1/27/2025 - AI Reasoning Engines: Deep Dive & Comparison of DeepSeek, Gemini, and GPT-o1

Transcript

1. 0:03but it's true all right
2. 0:22[Music] [Music]
3. 0:42[Music] is a place I can rest my head
4. 0:59[Music]
5. 1:15I'm singing like Timothy sham on Saturday Night [Laughter] [Music] Live yeah
6. 1:39[Music]
7. 1:55[Music] yeah should FL [Music] [Music]
8. 2:29good the good people you give this AI thing three days and someone comes along with a little model that just blows the whole damn
9. 2:38world up welcome to AI it's gonna be fantastic tonight Bob tell him what he's won oh hasn't still wonder thing this is
10. 2:49not a game show right carrying on um how is everyone tonight my name is Kyle Shannon this is the AI learning lab
11. 3:02we got some people filtering in uh they're watching me sing and they're filtering right back out [Music]
12. 3:26[Music] every time I see you now get that look in my every time see your mouth I hear that
13. 3:41smile New York state was a rolling Breeze and a sunshine with a blue sky falling J of old September [Music]
14. 3:57prep you again today you convince me again today you leaving this hotel looking for
15. 4:14someone else's cold and Rain say so long [Music] suzan hush now don't you cry
16. 4:31[Music] so long [Music] suzan don't you cry for me
17. 4:46[Music] [Music]
18. 5:01that is Champ Shannon on the vocals I got uh got Timmy back here on the drums Jasper on the bass my name's Kyle
19. 5:15Shannon I'm on the lead guitar and slightly flat vocals [Music]
20. 5:48yeah deep sea kind of blew up over the weekend that's an understatement what's what's amazing to me is like one of the ways that I kind of
21. 5:58follow the news is you know someone will launch something and you look at it and you're like oh okay all right I all and then
22. 6:08you'll see some chatter about it BL all right all right and then with deep seek it was like they launched it
23. 6:16and then there was chatter about it and then the chatter was like the big research company should be scared and then the chatter was like oh
24. 6:28the big researchers are in panic mode I heard it about meta first and I'm like they can't be in panic mode it's just another [ __ ] [Music]
25. 6:41model but it's how they it's how it's How It Was produced assuming we can believe it was produced how they say it was
26. 6:50produced which that I don't know if we can but they apparently produced it with 248 gpus two months so about $6 Million worth of worth of
27. 7:06training as opposed to you know hundreds of millions billions of dollars from the big the big companies and they're and
28. 7:12they're kicking their butts they also um today so it was like five days ago deep seek launched V1 three days ago four days ago it was
29. 7:24R1 the reasoning engine today they launched a multimodal engine I hear that they've got a voice engine
30. 7:31coming and then they've got apparently and this could just be rumor but rumor is chat TMZ style that they've also got a
31. 7:42uh like operator like a a web browsing agent so appar apparently they've got they've got all of that coming
32. 7:51so yeah Nvidia stock tanked like 20% today it's fun time being [Music]
33. 8:09a how how are we accessing deep seek so we're gonna play with that tonight let me let me switch this because it's hard
34. 8:15to read the letters and I got to put up my Black Box here for the people for the people there you go people um if you go to chat. deep
35. 8:26seek.com um it's just like chat GPT it's free you can also if you want to have your mind [ __ ] blown oh that's another
36. 8:34thing I should put up here hang on a sec I'll put up the mind blowing option go to gq.com gro.com
37. 8:52and you can flip that bad boy into deep seek R1 if you want to see if you want to see reasoning a reasoning engine at the speed of
38. 9:03thought uh go to gq.com and choose the R1 reasoning model all right fantastic Bob is it a Chinese data collection app for the
39. 9:17spying I don't know it might be as opposed to what the you you want our spying apps homegrown you only want our spying apps
40. 9:29here on uh here in [Music]
41. 9:44America I mean not for nothing Facebook and Google and all of them have been spying on us for 20 years right like you know that
42. 9:55right like everyone knows that yes is it worse because China's going to do something with it maybe I don't know
43. 10:02maybe if we become part of the social score of China at some point that would impact us so so if you wna you want to
44. 10:09head your bets for the future now um is it highly likely that if you go to chat. deep seek.com you should assume that nothing in there is
45. 10:21safe that that everything in there will likely be um you know popped into a uh a database sure you should assume that
46. 10:29that but you should also assume that anything that's [Music]
47. 10:44free can't sign up for deep seek been trying for a long time today and it keeps the gear keeps spinning that's
48. 10:50weird hang on I gotta move my Black Box um I just logged into deep seek that's Google studio that's that I'm in chat. deep
49. 11:05seek.com can't sign up yeah that's [Music] weird perplexity said deep seek downloads were
50. 11:20paused deep seek had an attack today ah so they are limiting new enrollments okay thank you all right so luckily
51. 11:30daddy AI learning lab got in there just in the nick of time let me see if it's actually working right now um tell me a bed time story
52. 11:48story yeah it's working okay by the way if you're worried about Ye Old Chinese servers spying on your [ __ ]
53. 12:00spin up a web-based GPU and install your own version of uh deep seek R1 there's all there's a bunch of different
54. 12:09sizes figure out how much you want to spend and just throw it in a hermetically sealed GPU instance in the
55. 12:15cloud and then none of your shit's going anywhere except to the to the GPU provider but generally they've got good
56. 12:22good things did this give us a good nighttime story the Allen the missing fireflies you know cozy fire forests nestled
57. 12:31between Rolling Hills where the moon painted silver Trails through the trees there Liv lived a wife Wise Old Al named
58. 12:38Oliver seems like a story all right there you go all right I all good people I downloaded deep seek off the App Store
59. 12:44L than an hour ago works fine okay yeah they've got an app and it's free and it's better than 01 or at least
60. 13:0001 preview probably maybe 01 I don't know if it's better than 01 screen share still off yeah that's okay I'm not I'm I
61. 13:08wasn't really sharing anything thank you let me share so I have carrying
62. 13:25on rather Cheerio pip pip you know we don't need that right now bless you oh bless me I thought you literally
63. 13:40literally meant like bless me like you were taking pity on me it was because I sneezed I'm I'm a little slow in the
64. 13:48uptake sometimes I don't know if you knew that H bless your heart a a't that cute he's on there trying to talk about the
65. 13:57AI with zero qualif ifications it's so sweet oh I'll tell you what I'll tell you what people I I'll tell you what I
66. 14:07do I go to that AI learning lab he he's just so sweet he's got a dog the dog bless his heart that little dog sing
67. 14:14sing sings now here's a funny thing and it's kind of sad I mean I feel for the gentleman um the dog sings better than he does and it's just
68. 14:26it's bless his heart bless his heart and then anyway but it's called the AI learning lab so you go there and what
