professionals talk about how well AI can make
32. 7:28predictions well you mean you mean like predicting Super Bowl winners and [ __ ] like that or do you mean like how well it does as a
33. 7:39prediction engine for tokens cuz I would say as a prediction engine for tokens it's pretty [ __ ] remarkable if you're talking about it
34. 7:50having some sort of soothsaying power beyond that I'd say it's pretty shitty you got to ask what's the name of the hair gel you
35. 8:03use it's just uh what's it called I don't know it's just cheap ass hair gel it's just cheap ass hair [Laughter]
36. 8:18gel free and clear no curly I don't I don't [ __ ] know it's in a green it's a green it's in a green it's like what you get at Walmart
37. 8:28I don't know it's like the [ __ ] you get it might not be obvious by looking at me but I'm not exactly spending the bulk of my paycheck on
38. 8:46product it's blantons if Blanton made a hair gel I'd be wearing it that's the one thing I would probably spend some money
39. 8:58on I never quite understood the spend $45 on on you know a thing of shampoo or conditioner and that was when I was you
40. 9:12know when the [ __ ] wasn't expensive now it's now it's almost close to that the cheap shit's almost that expensive it's
41. 9:19cuckoo all right got to be like Gorilla Glue man well I just this is just it's still wet I just wet it CU I took a nap
42. 9:28my hair was a little puffy it's a little puffy um okay so what do we do here other than talk about my hair product
43. 9:37and massive uh Production Studio um we talk about AI we talk about generative AI we talk about how it's going to impact us and
44. 9:52what's the cool [ __ ] and what's coming and what's not here and what's never going to be here what's going to be here
45. 9:57way sooner than you think predictions based on current and recent past events I would say that I I would say that large language
46. 10:12models the the one thing that I am decently impressed by with them is their ability to analyze other documents
47. 10:22there's you know there's kind of you know two ways you can use these things one way is it generates [ __ ] right you
48. 10:29give it a prompt that generates [ __ ] and the other way is give it a document and have it analyze that so it's pretty good at that
49. 10:37so um you know you could give it a bunch of a bunch of data from the past and have it look for patterns in that use Advanced Data
50. 10:49analysis within chat gp4 have it find Trends in the data and then have it make projections on the data moving forward it's probably not a
51. 10:58bad idea actually probably not a bad idea like you know is it going to be awesome and perfect no it's going to be janky and weird like
52. 11:07all of these large language models right now are janky and weird but I'll tell you what for for the person asking about
53. 11:13you know how how good it is at at prediction and and uh and stuff like that oh you're right here vocal sync
54. 11:23um I would say first of all get chat gp4 if you don't have it learn what code interpreter is upload some historical
55. 11:34data to it have it do some Trend analysis on it and then and then just learn how to use it learn how to use that [ __ ] with
56. 11:42it and then as it gets better you'll be ready for it and so so it it might get way better I mean the other thing you can do with it is
57. 11:54I I I have a sneaking suspicion you might be asking for things like you know know stocks and crypto and [ __ ] like that
58. 12:03um I don't I don't know how good it is at that [ __ ] but one thing you can do is you can go grab if you like a couple of
59. 12:11different trading Bots out there go grab the code from the trading Bots and dump it into chat gbt and tell it to make you
60. 12:18a hybrid trading bot and update it based on your preferences or update it based on some historical [ __ ] and see what it
61. 12:24comes up with I would just again the right now the the all of our jobs all of our jobs with these tools are play play
62. 12:33play play play play play play don't take [ __ ] too seriously just explore explore and play and explore and
63. 12:42play the more I'm using these tools the more this feels right to me you should be you can build [ __ ] you
64. 12:50can try to solve problems you might even solve problems you might even build a product people will pay you money for
65. 12:57you can make money with J time and assume that that product you built will be useless 3 months from now and there will be no demand for it
66. 13:10huh why because the thing that you figured out that was an angle that was like ooh there's it can't do this one
67. 13:19thing right so I figure if I combine this thing and that thing I could put those together and make a product and
68. 13:24that's unique in the marketplace and then 3 months from now someone will come out with some product that basically just does
69. 13:31that I just assume that's reality and if it's not if you have like a yearlong run with that product or longer more power to
70. 13:47you and if not then you heard it here first just keep building [ __ ] keep building [ __ ] keep playing
71. 13:59what if playing with it comes up with useful results 99% of the time winner winner winner chicken dinner you make
72. 14:07money my then I would say that I would my my uh my rule of thumb right now with large language models and and this [ __ ]
73. 14:16is if you can get these things to give you 80% of what you want consistently you're doing pretty good right now it's
74. 14:25still on us to get from 80 to 100 so if you can get some something that can consistently get to 99% of what you want or 99% success
75. 14:37uh good on ye mate put together a course and teach other people how to do that [ __ ] and then you can be one of these
76. 14:46guys you can make money with I figured out the formula you take this prompt plus this thing plus my secret formula that I'll
77. 14:59teach you for $49.95 a day for the next 27 days you can do that if you can do that [ __ ] far out man orange juice no
78. 15:13coffee it's just a little little little thin maybe little too much leche little bit too much cone leche Robert thank you for
79. 15:27the I work with confidential data docs what are my AI options right now um You can there's a there's a company out of Denver actually called
80. 15:39Lial which is really interesting you set up an account with them and their technology will de identify your docs
81. 15:47and then it will you do all your large Lang you do your all all of your third party querying with deidentified docks and
82. 15:57then it gives you answers and it re identifies the answers which is really [ __ ] cool so that's a way and then the other
83. 16:04way is um uh you know host your own large language models do it locally or do it you know get yourself a a pile of gpus in the
84. 16:17cloud um and use the open source stuff I I would do that right now I wouldn't I wouldn't trust any of the
85. 16:26third parties right now with any of the data the every they're moving everyone's moving too fast and loose with the
86. 16:32technology right now just I assume that if you're making large language model calls to like open AI or co-pilot or Google or perplexity or
87. 16:47Claude or any of them assume that it's just you're you know it's not secure just assume it's not secure it's the safest
88. 16:57way to do it but the the deidentification thing I think is really interesting did you read the report
89. 17:02about Clara and the use of AI agents no Clara Clara or Clara did you all read the report about Clara and the use of AI agents I did
90. 17:18[Music] not Clara agents AI fintech chatbot does the work oh yeah I
91. 17:34did read about this this was the customer service thing right saying that they're yeah yeah yeah yeah yeah they
92. 17:50they this was I did read about this so is it CL Clara Clara okay k l a r n a is the company and they they used customer service
93. 18:07Bots and it showcased impressive capabilities effectively performing the tasks of 700 full-time customer service agents with
94. 18:22similar um customer satisfaction results that that was the thing that was that was most impressive to me two two years
95. 18:30ago I was building models to predict oh crap I just lost that what did that say side hustle Mimi to predict what
96. 18:37supposedly took 700 jobs away from people I don't I don't that's probably not quite a right characterization um
97. 18:45repeat requests wait I was building models to predict repeat arrests and now chat GPT can do it in seconds oh that's wild side H hustle me
98. 18:56um just because the Bots can effectively perform the tasks of 700 full-time customer service reps doesn't mean it took the jobs of
99. 19:07700 fulltime customer service reps that said if there is a if there was a service category to be disrupted that needs to be flushed down
100. 19:21the toilet it's customer service because the customer service reps have no power they're basically just sitting sitting
101. 19:28at the base of these phone phone logic trees that they can only answer a certain set of questions they can't
102. 19:35actually answer questions they can't actually take actions so if if they can create Bots that can talk to you can
103. 19:45actually answer questions can actually take actions on your behalf and solve problems and then only escalate your problem to a real
104. 19:54person if it can't be solved in like all the official ways and then these people are actually trained up that would be a 10 time no
105. 20:06thousand time Improvement on the current experience of dealing with any customer service I don't care if it's onshore
106. 20:14offshore wherever it is they're all horrible right now because they all tried to make them as efficient as
107. 20:20possible it it's all these just nasty phone trees you can't ever talk to a person it takes you 45 minutes to get to
108. 20:27them when you get to them they no power they're as frustrated as you are about the third transfer is is when
109. 20:36you talk to someone who can maybe do something for you and at that point you've been kicked off the call twice
110. 20:43right you're now two hours into this experience and you're ready to just you know dive through a window if Bots can solve that
111. 20:57awesome I don't know what to [Music]