69. 14:33he's talking about on the AI learning lab is it's not always AI right sometimes he I and he it's learning he talks
70. 14:45about he's he's not a teacher he's not a teacher he's not qualified he's he's he's got zero qualifications um uh he
71. 14:53has a degree in acting and apparently got it mail order a mail order acting degree apparently um but he he tries he
72. 15:01he's so sweet he's super sweet he shows up all the time and and it's like if if you're going to give him an a I'll tell
73. 15:08you what AI learning lab they deserve a participation trophy they really a certificate a participation certificate
74. 15:15yeah trophy probably probably not probably not if maybe if they made like a little like you know like a little
75. 15:21Yoda trophy that would be fine but just more more of a you can make money with so that's what we do [Laughter]
76. 15:35here a fashion award for the bow oh good lord um ocean Kyle you know we have different capabilities and interests when we use
77. 15:49Ai and GPT needs different ways however how it's going to be able to cooperate and stay with the human brain oh it's not
78. 15:59ocean here I'll leave that up there because in case you want to figure that out one of the one of the one of the
79. 16:07questions one of the questions uh the question here how are humans going to keep up with these things we're not we're not
80. 16:27and what while that is potentially caused for [Music] concern shy of all of this AI activity shutting down which it's not going to
81. 16:43these AI capabilities I you know I listen I would argue that for me personally my cognitive wall I hit when 01 preview came out back
82. 16:54in what was it September um and I realized that I didn't really know how to um how to prompt it that I
83. 17:08didn't in fact the exercise that we're going to do tonight we're going to do a head-to-head between three reasoning
84. 17:14engines and the prompt that we're going to use we we'll write it live here so you can see my prompting process which is
85. 17:21very methodical and it's it I just pull it out of my butt basically um [Music] um but we're going to do we're going to do a heads up
86. 17:35competition competition is not the right word um because when you say competition that implies there's winners and losers
87. 17:43and in this world we can have no winners and losers although tell that to the Buffalo [Music]
88. 17:57Bills we're going to have a competition between deep seek Gemini their reasoning engine flash thinking I think they call it and then uh
89. 18:0801 and maybe we'll throw 01 Pro in there since I'm paying for it the whole point of me paying all that money is to have
90. 18:15access to these better tools all right so we're gonna do that and the uh The Prompt that I'm going to put in there is how would someone with
91. 18:25my capabilities for lack of a better term how would someone with my skill set best leverage a reasoning engine and
92. 18:34let's let the reasoning engines tell us how to use the reasoning engines and then we'll compare the answers and just
93. 18:39see see see what the [ __ ] these things are and how they work and why we should give a [ __ ] they're they're fascinating they're
94. 18:46fascinating I've I've been using them more in the past week and again part of the one of the things that I've been
95. 18:53talking about on the channel for the past I don't know three weeks a month is and this gets back to the human the human thing what's our
96. 19:04job if we have tools that are significantly beyond our cognitive ability what's our job and I think our job is to a person
97. 19:18figure out what we want to do in the world what difference we want to make in the world because we wait hang on did you just call me yeah
98. 19:34[Music] thought I was being called thought I was being summoned have you been able to prompt operator to do anything
99. 19:47useful yeah I I mean it depends on your definition of useful I haven't I I saw a couple of videos of people um using
100. 19:55operator to go to um repet and generate apps which I thought was kind of interesting I saw a video today of
101. 20:04someone having it go to make and create an automation which I thought was kind of interesting um last night I did kind
102. 20:14of a deep research project that a uh a thing came out for for um the University of Colorado and shoots which is just
103. 20:24like you know 45 minutes that way they just did some research Arch that they have um they've got a new drug that
104. 20:32works in animal models to repair myin sheath neuropathy which is what my boys have um in the eyes of I don't know rats
105. 20:43that had whatever the disease was that caused the myin sheath blindness um and and it's made it's it's uh promising to repair my and sheath
106. 20:53it's like the covering of your nerves um and so I just did a deep Di on that so I had I had um I had operator go out and
107. 21:03do that research and I it was fascinating like just sitting there watching it do the research and what you realize and you'll
108. 21:10see this with deep seek tonight deep seek is connected to the internet like it's hard to put it's hard for my
109. 21:20tiny ape brain to comprehend but like it will just go find a website like that and then it goes to it and it's Reddit
110. 21:30it reads the whole website or it reads the whole paper and in the case of deep seek it's like okay I found 49 websites
111. 21:37and now I'm analyzing them and now out comes your answer so so these things are incredibly capable so so that was pretty [Music]
112. 21:49interesting but good question I I think I think operator is is going to force us to think about well again what do we want Toom
113. 21:59what do we want to do in the world um you know solving some shitty disease that your kids have or or finding at least treatments for it
114. 22:09that's not [Music] bad so we're going to do that we're also going to start with jelly pod if you haven't been here if you weren't here last
115. 22:22week we did this uh we did this wacky zany this wacky zany um podcast maker called jelly pod I'm with Pate I don't like 01
116. 22:36or for 40 for Co coding for myself what what do you use Joker do you use um uh what you call it Claude yeah I don't I kind of feel like
117. 22:49coding um I've been coding for two weeks my agent will do all my coding all of the okay if you're new here if
118. 23:01you're new to AI apps and and these things these AI these AI tools today to a tool are janky pieces of [ __ ]
119. 23:12now they are getting better and better and better and better but there's still a mess I'm in three different Creator
120. 23:21programs that I keep getting access to their latest models and they still cannot make a car that that's Wheels
121. 23:28roll like the car moves but the wheels are stuck they're just sliding so they're just everything's janky including the coding tools
122. 23:41remember when IBM Watson was going to revolutionize the world yeah but we're in a we're in a different we're in a
123. 23:48very very different place here that was you know very narrow Ai and it it was a more traditional approach to machine
124. 23:55learning we now have generally capable you know um general intelligence not not AGI not artificial general intelligence
125. 24:04the threshold we haven't met the threshold yet um but but these are generally intelligent tools right they're they're intelligent on lots of
126. 24:15different things and the reasoning engines can reason um beyond their trained capability right they have emergent properties so that's very