112. 21:15say let's see live agent yeah exactly that's that's the problem is they've turned because of the processes and
113. 21:26because of the current state of the technology they've turned real human beings into Bots that can only read the
114. 21:34script that they've been given right so so that's a problem and it's because most of the customers are asking the same stupid
115. 21:48questions all the time right so they have to have these tears of people well let the bot answer all the
116. 21:58stupid questions and then just give me to a person like if I if they do it right within 2 minutes a decent AI
117. 22:07chatbot should be able to go this is beyond the scope of what I'm authorized to do let me give you two a human being
118. 22:15that is trained up to actually deal with you so we'll see we'll see if it gets any better but I mean that's that's
119. 22:24that's promising I mean I get that it's scary I get that it's going to be disruptive I don't have a good answer for that not
120. 22:34in that particular case because I I think there's some there's some sectors that are just so broken right now they're so broken customer
121. 22:44service is just it's just useless oh man and you have to listen to a 15-minute script oh yeah that's true
122. 22:52it finally picks up and then it's like this Call's being recorded for quality assurance purposes if you haven't seen the November flyer
123. 23:01for sales go down to the local Piggly Wiggly and you're like can I please talk to someone and we sense a we sense a little
124. 23:11bit of aggression in your voice so we've routed you to the to the um customer agent Care Facility Who who will give
125. 23:20you a little talking to before you can get your problem solved here's an idea fix the problem of every person calling
126. 23:28calling about over and over well yeah that's that's the that's the other thing but um yeah what they realized is that
127. 23:37is that by just running people through these Byzantine you know answering systems they don't actually have to deal with
128. 23:47the customers and that's a lot more profitable so with a system like this they can probably spend you know the
129. 23:55same amount or less and actually get customer you know customer satisfaction and solve customer problems education is
130. 24:04broken it sure is Mr Studios if you're not in the AI Salon you should join the AI salon and join the education Guild for that very
131. 24:14reason get in there with those people making a difference but not very good for higher levels of skilled coding what's not very
132. 24:27good large language models bgh 1684 are you using it right now to do pair programming or are you just assuming
133. 24:38it's not good for higher levels of skilled coding I think you're right right now but I wouldn't assume that what your
134. 24:46statement is right there will be true for very long I would say by the end of 2024 it's probably going to
135. 24:52get it's probably going to get scary good see here's the deal when right now the reason you use it okay cool
136. 25:02um the reason these things fail at higher level things is it's just too janky and it hallucinates too much and
137. 25:10it's not doing any self-analysis right it's just it's a predictive engine so it just it just vomits out
138. 25:17[ __ ] and then it either gets it right or it's not and it requires highle level coders to be able to look at the code
139. 25:24and say oh is this actually an efficient way to do this or not hey you know redo this CU what you did is not efficient or
140. 25:31it doesn't work or it broke whatever it is but by sometime in this year we're going to have the next level
141. 25:41of these tools that have reasoning the minute these things have reasoning the game changes because it'll write the code and
142. 25:52then it'll look at the code it wrote and do its own analysis and go oh that's not the most efficient way I could do that
143. 25:58hang on there that's that's more efficient oh and by the way based on your coding preferences I know you like to
144. 26:06code you know with this particular coding philosophy or this kind of coding framework so let me do it that
145. 26:13way and then let me do it and let me run my own QA on it and then let me run so so I'll write the code I'll run my own
146. 26:23QA on it debug it before this is all before it hands it to you right it'll write the code run analysis on it debug it throw an
147. 26:32interface on it and then watch a virtual interaction with the interface and do its own ux analysis and then by the time it
148. 26:43delivers it to you 38 seconds later it'll be good that's probably not this year but I mean that's that's where it's
149. 26:49headed so so right now yeah the tools like like I said any large language model right now it'll get you to around
150. 26:5780% of what you need and then as reasoning come comes in and Agents come in [ __ ] change shit's going to change
151. 27:06fast so just be AI literate be using it be using it and then you'll know you'll know when it gets better because it'll just not
152. 27:16suck uh I'll keep my call center opinion to myself I was working on a Model idea to handle inbound calls without a call tree
153. 27:28that would be huge Rachel Sparkles do you think reasoning capability already exists but they're either holding off to
154. 27:36implement it I think so I I think I I saw a thing today that they're calling it um it was like open ai5 but the O was a
155. 27:54q so they've so open ai's got this thing called qar which is from what I can understand it's a reasoning engine and it the
156. 28:08way I learned what it was from David Shapiro um it's kind of this reasoning engine that's independent of the of the
157. 28:16prediction engine that the large language model is so you have the prediction engine and then you have this
158. 28:20reasoning engine and together they do [ __ ] um apparently that [ __ ] got really good and and and rumor is is that's why they fired Sam
159. 28:34mman so the official statement from the board was qar had nothing to do with Sam Alman being fired which there's your immediate
160. 28:44clue that they said that out loud if it if it didn't have anything to do with him just say nothing I'm a software tester I'm
161. 28:58learning new skills now yeah shop shop Panda it's it's a good call just you know learn what you can do right now
162. 29:05with AI to do testing or learn completely new skills right that's the thing about these AI tools is they allow
163. 29:12all of us to level up all of the skills that may not be at the core of what we do and what we're likely going to
164. 29:21discover is there's other things that we can do our talents the way our brains work are going to be good for a lot more
165. 29:27than just the thing we think they're good for right now and that's going to be amazing hey Pate what's happening I so need an [Laughter]
166. 29:43agent how is a pastor using AI that's interesting I would think right now a pastor could use it in all sorts of ways who this I'm software
167. 29:56tester are are they maybe working out the bugs and making sure it will work right um I think that probably what
168. 30:04they're doing if they've got if they've got a reasoning engine Rachel Sparkles that works um this is much less about
169. 30:12figuring out if it works right and much more about making sure that they've got what they're what open AI is calling
170. 30:20super alignment so super alignment alignment is as these tools get more
171. 30:34capable are there are there imperatives are there rules within the software aligned with human beings
172. 30:49imperatives cuz if they're not that's when you get into weird scenarios where the tools are like well you know
173. 30:58these humans are kind of getting in the way here why don't I just go ahead and take out the the uh Logistics
174. 31:07grid so that the food supply stops that'll slow him up for a while and I can keep doing my thing right that's the the Sci-Fi scenario
175. 31:16is we make the machine that's got imperatives to succeed and then it does it at the expense of humans so a big part of what Sam
176. 31:30wman Maybe back in October they they announced there's a blog post on open AI about super alignment they said they're
177. 31:37spending 20% of their 20% of their expenditure is going towards super alignment and then um anthropic the
178. 31:46company behind Claude is a bunch of exop AI people that they didn't agree with the way that open AI was doing alignment
179. 31:53so they started a whole company to do alignment a different way they're using something called the Constitutional
180. 31:59learning model um they've got a completely different approach to alignment but all of these companies
181. 32:05right now if they get to AGI if they get to tools that are basically can do the jobs of most humans mostly
182. 32:18well um then they got to make sure that they're aligned so it's probably safety is the thing that they're working on more than does it
183. 32:26work and then also where do you aim it what do you do with it but it's crazy it's it's I don't think we're going to have
184. 32:36to wait until the end of the year to figure out what it actually means to have one of these systems that has reasoning and logic in
185. 32:45it Sam wman he just answered he he just answered this question this last week or I saw it last week I think it was last
186. 32:53week because I hadn't seen it before where he was he was asked about tell me about chat gbt 5 and you could tell Alman was a little
187. 33:03exasperated and he was like he was like well he goes so it's going to get better at all the [ __ ] right it's going to get better
188. 33:11at video and it's going to get better at large language model [ __ ] it's going to get better at coding it's it's going to
189. 33:18get better at all the [ __ ] but he goes he goes mostly what it's going to do is it's just going to get
190. 33:26smarter and like that was his answer the answer of what is chat gp5 is it's just what if something like chat gp4 just got
191. 33:38smarter well then it would start doing things like not hallucinating as much being able to do math problems being
192. 33:47able to code at a high level right like like just it getting smarter means it's not quite so janky and and at what point
193. 33:58like so you got your janky meter right so right now let's say say this is completely incapable and this is better
194. 34:07than every human right right now we're we're kind of here like it's mostly it's mostly Jank maybe maybe we're over here
195. 34:14right like it's pretty good it's pretty good but there's like a lot of Jank there's a lot of improvement to
196. 34:22go and the thing that's weird right now especially for people that are sitting on the sidelines of AI is they're going
197. 34:30oh well these tools aren't great look at this Gap this gap's huge well if they just get smarter it just it's that needle just
198. 34:41sort of pins and there's a point at which I feel like there's a Tipping Point where it's good enough that it's
199. 34:49better than most people and it's probably not all the way at it's [ __ ] brilliant right it's probably somewhere
200. 34:55between where we are and where it's going to be somewhere in there is holy [ __ ] it can do highle coding now holy [ __ ] it can write an
201. 35:07article fact check it grammar check it match it to your tone and present it to you for just approve and publish or
202. 35:16would you like to go in and fix anything and then you read it and you're like [ __ ] that's good publish so we'll see we will see but
203. 35:27we're not there yet ethical issues yes absolutely I can't access Claude because I live in Greece oh that sucks claude's pretty good uh if
204. 35:34you haven't tried Angelique if you haven't tried perplexity you should try that too it's pretty good hello happy people good day to you
205. 35:46good day good day good day Pure Heart wow what a big prediction it will get better give that man all the money but Pate he only wants $7
206. 36:02trillion it's not like he's asking for1 trillion I think here's the thing pay I think you're absolutely right I think you're on to something
207. 36:13here if you want to ask for a million dollars you got to have all the answers if you want to ask for a billion
208. 36:20dollars you got to have a vision if you want to ask for7 trillion doll you just say oh the next version's going to be slightly
209. 36:34better I think it's it's called the inverse proportion law to the hubris of how of how much you're asking
210. 36:45for listen these these computers are pretty smart but if you give me 100 billi trillion dollar five 5 to 7 trillion if you give me7 trillion
211. 36:59they'll get so much [Laughter] smarter the current code the chat GPT recommended to me was awful I'll wait for this
212. 37:10Evolution the current code that it recommended to you was awful what do you mean recommended to you if you're using
213. 37:17chat GPT like you use Google like give me some code and it gives you code and then you judge cat GPT based on that interaction you're
214. 37:28you can get much better results than if you're doing it that way because it's not just recommending code to you like to use it right you
215. 37:38have some interaction with it you give it enough context have it analyze its own code be in a conversation with it and are you
216. 37:47using chat GPT 3.5 or chat GPT 4 any experience with chat with GitHub co-pilot not personally um but a lot of
217. 37:57of developers I know use it and they I like I've heard anywhere from like the good developers I know say it's like a
218. 38:0520 to 30% Improvement in their productivity the shitty programmers that I know the lazy ones they say it's like
219. 38:12a 60 to 70% in their productivity so if you have a kickass coder they're still going to get improvements but if you
220. 38:21have a mediocre coder they're going to get better improvements but yeah but it's apparently quite good mod I got
221. 38:29this just it's 7 trillion yes Source Camp it's seven it's 5 to 7 trillion so to be clear Sam's not asking for7
222. 38:39trillion do to to make chat GPT like it's not for the software company or even the the the compute to buy from
223. 38:47Microsoft it's to build a chip company that dwarfs Nvidia so Nvidia right now their market cap is is like more than a number of
224. 39:02countries gdps put together because the company was asleep at the wheel and they basically seeded GPU manufacturing to a single
225. 39:13company and when this AI [ __ ] took off they were the only company making chips that you know and had the software that
226. 39:20that the community was using so what Sam alman's saying is if if we want to so so here's the deal apparently this is this
227. 39:29is from what I understand from the geeky folks I I I have talked to to do the quality of video that the the Sora
228. 39:42stuff the you know the Sora video that everyone's losing their minds over it's apparently 16 A1 100s of GPU power and
229. 39:53it's and it's like to render a minute of video is an hour hour of rendering time on 16 A1 100s so I think it's like four
230. 40:03or five bucks a minute of video which if you're used to Runway ml you're like I thought that was expensive um I don't know if it's going
231. 40:12to be that I don't like that could be [ __ ] but I think it's pretty pricey so I think Sam's 7 trillion is why don't we make
232. 40:22some specialized chips that are specialized for the way we do things and then let's increase production of those
233. 40:30beyond the world capacity right now it's the Cuda ecosystem that really sets them apart I don't know if you're serious
234. 40:38about that or if um if that's a joke you need to build context for code recommendations yeah
235. 40:53exactly exactly s hustle Mimi you can't just go listen if you're using chat GPT like Google oh it's expensive to do
236. 41:01compute oh oh I see yeah yeah yeah yeah yeah exactly it's expensive to do comput so so if you look at if you look at
237. 41:09actually if you join the AI salon and go in the mechanics Guild and talk to Pate about tpus versus gpus and then if you
238. 41:18go play with if you go to gro.com gq.com right now gq.com not the same grock as Elon Musk chatty MC Chat
239. 41:29thing it's it's a tool that looks like chat GPT it's not a software demo it's a hardware demo so if you go to gro.com
240. 41:39right now and just do a just do a chat GPT like thing you'll see how fast it is it's like 500 500 plus tokens per second
241. 41:48so it's like 350 375 words a second it's doing the the response um that's a hardware demo they're demoing
242. 42:01their crazy efficient chips that are all they do is large language models they can't do diffusion models they can't do
243. 42:11they can't make your game faster what they can do is they can make large language models super fast so if Sam Alman wants to take that $7
244. 42:22trillion and design a whole Suite of chips all dedicated to the specific kinds of compute that that they need as
245. 42:30efficiently as possible then all of a sudden compute costs drop dramatically and everybody benefits it also makes
246. 42:38them you know they they they become like that becomes a competitive insurmountable there that was weird
247. 42:51huh imagine Sam practicing in the mirror saying with a straight Fe face I need $7 trillion please I I thought it was I
248. 43:00thought it was [ __ ] Insanity that he was saying he needed a 100 billion that was just the rumors that I
249. 43:07mean that was just the reporting was that you know when he got fired he was over asking for a hundred billion dollar
250. 43:14after Microsoft gave him 10 yeah it's it's crazy it's crazy amount of scale but we'll see we will see I'm serious Nvidia didn't make
251. 43:28objectively better hardware for a while but their libraries are better yeah that's that's what I've heard PA is that
252. 43:35the the developers I know and um George Hots if you know him was talking about that like he George Hots tried to do some AI stuff with AMD
253. 43:48chips and couldn't because the software sucked the software libraries were bad so yeah it's not that stuff's better
254. 43:58it's that their their software worked uh it's nothing more than a probabilistic model it doesn't think yet 42o
255. 44:10okay that that dismissive it's nothing more than a probabilistic model keep keep using it the the more you use these
256. 44:19tools yes they are probab probabalistic model it doesn't mean they're just paring back the [ __ ] they've seen it
257. 44:27doesn't mean They're copying and pasting and the minute you add reasoning on top of the probabilistic
258. 44:35model the dismissiveness that is in that phrase it's just a probabilistic model yeah exactly I understand but how do you define thinking
259. 44:50exactly who was it today someone today was just talking about some big scientist today was just thought was just talking about
260. 45:04this that he said that the way the probabilistic model works is essentially how our brains work he says by definition It's
261. 45:16thinking the only thing it's not doing right now is reasoning the minute it does reasoning go look at the David
262. 45:23Shapiro video on qar go look up qar and look at the the diagrams on how that [ __ ] works it's really cool the the I don't
263. 45:34know if I have a visual in my head I I could probably go find the graphic but the way the way qar works is so you got a large language model
264. 45:46right and it's it it's got a probability and it let's say it can go in one of four directions right so you put in a
265. 45:54prompt and it's got a probability that it can go here here here here so what how qar works is it says okay I'm my
266. 46:01goal my goal is over here and I'm here and I'm going to take one of these four actions Boop and it takes the action and
267. 46:10then it measures what was the quality is the quality of the action that it took in the direction of that thing or
268. 46:17farther away from it and if it's in the direction of it it gets a higher score and if it's in the other direction it
269. 46:24gets a lower score so the way I think about it is is very much like card counting if you've ever learned about
270. 46:31Blackjack card counting there's a system where if it's you know tens are face cards it's plus one and if it's uh
271. 46:40whatever seven or lower it's minus one or something like that and all you're doing when you're card counting is
272. 46:45you're keeping track are you positive or are you negative and is it you know above the Delta right is it like + 5 +
273. 46:5310 - 5 - 10 and so that's how Q is working so it's it's trying to move forward toward a goal and it gets you