127. 24:26different and that was also IBM and it was also you know a single product a single division um this AI stuff is now you
128. 24:37know coming from every major tech company in the world Plus open [Music]
129. 24:52source um so what jelly pod is is this is this allows us to we designed these two voices Lily Vogue she's the one on
130. 25:04the right with the the bitching gray streak in her hair and then Buster mcgillicuty he's a little hipster dude
131. 25:11on the left he's stoned a lot um but there are hosts and then we went over to to Yan chat GPT and we had chat GPT write us some
132. 25:25episodes so so this was kind of a cool thing we we had it design the uh the hosts we had it design the the podcast
133. 25:36ideas and then Corey Sandler went off and made us episode covers so we're g to make one of these so let me go back to jelly
134. 25:50pod oh Brandon don't let me hit the submit button before I add the music because it's it's one of those counter
135. 25:58intive things where the submit button is big and orange and then the ad music is really tiny and gray and you can barely
136. 26:05see it um oh you want to hear one of these by the way here's one on the um the the uh episode one the premiere the
137. 26:13astonishing life of Mike the Headless Chicken oh wait you can't hear that oh God [Laughter] [Music]
138. 26:38the only podcast where absurd trivia and fascinating history collide with well a solid dose of charm wouldn't you say a
139. 26:46dose Lily I'd call it more of a heaping pile two charm heavyweights here I'm Buster your guide to all the weirdly
140. 26:55impossible stories you didn't know you needed so this is our little host all right so that's what that's what our our
141. 27:00podcast is going to be made out of so we're going to go new episode are are you viewing this yes you're viewing this
142. 27:06okay I've been using chat gbt over Google and any okay all right so we we're gonna go over I'm gonna go grab uh in chat GPT the episode
143. 27:25description and we'll go add source and then I'm go back over here and I'm going to grab the Wikipedia article about this
144. 27:37topic all right we're g to add that um or describe your episode uh I'm going to say this is episode 4 and it
145. 27:53is sponsored by the AI salon at the salon.
146. 28:10um do I need anything else other than that oh I'm adding that as a source okay fine copy that out of there all right so
147. 28:24create boom so now what it does is it's going to write us an outline which we're going to be able to augment which is super
148. 28:36sweet got into deep seek on Firefox on laptop okay great so yeah you might have had some sort of cookie some sort of
149. 28:45Bonkers cookie issue on on [Music] Chrome Valerie Cox what do she have to say while we're looking at that thing
150. 28:54I've been using chat GPT over Google on any questions that I have or on any explanation on a deep subject I have set up several tasks to
151. 29:07get me top news tell me a joke produce random pictures for fun yeah that's pretty cool are you you're using tasks for that all right that's
152. 29:17cool as an educator in Mar Meridius Marius I cannot believe how resistant they are to embrace in AI which is being something I assume you
153. 29:32know they're they're using AI detectors I it is staggering to me how far up their asses most educational institutions are right now with AI
154. 29:44you've got honest to God I mean just think about what it is it's it's one of the most profound knowledge augmentations in the
155. 29:53history no it is the most profound knowledge augmentation in the history of humanity and our educational
156. 30:00institutions are be like well we don't want people to cheat it's going to [ __ ] change everything it's going to change how our brains
157. 30:08work it it it's absolutely mind-blowing to me absolutely mind-blowing all right what the fact image archives number four
158. 30:18sneaker culture oh wait you can't see any of that because tabs morius mortius is a country in East Africa very cool
159. 30:30um I got to change this again hang on people the college kids can't even get a
160. 30:44smart laptop per instruction per instructions from their professors yeah well some of them are going to figure it
161. 30:51out anyway here's the Corey Sandler cover for what the what the fact from Humble Footwear to the hype machine so that's our
162. 31:01uh file too large all right we had this problem before I've got to I gotta go fix all these things probably before we go live shouldn't
163. 31:10[Laughter] I yeah just say yes 768 by 768 resample
164. 31:27the image let's do do it duplicate is Tiff sure fine whatever
165. 31:43[Music] export all right let's see if we got it here desktop we're going to work looks like we're going to work
166. 31:59all right we're going to add a chapter here it puts chapter the new chapter behind chapter one which is weird chapter title episode 4
167. 32:15intro and then I'm going to put the [ __ ] I wrote before this is episode four and sponsored by the AI
168. 32:23Salon I think that's all I need to put there am I missing anything Brandon I think that's it sure fine
169. 32:31whatever my response to my computer all the time exactly you have to have a sense of humor with these [ __ ] computers okay so now what we
170. 32:43do what we do I definitely know what to do here yeah okay there it is so now we're going to generate the script
171. 32:52bang and so it's now writing a script for us down here all this [ __ ] on the right where there's nothing there's going to
172. 33:02be something so it's going to take our little intro kicks the change the game Lily dramatically introduces sneaker cultures
173. 33:11Origins fun fact when I was doing agency.com we went out and we pitched Nike you know the sneaker company they make they make
174. 33:22kicks as the kids call them today oh yeah this is a free account that I'm on although he did give me some
175. 33:30extra credits so uh I couldn't be doing these on just a normal free account actually I don't know how many free
176. 33:35credit credits he gave me so we might not be able to to generate this we'll see um uh yeah we pitched Nike and they are a madly impressive
177. 33:50company to pitch because they've basically tried everything and the thing that you think is radical and Innovative
178. 33:57and marketing they're like yeah we tried that a decade ago that doesn't work um they're a little full of themselves but
179. 34:02they're [ __ ] Nike right but we so we didn't get the a account but no agency did what we did get was we got the
180. 34:10Jumpman 23 site so we actually built Michael Jordan's first website that was pretty slick okay um our thing has been
181. 34:18generated now we're gonna add intro music good catch there Brandon what generate audio first I did add intro music oh oh I got to
182. 34:35generate the audio here got it welcome to the Intriguing episode of what the fact where we uncover the stories behind things sneakers a
183. 34:45quick not quite Buster but thanks to the AI Salon this episode uh styish feet so let's Jump Right In um we'll say
184. 34:59check them out at the salon. [Music]
185. 35:14a by the way while we're talking about the salon. a go there if you've not gone there and check that [ __ ]
186. 35:25out that's the salon the salom bye Tony see you later Tony all right let's generate our audio Bang Bang audio generation started this
187. 35:41may take a few minutes all right then add intro music then publish beautiful beautiful I'll tell you another thing Tony hey Tony hey it's
188. 35:55Frank Frank how you doing I'll tell you what these these uh this this [ __ ] Chinese deep seek R1 comes in holy [ __ ]
189. 36:06comes in and knocks everyone off their [ __ ] Keers you know you know what I'm saying all the big AI companies are like
190. 36:12I just raised half a trillion dollars they're like hey I put together a couple of buckets and a string and I got a model better than yours
191. 36:25anyway also go check out are you ready for ai.com that's a course that uh I'm putting together with an Murphy and
192. 36:32Vicky Baptist uh to train people how to get ready for what's coming the fact of the matter is you can't get ready for
193. 36:40for what's coming we were going to create an AI um literacy course and then we realized you can't there's nothing to
194. 36:47master right now this stuff's changing too fast but what you can do is you can start to shift your mindset and so
195. 36:53that's the course we're putting together based on the um the the the really cool event that we had called AI festiv us so all right
196. 37:04let's go. [Music] MP3 uh where is
197. 37:19my huh let's go what the
198. 37:33name contains what the no what Dash the hang on what Dash Thee I just looked this up hang on a
199. 37:53sec champion no that's Mama what the fact there it is okay hang on
200. 38:10people boom all right there's our music welcome to another intriguing episode of what the
201. 38:24fact and then we're going to publish the episode should we hear it first how do you preview it can I preview this thing oh yeah the giant
202. 38:34play button oops sorry hang on go check out demystifying AI space in the salon for all the uploads including the
203. 38:42notebook LM podcast very cool thank you Pate we'll do that oh are you kidding me
204. 38:59hang on for people that don't know what my exasperation is about you have to change the way you're sharing your screen in
205. 39:12Chrome only because they're in some sort of [ __ ] Feud with Apple so if you're running a Chrome browser and you're
206. 39:19trying to share your audio you can only share it in a tab which is [ __ ] up if you have a a channel like chat add where
207. 39:27I'm flipping around between tabs like I I have memory
208. 39:43anyway welcome to another intriguing episode of what the fact sounds like doesn't it all right publish boom and we're done I think we'll see
209. 39:57once you publish episodes you won't be able to make changes you can subscribe on pocketcast yeah Pocket
210. 40:11Cast but wait so if I want to share this I click this yeah there's the episode share the episode and then we're going to go to oh
211. 40:20I gotta I got to unshare hang on dang it now I got to go back and reshare Tik Tock pin need more Cowell
212. 40:33yeah exactly it's definitely got some SNL Vibes to it that little that little songy song we'll do this with audio so I
213. 40:41can actually play this after we upload it black bar I know I'm I'm a professional I don't I don't need a
214. 40:54producer can you imagine to show with that one well yeah we can imagine it we experienced it for [ __ ] years it was
215. 41:07awful I would I would seriously sometimes talk for 45 minutes without sharing the screen I thought I was
216. 41:14sharing um deep seek R1 this is from Sam Alman an hour ago deep seek R1 can everyone see this yes is an impressive
217. 41:23model particularly around what they're able to deliver for the price why everyone's stock price tanked we will
218. 41:30obviously deliver much better models and it's also legit invigorating to have a new competitor we will pull up some
219. 41:38releases one of the things that I think all the naysayers I you know it is amazing to me how quickly um the tech bro culture turned
220. 41:51on all the big uh Frontier Model companies they're like oh deep seeks around you're all you're all worth L and
221. 41:57and you know out of out of business it's like does no one understand that [ __ ] changes so it's like the the
222. 42:07reason deep seek R1 is so good for so cheap is that we put an export restriction on the best AI chips so the
223. 42:19h100 chips the ones that all of our models were trained on we basically hobbled that thing and and gave China a
224. 42:27shitty version of it called the h800 the r800 whatever I don't know it's some different number it doesn't matter we
225. 42:34gave them a shittier chip now the rumor is that uh that deep seek actually has 50,000 h100s that they're not supposed
226. 42:43to have but they don't talk about that but we'll just we'll just let that rumor slide that's conspiracy theory because they had
227. 42:51shitty chips they completely rethought how they trained them and whatever they did to completely rethink how they
228. 42:59trained them made it really good really fast so you don't think that our Frontier companies are going to take that
229. 43:08methodology and adopt it and improve it and put it on the big ass things and then we'll have some crazy ass [ __ ]
230. 43:15GPT 17.4 of course they will anyway whatever just it's it's nuts okay so we are going to say um on this episode of what the
231. 43:35fact sneakers let's see sneakes sneakers on on this episode I guess we'll do four in parenthesis no
232. 43:52on how about here's episode four here's episode four of what the fact we drop that in there we'll go grab the
233. 44:09graphic what the fact can I add a link to this I don't think I can right no that's
234. 44:24stupid that's stupid Twitter hey Elon why don't you make your stupid sight so it's not so stupid all right thanks at jelly
235. 44:44POD at uh what was his name um Pier Pier Pier uh not Piers Morgan not Brock Pearson Pearson
236. 45:01Pearson Pearson something Pearson oh Pearson why do you have to have such a fancy name anybody remember
237. 45:13Pearson there's no guarantee deep seeks algorithm and training will lead to more powerful models it already has what do
238. 45:20you mean what do you mean there's no guarantee that that their algorithm and training will lead to more powerful
239. 45:26models they trained on on 2,000 gpus now according to them Pearson marks that's what it is marks um let's do atso because that was what
240. 45:42where the song was from uh let's tag Corey Sandler because she made us our swell graphics
241. 45:58Cory Sandler Pottery at Nike yeah why not at Nike we'll do at Jumpman I wonder if Jumpman 23 is still a
242. 46:13thing and no no and we'll do scobleizer not when you spell it that way SC Eliz
243. 46:35wait oh scoer okay all right there we go okay so here's my request for all you people out there we got 40 people over on the YouTube and
244. 46:49the X we have 68 people here on Tik Tok go here go to Twitter my account at Kyle Shannon and Pump Up episode 4 let's turn this into a hit
245. 47:04people well we don't even know if it's any good yeah know whatever oh tell me tell me also if that's still broken if you try to open
246. 47:15this thing on um on desktop at works and on mobile it didn't liked by Kelly Camp Kelly Camp gets the gold star she went first on x
247. 47:27she she liked this on X before you did I just got it on my phone got a notification if you want a gold star next
248. 47:44time okay beautiful beautiful fantastic Bob all right what are we gonna do I guess let's switch over to um so this Saturday I'm doing
249. 47:57um I'm doing an actual reading with actors of the musical Sydney that I'm working on so I thought maybe it might