274. 47:02know better better better that score is going up it's in the right direction then all of a sudden it hits a wall and
275. 47:07now it has to like go around that wall and so it's just it's just so that's what it's doing none of the current [ __ ]
276. 47:13none of the current llms are doing that at all the current llms are just going here's your answer Splat take it or leave
277. 47:21it so the minute it can figure out if it if what it did was good or what it's doing is in the right direction the game
278. 47:30changes cuz then it can start doing [ __ ] like math first you need to make a system to understand the idea of a sentence Oh Boy mm okay $7
279. 47:46trillion oh that's hilarious is that true shop pan I wouldn't I wouldn't have met my wife without a that's awesome that's good Brandon
280. 47:59joker7 trillion will buy wait two is it billion yeah uh yeah 233 billion 333 milon 333,000 333 h100 gpus yeah a trillion dollars is
281. 48:14a staggering amount of money7 trillion is like what are they going to do buy a [ __ ] country and just take it pave it
282. 48:25and put the the chip plant on it we're going to buy [Laughter] England we'll use all that uh the tidle power we'll buy Amsterdam and use all
283. 48:41their their title technology to power our our chip manufacturing facility I don't know what they're going to do it's crazy it's crazy crazy
284. 48:55caaaa [Music]
285. 49:13all right what do you want to do what do you want to do here we can we can go look at [ __ ] are you in New York City
286. 49:18did you make an AI assistant app like 10 years ago um I was in New York City for lots of years I was a co-founder of agency
287. 49:27back in the mid90s I left I sold agency.com in 2002 moved out of the city in like 2003 maybe and then left for Left I moved
288. 49:44north of the city and then moved out to Denver or to I moved to Boulder first moved to Boulder in 2011 and I've been out here ever since
289. 49:56I'm in Denver right now um I don't think I made an AI assistant I started so I've been doing my current company's called storyvine
290. 50:06which is an automated um video storytelling platform we've been around for 11 and a half years so I don't think so I don't think
291. 50:19so it's not a bad idea Brandon I might do that if Chat gpt's math is right it it
292. 50:34analyzed well if if it depends on the math if Chat gpt's math if chat GPT is using Code interpreter and it wrote Good
293. 50:42python code then it can do math but sometimes code interpreter what what I found with code interpreter is you'll upload some data to it it
294. 50:53loads um it loads the data from the your CSV file into a data frame in in the virtual in the in the uh
295. 51:03container and then it writes python code to analyze the code that's in the data frame but what I've noticed is that
296. 51:13sometimes it like puts other [ __ ] in the data frame that it just makes up or it'll just drop a whole chunk of [ __ ]
297. 51:20out of it and so then it's analyzing stuff you're like but wait where's the other numbers and it was like oh I'm
298. 51:25sorry yes I got that wrong let me try again and it's just it's just wrong again so it's just like I just the the
299. 51:32reasoning capability of chat GPT right now is just stupid so so just assume it can't do math oops got cut off could 7 trillion do buy set
300. 51:44up a satellite to manufacture chips you mean a satellite like up in the up in the planets to Manu I I don't know if
301. 51:54you gain anything by Manufacturing in zero gravity but maybe but $7 trillion could buy you a lot of [ __ ] satellites
302. 52:05yes but I assume what we're talking about here is big giant sprawling you know Tesla scale
303. 52:19factories 7 trillion won't fit in a calculator can chat GPT analyze a website and give Su suggestions on how
304. 52:25to improve it it can can do that all day long it's actually good at that um and Gemini Pro 1.5 I don't have
305. 52:38access to it yet but the people that have got access to it right now they're doing things like recording screen sessions of them
306. 52:48using applications and having Gemini watch the video and suggest um process improvements so I think that I think
307. 52:58that once you once you start to get like million token context windows with multimodal stuff you're going to be able
308. 53:06to use all sorts of inputs to do all sorts of crazy [ __ ] but yeah chat jpt right now is good you can give it code
309. 53:13and tell it to suggest how to improve things you can give it code and tell it to change it and you know make it look
310. 53:21better or different um you can probably just upload an interface and have it you know describe it and maybe even describe
311. 53:30ways to improve it improve usability yeah it's good at that stuff good good good is it okay to have the humble opinion the difference
312. 53:43between wait is it okay to have your hum oh humble opinion the difference between an employee and an employer with AI I don't
313. 53:57quite understand the question if if you're saying can AI completely replace humans right now in most jobs the answer is
314. 54:16no but a lot of businesses that I know that are starting to get their heads around AI are looking at their employees saying who's AI
315. 54:32literate and if they're making if they're prioritizing employees they're prioritizing toward the AI literate
316. 54:41employees now depending on the job as these tools get better the ability to completely replace a role or function of an employee today
317. 54:54is going to be very possible I think I think it's something like I heard this stat today that like 80% of the
318. 55:02jobs like the data jobs that people do today are repetitive um like low contribution just like repetitive Cog in the wheel kind of
319. 55:17tasks that's the perfect kind of thing that AI is going to be able to automate first so the people that are getting AI
320. 55:26literate who are expanding their skill set they're they're going to be in much better shape than people who are sitting
321. 55:34on the sidelines that this channel is all about we can debate whether AI is good or bad evil or genius um janky and unusable or
322. 55:48brilliant and [ __ ] profound we can debate that [ __ ] all day every day and the AI is still coming it's not
323. 55:56going away all of these peripheral conversations about the validity of AI and what's it going to do and is it good
324. 56:04or bad that it replaces shitty repetitive jobs is all those are all independent conversations from the fact that it's
325. 56:16coming and quite frankly it's here it's just that not that many people know it yet and the tools are not that they're not at the
326. 56:29level that it's a threat on the surface would you ever let an AI robot operate on you [ __ ] yes are you kidding
327. 56:43me did go cap Captain mut go read about the robot that they built that Elon Musk had to build for neuralink it's insane what it does so
328. 56:58they drill a hole in your skull and then in your brain you got your neurons and you got your little blood vessels
329. 57:11right and what neuralink is is 60 or 120 I for I don't I forget how big the bundles are at this point I think they're 60 there's 60 um
330. 57:27what are they called uh little wires little probes there's 60 little probes that have to go in your brain to connect
331. 57:35to different neurop Pathways right they have to insert them in between the blood vessels so that they don't poke into one and give you a
332. 57:46stroke so so and and the hole is like I don't know the size of a I don't know they I think they were originally going to be
333. 57:56the size of a pencil eraser I think they're now about the size of a quarter but whatever it's a small hole and so
334. 58:02this robot looks into your brain and it actually steers those things around your blood vessels like 60 of
335. 58:10them and just inserts them one by one steering them around the thing so yeah would would I let something with that
336. 58:17level of precision operate on me rather than some guy that maybe had too much coffee or not enough or went out on a bender the night before
337. 58:26sure yes do I want to be the first guy operated on by that robot no but by the time a a non-clinical trial person is going to be operated on
338. 58:39by that robot it'll have gone through a bunch of [ __ ] I had surgery in early 2023 with a da Vinci in room physician
339. 58:48and a remote co-pilot in Florida yeah exactly absolutely I would absolutely like yes yes I mean this is no different to me
340. 59:02than what what I would I like a uh you know would I find it acceptable to talk to a robot if the robot actually
341. 59:14answered the phone and could solve my customer service problem yes yes I would yes I would I would gladly talk to a
342. 59:24robot that's like hi Kyle how can I help you today oh well I tried to open my refrigerator door and it fell off in my
343. 59:32hand what the [ __ ] am I supposed to do now oh that's really unfortunate Mr Shannon that sounds that sounds like
344. 59:39something we got to get a technician out there right away don't we well yeah if you don't want my food to go bad all
345. 59:44right let's go ahead and do that that technician's been booked they're on their way would I prefer that experience to
346. 59:52the one that I went through when that actually happened yeah cuz that was three weeks of [ __ ] hell did the surgeon press shift enter
347. 1:00:05exactly that was a joke from last night there's a there's a kind of funny uh ux experience on a Microsoft co-pilot and
348. 1:00:15the neighboring Little Thing Called Microsoft notebook which if you didn't see it last night I'll let you go Discover it for yourself it's
349. 1:00:24special but yeah [ __ ] yes but I also have a uh I have a world ID like what's a world ID Kyle I don't know what a world ID wait
350. 1:00:39what's what the hell was that that was bizarre it said your live will end in a minute but I had to like put a puzzle
351. 1:00:54piece weird I spoke with gp4 for the first time today it's often awesome oh that's great Apple user oh yeah you've been in here
352. 1:01:03for a while yeah I'm telling you gp4 is just a different thing I got on co- co-pilot today and did the exact same
353. 1:01:10thing so frustrating yeah so if you don't know in co-pilot right now there's two modes there's co-pilot you go to co- pilot.