250. 48:06be fun to play some songs from there but I don't have to we can jump right into maybe we do a little little poll on on
251. 48:13Tik Tok or something like that um or maybe maybe I'll just make a call as if I'm running the show but we can either we can listen to
252. 48:23some Sydney songs folks waiting for 01 fine we'll do reasoning F we'll do it live we'll do it live all right we'll do reasoning we will
253. 48:36do um maybe we'll maybe I'll do a Sydney thing Thursday night or maybe I'll I don't know when I'm GNA get
254. 48:47in yeah maybe I'll go live Friday night and I'll just do a special Sydney episode Friday night I think that's what I'll do all right
255. 48:58it's as if I don't know what I'm doing you're not sharing sound on YouTube oh God Jesus well that's okay we're not gonna
256. 49:11need to share sound for what we're about to do so let me go back to sharing everything yeah Friday night date night
257. 49:22with a soundtrack I think that's right it opened for me okay great fantastic yeah maybe it just didn't work that
258. 49:29first night because the the you know URL hadn't propagated or whatever the [ __ ] has to happen with
259. 49:35domains there okay there's that that was that comment okay so what we're GNA do now if you want to play along jump on over to chat. deep
260. 49:55seek.com and this is very much like chat GPT God I hate how this thing displays on Tik Tock there's just no way to
261. 50:13control the look and feel oh well what are you gonna do what are you gonna [ __ ] do more powerful means better reasoning
262. 50:26and more accurate results it has it is more powerful and it has more accurate results that's the that's the that's why
263. 50:33everyone's freaking out is that this thing is is really quite good okay so so what we'll do so so new app
264. 50:42profile am I logged in I must be where's my history though oh failed to load my history failed to load try reloading
265. 50:57shut up champy okay there's my history great all right all right fantastic they need a better UI yeah they're I mean their UI
266. 51:06is just like chat gpts they just stole it from chat GPT so so yes you're correct they need a new UI okay so here's what we're gonna do so
267. 51:18so so what we're looking at here we're playing with reasoning engines the difference between your chat GPT and a reasoning
268. 51:27engine uh is what's called Chain of Thought reasoning so when you type a prompt into chat GPT it does this probability calculator
269. 51:38thing where it's predicting the next token or the next batch of tokens and you can effectively translate that into
270. 51:44it's predicting the next word or the next series of words and so you put in a prompt and it basically answers you and it's got no
271. 51:55real idea of what it answered it's literally just spitting out fragments of words that it stitches together and then
272. 52:03we as humans read and go yes that's awesome or no that's completely wrong that's why we get things like
273. 52:09hallucinations right where it just does [ __ ] wrong reasoning engines rather than just vomiting out the stuff they have this
274. 52:19whole background conversation going on where you give it a goal and it understands that goal and then it creates a plan
275. 52:27right and it it and as it's creating the plan it's talking to itself and figuring [ __ ] out and and going step by step
276. 52:35through the process there there was a prompt for chat GPT think step by step will will improve a chat GPT prompt what
277. 52:42what that was effectively doing was telling chat GPT to act like a Chain of Thought Reasoner um they're just doing
278. 52:48it in a more sophisticated way okay sound is amazing over here versus Tik Tock great fantastic thank you Lord digital gods for
279. 52:59thees that ni I like it I like it a lot it's very nice okay so what we're gonna do the the one of the things that
280. 53:16um that emerged over the weekend for me is that the the reason people like deep seek so much other than the fact that it
281. 53:26was you know way cheaper than than Chad GPT um is that its personality is pretty good so when it like people are more excited about the
282. 53:40reasoning than they are necessarily about the answers and so so we're going to do that so what I'm going to do is
283. 53:46I'm going to come up with a single prompt and then we're going to apply that prompt in three different reasoning
284. 53:51models and then we'll look at the The Chain of Thought reasoning and then we'll look at get the answers and then
285. 53:57we'll just kind of compare them that's kind of the plan for tonight um with deep seek you can turn on reasoning with
286. 54:04with the left button you can turn on web search with the search button and so you can actually do a reasoning response
287. 54:13with web search which chat GPT doesn't have and I we'll find out I can't remember if if Gemini has it it may or
288. 54:20may not I can't remember but I'm I'm not gonna I'm not going to use web search I'm just going to use reasoning
289. 54:28oh yeah and there's there's a new one that came out um today called keimy 1.5 U let's not worry about that right now oh what I had a I had a
290. 54:43metaphor that popped up today like trying to keep track of all these things oh it was so when I got really into
291. 54:50blockchain and nfts I I did like a two-year Deep dive into that world and when I was trying to keep up with the
292. 54:57nft projects I started creating spreadsheets because I couldn't keep track of everything and that's where I
293. 55:03am right now with all these AI models with all the like the video tools and the image tools and now the reasoning
294. 55:09tools and who's got what features and wait was that feature on that tool or was it on that tool I feel like we're
295. 55:15just in this place where if you're trying to keep up with this stuff you you like you literally need AI to help
296. 55:21you make sense of it um YouTube question is deep seek safe well if you're on chat. deep seek.com you should assume that any data
297. 55:34you put in here is going straight into a Chinese database period end of story now what they did with this model
298. 55:46however is they they gave it to us in open source so if you go to I assume it's on GitHub or or on uh hugging face
299. 55:55it's on one of those two I assume it's on GitHub hang on I gotta do that um you can you can run these models
300. 56:07either locally on a machine they G they gave different Siz models from 1.5 billion parameters up to 70 billion
301. 56:15parameters and there's people right now that are putting that 1.5 billion parameter version of Deep seek on phones
302. 56:22they're running them locally on phones so if you're running it locally your data is not going anywhere so in that CA
303. 56:28you know in that in that sense it's safe within your hermetically sealed container where you're running it right
304. 56:35the most safe would be run it on a local computer or a local device the next safe would be spin up a GPU instance in the cloud that's
305. 56:45secure you know and then the least secure is go to chat. deep seek.com but if you're lazy like me or if you go download the iOS app
306. 56:55that's that's using their servers just assume that anything you put in here is not safe um but but you know is it is it
307. 57:04is it unsafe no it's not unsafe it's just assume that your data is gone being used elsewhere okay I'm gonna go to this thing and go [Music]
308. 57:18here all right and then I'm going to go g g g and I'm going to go uh this is the Deep seek prompt and let's change this to title as
309. 57:35well yeah so you can see it we'll call this we'll call this the reasoning prompt so I'm going to write one prompt for all of these
310. 57:50engines okay let's see I am I
311. 58:06am a a creative professional that does not
312. 58:23have PhD level problems to [Laughter] solve however comma I
313. 58:38want to take advantage of the new AI
314. 58:53reasoning engines that use Chain of Thought reasoning and
315. 59:08I'm hoping you can help here are some of my interests and
316. 59:23uh areas of aptitude if you're thought right now is couldn't you have done this
317. 59:38before um good point yes yes this must be insufferable to watch but that's
318. 59:53okay because I'm I I think this is well I I'm hoping that this is valuable because we're all going to have to kind of muddle through this
319. 1:00:05like we've never had I don't I have never personally had like a 30 PhD buddy right so I don't know what to ask
320. 1:00:15a 30 PhD buddy the closest thing I have is pate and when he answers me I'm like I don't know what he's saying and then
321. 1:00:21he's like all right I'll dumb it down for you he does it pretty good actually um okay let's make this a little bigger so you can see it
322. 1:00:40okay let me Let Me Entertain [Music] You all right there we go beautiful fantastic blob okay
323. 1:00:56Tik Tock question Kyle Irregulars Irregulars post of my random words as a prompt in deep seek okay cool hey I'm a college student trying to
324. 1:01:08be a software engineer remember to always check the sources yes I agree um always check the sources I saw a Tik Tok video on that today that
325. 1:01:20someone put a legal brief in and had it write a summary of the legal brief and then um chat GPT got them all wrong and then
326. 1:01:29basically just she she called it it quiet quit on her where it said it was trying but it just stopped trying uh and
327. 1:01:35then deep seek um actually got all the the summaries correct so um so it this might be better for summarization but
328. 1:01:43anyway okay so here's my some of my areas of aptitude um um creative writing for theater um fiction
329. 1:02:01marketing um uh let's see um performance and acting um
330. 1:02:17[Music] entrepreneurship and being a Founder um first online
331. 1:02:32magazine first digital agency um patented video uh
332. 1:02:47Auto mated video platform uh uh uh Community Builder um
333. 1:03:01World [Music] Builder worldwide web artists consort 1994 and AI Salon the
334. 1:03:18salon. 2022 um educator Tik tock.com
335. 1:03:34slash AI learning lab so what I'm doing here is just trying to give it who I am what I'm you know the kinds of things that I do and
336. 1:03:46know how to do um what else is on this list um leadership and uh people management
337. 1:04:00consulting um comedy jokes oh musician
338. 1:04:16um you should ask chat gp01 mini to write this bio yeah I could do that well that's okay all right so that that's
339. 1:04:25enough that's enough kind of stuff there okay so now let me go back here and make this big again for you so I'm going to
340. 1:04:37say I want you to think long and hard about the kinds of things I could use a
341. 1:04:54reasoning engine four and break the use cases
342. 1:05:14into pink bow owner um break the use cases into um categories
343. 1:05:30you think will be Val Val valuable and then in each category
344. 1:05:45provide a provide three use cases and then we'll say
345. 1:06:01basic intermediate and advanced and describe the use
346. 1:06:15case why a reasoning let's see um why it makes sense and then
347. 1:06:35include the prompt for the prompt to use in a reasoning
348. 1:06:51engine okay good Lord Okay so I basically say I don't I don't have the training or or the intellectual capacity to really know
349. 1:07:06how to use this [ __ ] tool that's in front of me however here's some [ __ ] I've done right so you could do this you make your own list of
350. 1:07:14[ __ ] and then basically I want you to think of ways that I could use a reasoning engine right does that make
351. 1:07:21sense this is amazing on my huge TV that's terrifying but but awesome okay what a great idea Valerie what did
352. 1:07:28Valerie say oh I watch him on TV on YouTube and chat on my phone oh that's yeah the dual screen thing people do
353. 1:07:34that a lot we're all here for the fun that's great that's really cool Tams is loving it all right rock on Dax is
354. 1:07:42loving it cool all right Pate says he has some PhD problems you can borrow yeah well actually you know maybe that's
355. 1:07:49something we do here one night Pate is because I got 30 days of deep Dives maybe actually Pate you know what could
356. 1:07:55be kind of fun is you come up with a uh you know PhD problem or two and then maybe I'll bring you up on stage and
357. 1:08:04we'll we'll just dick around in these reasoning engines and see if we can get them to do anything interesting um Source Camp my prompts
358. 1:08:12include some form of I don't know what I'm doing yeah ex great oh that's perfect I'll do that um you know uh wait
359. 1:08:21hang on let me make it a title so you can see it uh as reminder uh I'm a bit of a dumb [Laughter]
360. 1:08:37dumb but um I I'm really sweet and well-meaning that's gonna become part of our prompt all right so we're gonna copy this whole
361. 1:08:52thing pat down for whatever great cuno impostor prompting yeah exactly exactly all right so here's my impostor
362. 1:09:04prompt and and listen I mean I'm doing this a bit tongue and cheek right you know that that that I'm a bit of a dumb
363. 1:09:11Dum and and you know I'm not really smart enough to use this thing but but to I I I mean the point still stands
364. 1:09:19just in my life you know I don't have a lot of people presenting me with with big hairy problems to solve um or or
365. 1:09:30like the ones that like like when I wanted to come up with what storyvine has become and what that technology has
366. 1:09:36become you know that was two or three years of solving a problem I could now probably take the problem that I solved
367. 1:09:43with storyvine and I probably could have solved it you know in a [ __ ] week right that's that's the difference for
368. 1:09:49where we are right now but knowing having the problems is part of the challenge so let's go use these tools all right I'm an
369. 1:10:25problem hang on hold please I'm like that doesn't make any sense that you can't paste it in a long prompt boom okay so there we go I'm I'm
370. 1:10:39a creative professional that does not have PhD level problems to solve however there's what I do I want you to think long and
371. 1:10:47hard all right and then we're going to go make that bigger put this in here so you can see it error between seat and keyboard
372. 1:11:02chat huh in the prompt never mind chat okay
373. 1:11:17whatever I'm using that I'm Kelly and I'm anomor um all right so here we go bang thinking oh it's slow everyone's using
374. 1:11:33the [ __ ] out of this thing this was much faster earlier okay so the first thing you so what it's giving you right now so
375. 1:11:39this little line down the side says that it's giving you its thinking so it's now it's thinking so this is the what we're
376. 1:11:47getting to witness here is the Chain of Thought reasoning so this is its conversation with itself what's cool about it is it's very
377. 1:11:56conversational okay let's start by understanding the users's background and needs they're creative professional with