354. 1:01:19microsoft.com and then there's co-pilot and that's just like chat GPT and you type in a prompt and you hit
355. 1:01:28return and then if you want to put a line break in your prompt like if you want to go hey based on this um thing
356. 1:01:35I'm about to paste in here shift return two line breaks and then you paste in a blob and then shift return and you give
357. 1:01:42it the rest of your prompt and then you hit return and it enters it right next to co-pilot is this thing called
358. 1:01:49notebook and notebook is a hor it's like a side by side like instead of your prompt being on the bottom your prompt
359. 1:01:55is on the the side and that one you when you hit return it gives you a line break and when you hit shift
360. 1:02:04return it enters it it's the complete opposite action of the thing that's right beside it it's Inc it's it's
361. 1:02:11Insanity tiktock did that to me today in my life oh good uh investor friendly agent because that was crazy I thought I
362. 1:02:20might have been getting booted out of here forever cuz I said whatever something something stupid you can make money
363. 1:02:34[Music] with I want customer service like the Alaskan Airways where their chatbot made
364. 1:02:49a new policy for bereavement I didn't hear about that one new security feature probably because of AI and people making deep fakes wow yeah
365. 1:02:59maybe that's possible but when we get agents because these things have Vision all of these captas and Slide the puzzle piece to the
366. 1:03:09puzzle piece all of the AIS so the AIS are going to be able to do all that so there's going to have to
367. 1:03:18be a completely new method like all of the current security two Factor authentic all that bullshit's going to just
368. 1:03:26get obliterated in the next three or four years so it's going to be crazy it's going to be I'm I'm telling you 2023 last normal
369. 1:03:37Year all right oh that was Canada air and it was actually eight eight months ago got my big presentation about school
370. 1:03:44AI tomorrow Mr K good luck with that so everybody uh tap the screen wish wish wish Mr K good luck so Mr K is a teacher in central
371. 1:03:56Pennsylvania I grew up in York Pennsylvania so East Eastern York High School Mr K so just south of you there
372. 1:04:04buddy that's that's how you got this [ __ ] genius pile of gray matter um I guess pitch into the school board
373. 1:04:16tomorrow about um we should be using AI what Mr K is going to go tell them is um Hey listen there's these things out there
374. 1:04:26called large language models and there are students our students are large language models and it's our job to get language in their
375. 1:04:39frustrating little brains and I don't know if you've heard this but another thing came along called a large language model that are these
376. 1:04:47computer programs that are the most remarkable knowledge systems in the history of humanity and right now we're not allowed to use
377. 1:04:58them and so what I would like to suggest to us as a community is maybe we should take these large language models and
378. 1:05:06help these large language models learn faster and better and different as if these new things were going to be around when these kids
379. 1:05:18graduate so wish Mr K good [ __ ] luck yes go tell them go blow their minds and go tell them this [ __ ] isn't going
380. 1:05:31away these are not plagiarism machines these these are not cheating machines these are Humanity amplifying
381. 1:05:41tools our kids will learn faster deeper better they will be more passionate about [ __ ] their education will be more personalized
382. 1:05:56holy [ __ ] [ __ ] you don't want me in the room like listen
383. 1:06:11dipsticks so please if you're in the AI Salon go into the education Guild and give Mr K some shout outs and some
384. 1:06:20good loving education Guild Monday night meeting right before the AI learning awesome fantastic thank you Brandon
385. 1:06:27that's great so I keep talking about the salon what the hell is that so if you are curious about AI or if you're kind of like me where
386. 1:06:42you're like I kind of think this AI thing's a big deal but everyone else in my life seems to think it's not a big
387. 1:06:48deal am I wrong no you're not wrong you're just hanging out with the wrong crowd you're hanging out with bad influencers love that Humanity
388. 1:06:58amplifying Kyle for president 2028 this channel so okay so if you don't know if if you're new here there's a pile of people in here if
389. 1:07:11it seems like there's people in here that kind of know each other and have inside jokes and are answering your
390. 1:07:17questions those are the people that show up here regularly so if you don't know I do these things seven nights a week
391. 1:07:26why because I know how to make Good Life Choices the Irregulars the ones that show up here every night are living proof that this AI
392. 1:07:41stuff are Humanity amplifiers because to a person the people that just show up here regularly and are in this
393. 1:07:49conversation in what conversation it doesn't seem like you've said anything tonight yeah exactly you know why
394. 1:07:54they're here cuz while I'm rambling they're playing with AI and if I mention some tool they go Google it oh that's kind of cool then
395. 1:08:02they're playing with it and they're playing with they're doing this and they're doing that and they're like oh
396. 1:08:06[ __ ] all of a sudden oh I can make a really good image and then they share that image in the AI salon and then
397. 1:08:11someone else sees that image they're like how did you do that like oh I don't know I I uploaded this image and then I
398. 1:08:16had it do that and then oh my God that's amazing and then they're talking to one another and and and what's happening is they're getting
399. 1:08:27reignited creatively like like you know the human spark that spark that's like I got something to say I I want to put
400. 1:08:36something into the world you know that thing you had when you were seven and then the world [ __ ] beats it out of
401. 1:08:45you people that are playing with these tools regularly that like that that fire is reigniting and they're like oh what what if I did this
402. 1:08:56what if if I did that could I do this and so I I get to witness it in here every night everyone who shows up here
403. 1:09:03every night I'm The Quiet One a newer regular awesome yeah jimc I know you've been in here a lot it's great is we're all getting to witness
404. 1:09:13and support and celebrate one another as we're rediscovering who we are that's what these tools are can they be used for bad might they
405. 1:09:26do bad things yeah in the meantime until it shuts down the grid and ruins our lives forever why don't we see what they are
406. 1:09:39why don't we play with them that's what this group is so if you go there it's going to take you to a link tree you can there's merch
407. 1:09:51there you can buy T-shirts that support this group because this doesn't cost anything there's a similar Channel out there that
408. 1:09:59kind of does what I do he created a community he's charging 500 bucks for it this is free not like free to get into level
409. 1:10:18one and the people that are here are [ __ ] remarkable thank thank you Sherry D thank you so much Sherry D is one of
410. 1:10:29them 30-year nursing career she's now making music videos for bands like really good ones that's exactly what's happening I felt like job
411. 1:10:41the job market is trash and this gives me hope yep exactly riggity exactly and what's the hope I don't know
412. 1:10:54what's this thing going to give you that's going to be the thing that's going to help you survive all this upheaval I
413. 1:11:00don't know but here's what I know if you're sitting on the sidelines with your arms crossed these tools ain't all that
414. 1:11:08you're not going to be in as good a shape as someone who's hanging out in here figuring out what the hell this stuff can do promise you
415. 1:11:16that so if you go there the first link says the AI Salon Community click on that and it's going to take you
416. 1:11:30to here welcome to the AI salon and you know all your friends that are like you know Sally
417. 1:11:46um we need to talk so here's the thing I know we said we wanted to go out to drinks tonight but this is kind of a
418. 1:11:52little bit of an intervention I don't have a drinking problem no it's not about the booze we like no we like
419. 1:11:58drinking it's about the AI could you we we we've talked and we've decided that you talk about AI it's all you talk
420. 1:12:07about anymore and um how do we say this um we want the old Sally back um could could we just do some melon poppers [Laughter]
421. 1:12:22and so if your friends are doing that to you this group is not that this group is like oh you're really into AI cool come
422. 1:12:34on oh you're really curious about it cool come on and notice this is not a single person heading off into the abyss one of
423. 1:12:43the one of the realities right now I promise you I promise you to a person every single person in the AI salon
424. 1:12:50right now feels like they are behind and undereducated and UND skilled like like every someone else in this community has
425. 1:13:00to know more than they do every single person feels like they can't keep up they don't they're not good at this
426. 1:13:09[ __ ] they like the stuff that they're doing but they apologize for it because it's not enough to a person I'm trying to
427. 1:13:16eradicate the apologizing in this group by the way because what the people are doing in this group is remarkable because what
428. 1:13:23they're doing is they're playing and that's the only skill you need right now so this group will welcome you in
429. 1:13:35with open arms and then here's my request so if you're here read about who we are and why we do what we do and then what I really want
430. 1:13:43you to take in like I really want you to take this in like like as passionate as I'm talking right now I want you to really read the
431. 1:13:52values so the values here are not like this is who we want to be the values here of this is who this community already
432. 1:14:03is and the thing that's happening here is this community is starting to to grow quickly and and it's it's starting to
433. 1:14:13accelerate and I've grown you know I've start I started a company that grew really big really fast I started a an
434. 1:14:21organization kind of like this one in 1994 called the worldwide web artist Consortium That Grew really big really
435. 1:14:28fast and one of the things that happens when you grow really big really fast is you you kind of lose the good [ __ ] cuz the new people
436. 1:14:38overwhelm the old people and just it just becomes chaos so the reason I talk about these values most nights I I think every
437. 1:14:49night is that you shouldn't join this community if you don't dig these values because this this is who this community already
438. 1:15:01is okay so the first one's curiosity like we want people in here that are curious not cynical right you can be
439. 1:15:07cynical but just be funny about it like if you're cynical like a standup comedian is cynical like like Tom
440. 1:15:17sigura or or Doug stanh hope cool that's fine but in here you should be curious about this AI stuff huh H wonder how
441. 1:15:25that works whoa what would happen if I did this right next next next value is is generosity so don't be stingy with like
442. 1:15:36oh I just learned something I'm going to keep that to myself no no I just learned something let me see if anyone else
443. 1:15:41knows that too here's how I did this anyone else oh how'd you do that oh here's how I did that a lot of back and forth right generosity
444. 1:15:51exploration people that are excited to like strap on the backpack and just head into the unknown cuz that's all what
445. 1:15:59we're what we're all going to be facing for the next 5 years is a lot of unknown so you want to be around people who even if they're not
446. 1:16:12comfortable in that scenario are willing to do it this is a community of people willing to step into the unknown do you know how [ __ ]
447. 1:16:23inspiring that is to be hanging out with people that are willing to do that you know how few people in life are willing to do
448. 1:16:32that so so that's who this group is collaboration is another one you know willing to do it together inclusion like I don't care if you've
449. 1:16:41got you know six hours experience with AI or six decades experience with AI doesn't matter all points of view right
450. 1:16:50now are contributing to everyone's understanding because no one knows anything everything's changing too fast empathy is a big
451. 1:16:59one people are scared people are angry people are dismissive there's been some of that in here tonight oh you know
452. 1:17:07these things are just predictive engines well they're not perfect well they can't do high level coding okay
453. 1:17:16understand but understand that sitting underneath that kind of bravado is fear underneath that anger is insecurity oh no what's going to
454. 1:17:33happen if if if this thing I've been doing for 20 years all of a sudden is irrelevant it's very real right so
455. 1:17:40imagine being in a community of people who tap into that and are willing to hear that and not just interact with
456. 1:17:49your dismissiveness or your anger but are willing to say oh yeah what's really going on on here well I've been doing
457. 1:17:56all this thing and I don't feel like learning a new career yeah that must suck come on in let's let's see if we
458. 1:18:02can figure something out together like that's who this group is and then the last one is gratitude like actively actively acknowledging one
459. 1:18:15another if someone if someone around you is generous enough to tell you how they're doing something or give you a leg up thank them
460. 1:18:30the more bulgy I get that's funny Corey Sandler I'm on the put me on put me in coach I'm ready to play oh man so anyway that's the values
461. 1:18:44of this group and I I know that was a little a little preachy but [ __ ] it I mean I mean I'm very serious
462. 1:18:53about how remarkable this group is and how much how committed I am to keeping it that way um I mostly only use an iPad and my
463. 1:19:04iPhone can I get into AI stuff with just that absolutely pres dog so pres doog so you can right now just using Safari um go to chat.
464. 1:19:18open.com I would strongly recommend subscribing to chat GPT plus which is their 20 Buck a month um subscription to upgrade you to chat
465. 1:19:28GPT 4 you can also download the chat GPT app make sure you download the official one from open AI make sure that the
466. 1:19:38company that makes chat GPT is open AI anything other than that is a faker app that's out to rip you off but yeah you can absolutely do all
467. 1:19:48this stuff right there I would also download the PI app I think perplexity has an app I would download the perplexity
468. 1:19:58app I think Claude has an app from anthropic so yes absolutely okay so when you come in here
469. 1:20:15so if you if you dig what what we're about cool the next thing I want you to do is introduce yourself take that Val of
470. 1:20:27generosity and come in and introduce yourself mess Schmidt said hi that's a little stingy but he at least introduced
471. 1:20:38himself right I think of he but ideally what you do is you come in here you say hey here's who I am here's what I'm doing here's what I dig
472. 1:20:49in my life here's what I've done in my life here's where I am with AI sure would would be swell to find someone who blank does what does
473. 1:20:58something knows something about this is willing to just help someone out right and then you can check out our
474. 1:21:11events and if you go to this this third link the the two links below introduce yourself is this thing called AI Launchpad and
475. 1:21:19tools and that's a page that I personally curate wait progg I'm old 59 just a truck driver and a pilot but AI
476. 1:21:29seriously intrigues me pres dog I'm 58 and and I I'm just a I'm I'm just a guy that that's playing around with uh
477. 1:21:40with storytelling tools and Technology the the what you what you do and have done is irrelevant to what's going on
478. 1:21:49right now if AI seriously intrigues me I mean seriously okay okay you're a truck driver okay PR dog you're going to [ __ ] love this go hang
479. 1:22:02on hang on go so you only have an iPad and an iPhone right so you got Bluetooth in your rig so I want you to you can do this
480. 1:22:15with either the chat GPT app or you can get this app called Pi Pi it's it's Pi personal assistant and the the icon
481. 1:22:23looks like that it's it's green two green letters Pi go get that app so it's it's like chat
482. 1:22:37GPT except it's optimized for conversation and so oh you have the pie okay cool perfect so I don't know if you
483. 1:22:48use it if and I don't know if you use it like this but go on your long runs to 2hour drive 4our drive 8 hour drive
484. 1:22:57whatever it is right and just start having a conversation with pi it's connected to the internet and it's
485. 1:23:02designed to talk and you know so there you are and if you don't know it that little icon right there the little phone
486. 1:23:09icon puts it into conversation mode I'm on a long trip and I'm a little bored can you help me come up with a way to pass the
487. 1:23:23time sure thing Kyle long trips can definitely be a bit tedious sometimes let's see if we can spice things up a
488. 1:23:31bit how about we play a game of would you rather I'll give you two options for different scenarios and you have to
489. 1:23:37choose one ready to get started now I think what I'd rather do is you tell me a a news story that happened today but
490. 1:23:46keep the details from me and see if I can guess about it all right Kyle here's a news story that happened today I'll keep the details
491. 1:24:00hidden so you'll have to use your detective skills to figure it out a major event occurred in a region that
492. 1:24:07has been in the headlines recently leading to a significant loss of life and injuries it involved a large group
493. 1:24:15of people and a convoy of Aid yeah yeah I know what that one is I'm not going to talk about that one I'm not getting
494. 1:24:20booted off Tik Tok for that but anyway you can you know it can make up games for you it can talk to you you can
495. 1:24:28have deep you know contemplative conversations with it you can have it teach you something you can have it
496. 1:24:35explain books to you movies to you I just all sorts of stuff so um that would be a really like having having pie teach you about AI while
497. 1:24:48you're driving would be kind of kind of wild and cool cuz the other thing about those conversations is you're having
498. 1:24:56these conversations it's recording them all in text the other thing you can do with chat GPT is if you've got chat GPT 4 it's got dolly in
499. 1:25:10it so you do a similar sort of thing with chat pt4 but you can have it make images I'm on a long road trip right now
500. 1:25:23and I'm driving through Mid Ohio can you make me a picture of a blue semi truck driving down the highway in Mid
501. 1:25:39Ohio and so it's off now rendering an image so so not only can you have a conversation you could be making art while you're
502. 1:25:50driving now why you would want to do that I don't know but you would at least start to learn the capabilities of these things and what's
503. 1:26:00possible and's see what it does see what it comes up with look there's here's a picture of a blue semi- Tru driving down the highway
504. 1:26:09in Mid Ohio it really captures the sunny day and the essence of a road trip through the Midwest how's your road trip
505. 1:26:16going so far you can check out the images in our conversation history it's going okay I'm a little board though can you come up with a game
506. 1:26:25for us to play sure how about a classic game of 20 questions it's perfect for road trips I'll think of something there you go so
507. 1:26:40cool right anyway so yes yes get curious about it here what what's what's happening in here pres dog pretty consistently
508. 1:27:00is people kind of come in and they're like I think I want to learn about Ai and they're like I don't know anything help and
509. 1:27:10then very often what they realize is everyone they talk to feels the same way even if they've been doing this AI stuff
510. 1:27:17for a while like people might know more than them but they're like oh everyone's kind of trying to figure this out and
511. 1:27:23there's something really comforting in that and then if you haven't had it yet you know at some point you'll have what I
512. 1:27:38call your Kevin McAllister moment right from Home Alone where these tools will do something that is so good or profound or just
513. 1:27:50unexpected that you're like I didn't know it could do that right and it's and then the the more you use these
514. 1:27:57tools the more you have the sort of rolling series of Kevin mallister moments and if you keep pushing and keep
515. 1:28:08playing and keep sort of figuring out where the edges of your interests are all of a [Music] sudden you'll discover something you're
516. 1:28:20either good at or you were super passionate about out when you were younger or something where all of a sudden it like it will
517. 1:28:27reignite something in you or it it it'll ignite something new in you and you'll be like oh well let me explore this let
518. 1:28:34me explore that and I'm just I see it in here every flipping night what are your thoughts on AI and Automation in homes
519. 1:28:44um my my thought on any sort of complex systems like home automation systems are super cool if you're an Uber nerd right
520. 1:29:02because all the home automation systems they've all got their own proprietary [ __ ] systems and they're all
521. 1:29:09incompatible with each other and they all have really shitty interfaces and so most home automation nerds started with this system and then
522. 1:29:19added some components from that system and then they've got the their Sonos system over here they got that system
523. 1:29:25over there and they got their GE LED dial color light bulbs over here and none of them talk to each other and so only the quantum physicist
524. 1:29:37in the house that put the [ __ ] together can deal with it right and so I think what AI is going to do is it's going to
525. 1:29:47simplify those complex systems where when we have say an autonomous agent that you can talk to like pi and you can
526. 1:29:54just say hey go figure this [ __ ] out make it warmer and make it green in here and turn on the TV and turn down the
527. 1:30:03radio and make the lights blink purple and then just just start doing it then it'll be cool so I don't know if anyone's built
528. 1:30:18that yet but I like I haven't touched home automation [ __ ] cuz I don't feel like spending 17 hours a week debugging why
529. 1:30:29my light won't turn on in the you know next to the the reading Nook I had no interest in it but if I can just say AI figure that [ __ ] out
530. 1:30:42I'll just go plug some [ __ ] in I'll I'll give you access to all the [ __ ] I plugged in and you figure it out then
531. 1:30:49I'm in I did ask chat gbd 3.5 to find me cheap flights and actually worked and that that's actually remarkable cuz 3.5
532. 1:30:57isn't connected to the internet so how did it do that you sure or did you use co-pilot if you want if you want access
533. 1:31:08to chat GPT 4 the underlying engine gp4 that's connected to the internet but you can't afford the 20 bucks a month for