378. 1:12:00experience in writing performance entrepreneurship Community Building education and more they want to use AI
379. 1:12:05reasoning agents but don't have PhD level problems great first I'll break down their listed interests into
380. 1:12:11categories the user mentioned creative writing performance entrepreneurship Community Building education right gives
381. 1:12:17me the long list these use cases each each of these should be a category for each category I I need to so so look at
382. 1:12:25this for each category I need to think of basic intermediate and advanced level use cases so it took that basic
383. 1:12:33instruction I did and it kind of internalized it right so it's responding as if it's a just like a a person right
384. 1:12:41let's play with Janice Janice Pro images uh yeah I would love to do that actually um ramik do you know what uh
385. 1:12:49youl like where where are they hosted I assume they're not on on chat. deep seek.com yet but if they are let me
386. 1:12:57know um okay starting with creative writing um it's coming up with ideas next is performance or acting it's coming up with ideas here's
387. 1:13:07entrepreneurship ideas here's community building ideas here's education ideas right so it's giving itself all the
388. 1:13:13ideas then it says each use case needs to prompt needs needs a prompt that guides the a AI through the reasoning
389. 1:13:20Pro process the prompt should be specific and trigger a Chain of Thought right so it's so then now here's the
390. 1:13:30answer so now we get it in bright white right so the thinking is in this gray with the little line and then here's the
391. 1:13:36answer let's break this into categories aligned with your skills and use cases basic to Advanced and tailored prompts
392. 1:13:43for reasoning engines like chat gbt gemini or Claude these prompts will then guide the AI to think step by step okay
393. 1:13:50great well we don't need it to add the think step by step all right but let's see what it did creative writing basic use case
394. 1:14:00brainstorming unique story premises why overcome writer block The Prompt act as a creative writing partner I need five
395. 1:14:08story premises that blend genre theme for each premise suggest a protagonist flaw and a symbolic okay so that's that intermediate character
396. 1:14:20motivation Mapp mapping deep charact deep deepen character arcs by forcing contradictions all right these prompts
397. 1:14:29seem a little thin I'm going to I'm going to tell it that I'm going to say um your prompts seem a little
398. 1:14:44thin and lack specificity um expand on each use
399. 1:14:58case with more real world examples to inspire me to think outside the box or whatever that [ __ ]
400. 1:15:16[Music] means all right let's go back up here while it's while it's thinking we see see some of these other ideas as it gave
401. 1:15:25us so there was creative writing um Advanced meta narrative structure design experiment with nonlinear storytelling that's pretty
402. 1:15:33cool that's actually a really good one for a reasoning engine right entrepreneurship and Consulting audience paino generator pretty
403. 1:15:42cool business model Paradox resolver tackle contradictory goals scale fast but stay Niche cool anti-brand strategy
404. 1:15:52all right now oh it's thinking again all right let's see let's see what it does okay let me try to figure out how
405. 1:16:00to approach this the user found my previous prompts a bit thin and lacking specificity they want more real world
406. 1:16:08examples to inspire them to think outside the box so I need to expand each use case it's fascinating like how this
407. 1:16:15thing interprets my prompt and then expands it into what it's going to do right so this is now a plan but it's
408. 1:16:21kind of internalized it like it's my job to help this guy all right sidebar question on Tik Tock um I want to train
409. 1:16:29my chat gbt but my son uses the same account must I pay for a new a new thing um yeah you do you can get a team account
410. 1:16:41so that you you he's got access and but that I don't I just yes you do talk to Cindy [ __ ] on here or on the on the AI Salon she's got
411. 1:16:55this happened with her and her daughters her daughters were were like Mom would you shut up about the AI we're we hate the AI the AI is
412. 1:17:03stupid and then she had them do something where it solved a really cool problem for them and they're like oh my
413. 1:17:09God thei so cool we want our own and so so I think she's a three chat GPT house at this point so yes that's a pain in the ass
414. 1:17:22you could just have him work on on deep seek deep seek is free and it's better than chat GPT it's some of the basic
415. 1:17:30[ __ ] it's also Wicked slow right now Wicked slow this this was not this slow earlier first let's look at the categories they
416. 1:17:41mentioned all right so it's still thing okay so let's go on so now let me go let me go blow your mind with something
417. 1:17:47we're gonna go back to get the prompt so we'll copy the prompt from here while that works we're going to go to grock
418. 1:17:56um and grock is not grock is not Elon Musk grock but this is grq so how is this prompt different than a chat gbt the The Prompt is not
419. 1:18:10different how a reasoning engine responds to The Prompt is different what I'm trying to get is not so much the
420. 1:18:17prompt I don't really give too much of a [ __ ] about the the the phrasing of the prompting however I I wanted it to be a
421. 1:18:24little better than it it gave me but I I want to come up with use cases just thinking of like what is a thing I could
422. 1:18:31have a reasoning engine do versus normal chat GPT part of my challenge is normal chat GPT is already well
423. 1:18:42beyond like it can solve almost anything I throw at it right so my question is why am I using a reasoning engine on top of
424. 1:18:51that all right Janice Pros on hugging f okay great we can go play with that there yeah so so um deep seek R1 is the reasoning
425. 1:19:03engine and today they came out with a multimodal um also open- Source engine um that can make images so so just like
426. 1:19:13DOI makes images it's just like that only it makes images so all right Tik Tock pin might want to engage all
427. 1:19:20right no it's not it's dangerous it made fake FDA letters on its own you can get AI to make anything you can you can get
428. 1:19:29AI to make fake FDA letters you can get it to make you know real well researched ones we're still at a point with
429. 1:19:38AI that we need humans in the loop right and AI is like any other tool you could make a fake a you could make a
430. 1:19:49fake FDA letter with Microsoft Word and a fake FDA logo with Photoshop right the the tools are the tools now I'll give
431. 1:19:59you that these are much more powerful tools and people that don't understand them and don't understand they might
432. 1:20:05hallucinate they might inadvertently do stupid [ __ ] but that's on us as people it's on us as people to to understand
433. 1:20:15one of the reasons I go live five nights a [ __ ] week to teach this stuff is to encourage everyone to play with these
434. 1:20:25tools not to believe in them like they're you know [ __ ] you know the savior of of of humanity but to play
435. 1:20:33with them because they're not going away and they're getting more and more powerful and it's our job in my opinion
436. 1:20:41to to understand what they're capable of and where the boundaries are and right now the boundaries are are relatively