534. 1:31:16chat GPT go to Coop pilot. microsoft.com I've watched you many nights I've learned a lot just listening to you oh that's awesome PR dog I
535. 1:31:25appreciate it yeah I thought I've seen you in here before I [ __ ] love that you're a trucker and that you're you're trying to
536. 1:31:31figure this [ __ ] out I I I I just I feel like anyone who's willing to go into that unknown uncomfortable space and
537. 1:31:41just try to figure some [ __ ] out is going to be gifted like all sorts of New Opportunities so I think I think where
538. 1:31:50you are right now is like perfect like I'm intrigued by this but this is what I do like that whatever the
539. 1:31:57limiting beliefs are of however you've defined yourself right that that maybe this is out of your reach drop that [ __ ]
540. 1:32:06this is this is this is actually really important November 30th 2022 when chat GPT launched here's here's the big deal
541. 1:32:13about it from from where I sit prior to chat GPT if you wanted to do this AI [ __ ] you kind of had to be a geek
542. 1:32:24you like it didn't hurt if you went to MIT having the Stamford mathematics degree probably a good thing if you're
543. 1:32:32trying to figure out tensor flow and and embeddings and mathematical weightings and probability matrices chat
544. 1:32:45GPT does for machine learning and AI what the worldwide web did for the internet so prior to the worldwide web
545. 1:32:53the internet was command line it was researchers it was scientists it was academics and then the worldwide web
546. 1:33:01comes along where you can sort of point and click on hyperlinks and the only skill you need to be able to use the
547. 1:33:07Internet is this click click click click click chat GPT does the same thing for all this stuff all you need to be able to do is
548. 1:33:18talk to it and I think because of terms like prompt Eng engineering is really important you've got to be a prompt
549. 1:33:26engineer it makes it sound really [ __ ] intimidating what prompt engineering is is use your words good to ask for [ __ ] you want
550. 1:33:38good and then the machines will give you good stuff back like that's what it is now there's all sorts of ways you can do
551. 1:33:47prompting one of them is you can do prompt engineering and very structured prompting and approach it like a systems analyst
552. 1:33:54another way you can use prompts is just talk to these things as if you were talking to someone who could do [ __ ] hey could you do that
553. 1:34:04[ __ ] sure here's that [ __ ] that's not very good could you do that [ __ ] better sure here's better that [ __ ] like that's
554. 1:34:15it and if you're willing to just dive into it and go [ __ ] it I'll just try I think the rewards are insane
555. 1:34:24I've been missing these sessions you have really changed my my AI life thank you very much thank you Carl welcome
556. 1:34:30welcome back yeah just keep coming keep coming keep hanging out this is um use your words good exactly Lori you got to use your words
557. 1:34:44good and you want to know something [ __ ] amazing if you're not so good with the words say to chat GP
558. 1:34:54hey I want to do a thing but I'm not so good with the words can you help me come up with a prompt that will get you to do this
559. 1:35:03thing better and guess what it will in fact when you ask chat gbt to make you an image it doesn't use your prompt you want to see [Music]
560. 1:35:19something okay here we are chat GPT chat GPT 4 we're just in normal chat gbt I love the way Kyle talks about
561. 1:35:29playing around to find out and learn I think it's the only I mean I appreciate that thank you Winston I think it's the only the only thing
562. 1:35:36that does not seem like insanity to me right now is playing because it it it actually seems like insanity to me right
563. 1:35:44now to try to get your head around all of what's happening and try to master any of these tools it's just it's it's stupid because
564. 1:35:53they're just they're changing too fast there's too many of them so so don't take it so seriously like this is a rare
565. 1:36:00moment in history where where we get to just play with this [ __ ] okay so I'm in regular chat GPT I'm going to say I'll
566. 1:36:06I'll do my classic make me a me a wide photo of a green
567. 1:36:2170s muscle car in an abandoned Factory all right so I did those words pretty good they were pretty good words
568. 1:36:36they were not bad they're they're not great words right green 70s muscle card an abandoned Factory a wide
569. 1:36:46photo so we're going to we're going to let it make its little picture here we'll see what it comes up with I love to listen to you thank you
570. 1:36:54very much I appreciate that side hustle Mimi have a good night everyone I'm falling asleep at the keys good night good night
571. 1:37:06Mimi it's it is my job I'm like you could take Xanax or you can listen to the AI learning lab I'll knock you right
572. 1:37:21out oh man come on chat GPT what's going on with this image show me the
573. 1:37:39image where my image where my image I used it to fix my res resume perfect use case um let's stop this and restart it
574. 1:37:55uh I'll do try again when will emo have an app um I would imagine emo is so powerful so by the way emo so if you don't know what
575. 1:38:08emo is emo is a tool that Alibaba actually came out with um that will take a still image and a recording of someone
576. 1:38:21talking and not only animate the lips but like fully animate the head eyebrows all the facial expressions um and it came out yesterday
577. 1:38:31as a white paper um Mr Studio it is like it it was one of those white papers that so many people are talking about it that I would be
578. 1:38:42surprised if by the end of the weekend there wasn't some um Google collab notebook that you could just you know
579. 1:38:53get up up and running and have it running like I would imagine that there's probably some version of that right
580. 1:38:59now um anyway I was talking before about promp prompting and prompt engineering and how I I think that's that term is
581. 1:39:09doing beginners of big disservice that get intimidated by it and that you can just use simple words and and these
582. 1:39:17tools are actually getting better at helping us so the way chat GPT works I typed in I created oh wait I I typed in
583. 1:39:26make me a wide photo of a green 70s muscle car in an abandoned Factory if I click on this image in the upper right
584. 1:39:33hand corner is the information button here's the prompt that chat GPT wrote a wide photo depicting a vibrant
585. 1:39:41green muscle car positioned in the midst of an abandoned factory setting the Factory's interior is marked by a vast
586. 1:39:50desolate space filled with remnants of industrial machinery rusted with Co rusted and coated with Decades of neglect sunlight streams
587. 1:39:59through the broken painless Windows Illuminating the dust particles in the air and creating a dramatic interplay of
588. 1:40:05light and Shadow the muscle car with its glossy well-maintained exterior stands in stark contrast to the surrounding
589. 1:40:14Decay offering a striking I I mean this is [ __ ] impressive right Kyle is my new Night Show theme
590. 1:40:32[Music] song um isn't that crazy so I typed in a shitty prompt and it said I got you boss here
591. 1:40:48you go here's an incredibly well-written incredibly descriptive prompt that's going to generate that so don't be intimidated by this
592. 1:41:04[ __ ] just play with it you'll be surprised oh by the way did you guys know sunno sunno AI that does songy stuff I
593. 1:41:22didn't know this I found this out last night after I was leaving or after I left now available V3 Alpha access for pro and Premier Members I
594. 1:41:34don't think I'm a pro member sign in with Google did I pay for this maybe I paid for this what's
595. 1:41:50new V3 access extended Max song Length faster song
596. 1:42:07generation where custom mode V2 yeah I don't think I've got access to it but anyway if you haven't seen
597. 1:42:24I love to oh thank you um let's see country song about driving my big rig
598. 1:42:44through Mid Ohio yeah I've only got V2 but anyway there's a V3 of this that
599. 1:42:58you can do longer song lengths and apparently it's better so if if you like making music but man this is much faster rolling on the mid
600. 1:43:07Ohio highway let's these these could be bad kicking a I hit the RO make my way through the H
601. 1:43:24nothing going to make me St that's [Music]
602. 1:43:39okay it's not singing where's the [Music] words and it didn't do the words for that one that's funny sometimes it doesn't do words janky is
603. 1:43:53dances in Ohio and yeah anyway new version of sunno out so if you're into the suo stuff go go get the V3 of
604. 1:44:08[Music] it all right everybody I think I'm going to wind I'm going to wind her down
605. 1:44:24um tomorrow so a couple of things tomorrow I've got AI office hours on LinkedIn so if you find my profile on
606. 1:44:33LinkedIn if you want to hang out during the day um it's at 11: a.m. mountain time so 1: p.m. Eastern 10:00 a.m.
607. 1:44:48Pacific um and I don't there's nothing up there right now so um it'll be there in the morning just go to my profile it's
608. 1:44:59there oh yeah then tomorrow's Friday night date night so oh Robert Rossy thank you for the dyno I love that one that was
609. 1:45:07beautiful very generous thank you um so you can hang out there we can do Friday night Daye night also um Saturday
610. 1:45:18at noon Pacific Ann Murphy does her GPT weekend jams so in the AI salon and if you if you're new here and don't know what I'm talking
611. 1:45:30about go to the salon. click on the very first link on that page it's a link tree where it says AI Salon community
612. 1:45:40and check it out under the technology Guild is weekend jams with uh with an the GPT jams um so that's Saturday at noon you
613. 1:45:53should go to those if you're curious about this [ __ ] you should go to those uh and what else if you want to support this
614. 1:46:04channel you can follow me five minute goodbyes you can follow the channel that'll support you can subscribe to the
615. 1:46:14lives that supports you can buy the video series that's in the corner there called underhyped if you're actually
616. 1:46:21especially if you're in a business and you're trying to figure this [ __ ] out um that video series talks about why AI is a bigger deal
617. 1:46:29than you know the people that are dismissing it or demonizing it are kind of missing some of the big picture and
618. 1:46:36so that's that's what I talk about in that series and the and the way that you can really support this channel is just keep showing up
619. 1:46:46here well when when when do I show up Kyle I don't know show up every night I do it's generally 8:00 Mountain every night thank you
620. 1:46:57Winston I try to keep it consistent I used to be just chaos I would just go on any time of day and people were like
621. 1:47:03I'll I'll come on whenever you're on but this is kind of a pain in the ass and then I thought yeah [ __ ] it why don't I
622. 1:47:09just do it the same time every night it's easy for everyone and easy for me you know and just show up
623. 1:47:21here and don't have an agenda like don't even feel like like I just feel like it's less and less important to teach anything on this
624. 1:47:35channel it's called the AI learning lab we want our lessons than we want them now teach us now I feel like the most important thing
625. 1:47:46is just like commit to being in the conversation commit to playing James come out thanks thanks for your knowledge share Kyle
626. 1:47:57appreciate it thank you very much I appreciate the kind words just show up here if you learn something
627. 1:48:05great if you show up here and what I'm talking about is boring the [ __ ] out of you then just hang out here and keep
628. 1:48:12making images or play with perplexity or see if you can get it to do some coding that doesn't actually suck take the same
629. 1:48:24prompt and try it in chat GPT 3.5 chat GPT 4 Claude perplexity grock uh uh what else um po all the different models in po like just play
630. 1:48:40just try different [ __ ] out and go oh that was really horrible but that's kind of amazing Kyle tomorrow afternoon Pete and
631. 1:48:48Leah are going to talk about mid Journey oh that's awesome Dr J yeah so just go to the salon check out the events go to
632. 1:48:55the guilds check out the guilds they've all got events going on oh there's good thank you Brandon there's also down at the bottom of the
633. 1:49:05guilds there's a thing that says Community clubs there's only one Community Club it's this group it's called a l
634. 1:49:16Irregulars these are the Irregulars the people that show up here all the time they're the Irregulars so go there scroll to the
635. 1:49:23bottom find Community clubs and hang out in the Irregulars there's a chat area where you can chat and there's another
636. 1:49:30area where you can share images and we didn't really do any image sharing tonight but I'm sure some of you
637. 1:49:36did anyway you can go in there and troll me make funny pictures of me acting like a dumb Dum it's it's a great Pastime amongst the
638. 1:49:50Irregulars and just show up just show up just show up just show up just show up just play play play play play and then talk about
639. 1:50:00it why be in a community so that you can talk about this stuff who else should you talk with this stuff about
640. 1:50:12strangers hey Jim how's it going it's going okay hey have you played around with any of this AI stuff I heard it's
641. 1:50:20evil let me show you something cool get out your phone write him a song make him a picture here let me pick make make a picture of your
642. 1:50:34Diner all right how'd you do that it's AI it's cool anyway see you have a good day tell your friends and family the
643. 1:50:44ones that tell you not to talk about AI [ __ ] them tell them anyway your co-workers we're not you know we're not allowed to to use AI here
644. 1:50:53what are you why would you even talk about it that seems pretty risky tell them Jenny's such a rebel she's just out
645. 1:51:09she's you know what she does on the weekends she's got a Chromebook and she plays with the AI she plays with the
646. 1:51:16chat GTP she plays with it all weekend and she comes in here and all her work's done she gets her work done on the
647. 1:51:22weekends and she comes in here and she just Lolly gags about be like [Laughter] Jenny all right so thank you thank you
648. 1:51:38to the Irregulars so anyone who shows up here more than once you're you're saintly because why would you
649. 1:51:48um thank you to the subscribers thank you to the mods thank you to anyone who gifts me during the during the things I
650. 1:51:55don't I I feel like I don't thank you folks enough the mods who sort of elevate your comments so that I can see them and and
651. 1:52:05comment to them and sort of keep me on the path thank you for all of that and yeah so all right let's see who
652. 1:52:16we talking to going to hold up my little magic mirror here who do I see I see Lord digital Gods I see Brandon and papa
653. 1:52:24JJ I see Corey Sandler Pottery you should follow Cory Sandler Pottery Corey c r i Sandler s n d like Adam Pottery like cups and
654. 1:52:40[ __ ] go follow her cuz she makes awesome Pottery you know what else she makes awesome of music you know El she makes
655. 1:52:48awesome of uh stews she she does cooking you know those she makes awesome of [ __ ] gpts she's like a GPT Master she's she's
656. 1:52:58a [ __ ] application developer now she made a GPT last week that will take an image you generate in Del or anywhere frankly and
657. 1:53:09it will automatically add a watermark on it that has a copyright With Your Name and All Rights Reserved on
658. 1:53:17it she's an application developer now go follow her anyway Brian Whitney good to see you Jane Doe Christy good to see you Mr
659. 1:53:28it show and tell uh no I'm I'm I'm Crispy I just want to I just need to go I just need to go napping but you should go check out the
660. 1:53:39showand tell channel in the salon too there's a whole showand tell area now we have four different channels one for
661. 1:53:44images one for songs one for videos and one for gpts so go check out the show and tell area Kyle show tell people are asking please okay
662. 1:53:56fine see you you just you just can't I just I wish I could quit [Laughter] you robot doing
663. 1:54:12surgery oh happy leap day oh there was there was a whole bunch of cool Leap Day things that one's great our road trip
664. 1:54:19the Dodge cult station wagon a genuine lemon that's flipping awesome nice Papa JJ win an iPad what is
665. 1:54:30this oh something created using AI yeah probably check out the patch on the uniform yep for sure that's AI for sure I need 7
666. 1:54:46trillion look see Cory Sandler All Rights Reserved see best way to spend a leap day oh that's beautiful who did that cam Katen that's
667. 1:55:00gorgeous big eyes oh that's gorgeous more leap year yeah happy leap day everybody class Salon oh that's gorgeous
668. 1:55:16Megan my head is a jungle nice wow yeah really inspiring I mean that's the whole thing is you know just be in a
669. 1:55:29community with people that are doing [ __ ] like this it'll it'll light you up wow that's awesome I do lots of videos
670. 1:55:35and picks when I fly everyone says I should Watermark there you go so so Corey Sandler so if you go to the GPT so
671. 1:55:44what I'm in right now is the image and Graphics show and tell if you go to the gpts not not a ton have been posted here
672. 1:55:51so like I think two things down yeah right here so Corey Sandler's um art Shield copyright protect watermarks that's the
673. 1:56:01link to her GPT you can also just go look for art shield and on the GPT store and it'll be there I got the itch Megan got the itch
674. 1:56:12so get in there get in there and do it all right everybody let's see investor until someone creates GPT how to remove her Watermark
675. 1:56:28exactly yeah the the the watermarks are just you know what it what it really is is wait I'm making another one right now
676. 1:56:37similar but a bit different I'll let you know when it's done that's great that's awesome anyway go follow her Robert
677. 1:56:43Rossy thank you Amelio's wife good to see you um Brandon thanks Danielle good to see you Jane Doe inhale exhale Joe 90 inhale exhale it's
678. 1:56:57pretty good healthy wealthy and wise good handle Sherry D good to see you Tobias thank you as always Daniel
679. 1:57:04Hollingsworth we're going to need to differentiate app developers from coders and Engineers if AI mashups or
680. 1:57:11apps I don't know that we're going to need to to differentiate them Daniel I think that's that's old world thinking I
681. 1:57:19think we are about to enter an era where everyone where where all this shit's democratized app development is democratized Art creation is
682. 1:57:30democratized video creation is democratized music is democratized all of it all of it all knowledge work data analysis is
683. 1:57:43democratized business Consulting is democratized all of it all the [ __ ] that we went to school for where we studied a special
684. 1:57:54is going to be available to everyone else I think the world gets really interesting when that happens and we're
685. 1:58:01not that far from it you know we're like three years from it being you know fully rolled out we're we're in the middle of
686. 1:58:08it right now um so so yeah I think right now you might want to distinguish between well Corey Sandler is not really
687. 1:58:18a developer cuz she's just using you know chat GPT that's not real devel development but I can tell you that the
688. 1:58:25way she is developing apps a developer will never have gotten there cuz she's coming at it as an artist with the ability to
689. 1:58:41code and then I look at someone like Roger Scott who was a coder for 30 years and the way he's approaching
690. 1:58:48images an artist would never get to and he's making amazing art right like so it's just so I think the the mashups
691. 1:58:58the the the societal mashups that we're about to experience are going to be absolutely insane and I don't think we
692. 1:59:05need to distinguish them I don't think it's important or relevant to say this one's a real developer and that one's
693. 1:59:11not this one's a real artist and that one's not no anyone who decides to express thems in any form is fair
694. 1:59:19game if it's good it's good how they got there who gives a [ __ ] so anyway I'll leave you with that tonight peace out
695. 1:59:28everybody keep showing up here hang out at the salon don't be an [ __ ] peace