437. 1:20:50inclose like they're within they're within grasp of us right so when it does research of a legal brief and it summarizes it and it puts
438. 1:21:00citations in there that are not correct that's on the person that made that summary to check that [ __ ] there will be a point at which
439. 1:21:09these things are so good that they won't make those mistakes but we're not there right now but yeah any tool can be used
440. 1:21:15for any any bad because it can be used for evil doesn't mean that it's inherently evil just doesn't okay I ask for practical non-digital use
441. 1:21:25cases that's great yes but it said it hacked itself and was 18% its previous version it also made up I cracked Claud yeah if listen if you
442. 1:21:40want to go find examples where it did shitty things and it did stuff that's scary you can do that I I don't I don't understand the
443. 1:21:49point of your thing like should no one use these tools because they did some things that were kind of scary that kind of felt like a sci-fi
444. 1:21:58movie or do we want to learn that they can do that and work with companies to say let's make them more safe so they
445. 1:22:06don't do that right the whole purpose of putting these tools in the hands of people is to understand what these
446. 1:22:13things are and what our responsibility is and putting them out there all right archetypal architect on YouTube oh
447. 1:22:21that's my fault I had it thinking about a prompt that would really check challenge it and then no reason on that prompt all right
448. 1:22:28Sky Timbo um I've even reached a factual error with notebook LM even though it only responds to its uploaded docks yeah
449. 1:22:37these tools are jankier than [ __ ] just like again these are not all good they're not all bad they are what they are they are
450. 1:22:48algorithmic knowledge engines what I would argue is that AI in its current state what these things really do is
451. 1:23:00they take any idea that you have and they amplify it right they allow you to get it out of your head and get it into
452. 1:23:06the world and you can get any D any idea in your head out in the world to about an 80% complete level not much more than
453. 1:23:16that that 20% is really [ __ ] important right right now that's on us as these models get better and better
454. 1:23:25that 20% goes to 18% then 15 then 10 then five three years from now you're going to be able to put things out of
455. 1:23:33these things that you know are very very close to final if not final but we're not there right now anyway all
456. 1:23:40right g g g so what I want to show you here is grock gro.com gq.com is a hardware company and they create
457. 1:23:53these special specialized inference chips and inference is the moment when you give it a prompt and it gives you
458. 1:23:59the answer that's inference that's the the the moment of inference and so these chips are hyp specialized for just text
459. 1:24:07inferencing text large language model inferencing so you're going to see you saw how slow um chat. deep seek.com was
460. 1:24:17watch how fast this is so it's still the same model deep seek R1 but it's running on this faster Hardware so there's myig ass
461. 1:24:26um apologetic prompt and watch this bang like it's just wow it's thinking deep holy
462. 1:24:43[ __ ] like it's all done okay so I'm trying to figure out how to use I use AI reasoning engin for my creative projects first let's look
463. 1:24:57into categories starting with creative writing okay so it did all the different categories so yeah it approached it the same that makes
464. 1:25:04sense creative writing prompt analyze so it's it's still given us short little prompts theme suggestion timeline
465. 1:25:14consistency create a timeline of events based on my plot ensuring oh that's kind of cool so you could give you could give
466. 1:25:22an outline of a story and it'll create a timeline that's kind of cool performance and acting character dialogue analysis
467. 1:25:30that's that's although that's regular chat GPT could do that physical mannerisms this is pretty this is pretty
468. 1:25:37shitty it looks like they've got um like the the amount of token outputs for this really low um so while it's fast it's
469. 1:25:46very thin so I don't like that all right let's go to Gemini okay so where I am right now is I'm at AI stud i.g
470. 1:25:57google.com and there's a lot going on on screen here on the left hand side is so so this is if if you've ever used um
471. 1:26:05open AI playground where you can design your prompts for applications You're Building you've got things like
472. 1:26:12temperature and you've got you know some more uh input options like one of the things that we can do here is I can
473. 1:26:20actually put in a system prompt so I could say that I want I want this uh this uh response to only respond in you know
474. 1:26:31in comedy or or whatever right so it's faster I thought open AI intentionally slow rolled responses so that it felt more human
475. 1:26:43um I don't I don't know what that's in reference to John um I think that um open AI slow rolled oper operator um just so that people could
476. 1:26:57keep up with it chat GPT is free but China based if that matters it's cheaper faster stronger oh that was to the tech text
477. 1:27:08response well what I was just demoing over on gro.com this is a this is a hardware company demo you just happen to be able to you
478. 1:27:19you can choose a bunch of different models um and I just chose they they put deep seek R1 on here so you can just
479. 1:27:27compare the speed of this output with the speed of of uh of the regular chat why not use beam compare all the all the
480. 1:27:37llms in seconds using the same prompt because I'm a dumb dumb we established that earlier um um okay so in in Gemini
481. 1:27:55over here on the right is where you choose your um your models there's a bunch of different models here you
482. 1:28:04know oh this is good one thing that they've done well here is for each of the different models they now give you
483. 1:28:11what it's good for and what its cost is and so we're going to go to flash uh thinking experimental so this is their
484. 1:28:18reasoning engine this is Gemini's reasoning engine so we're going to take our big ass prompt is this might
485. 1:28:28no we go back over here we'll copy this we'll go back over here we'll paste [Music] this all right there's our prompt
486. 1:28:44run all right thinking expand of view thoughts all right so here's our thoughts so remember R1 was pretty um
487. 1:28:55good question on Tik Tok oh my God how many Tabs are open right now uh I'm going to guess 84 oh that's not the
488. 1:29:10question the question pinned oh Tik Tock what is the point of open AI now so your question GR grat crack grat
489. 1:29:26crack you um your question is why the NASDAQ tank today about a year ago maybe a year and three months ago there was a Google
490. 1:29:38engineer that I think left Google no no I think he was still at Google and he wrote it I don't know if he's still
491. 1:29:45there P might know um he wrote a piece called there is no moat and what he basically talked about is that the cost of intelligence is
492. 1:29:54going to Trend to zero and everyone will be able to replicate what everyone else is doing and the fact that this Chinese
493. 1:30:02company created a state-of-the-art reasoning model that beats open Ai and you know the the the other things that are out there
494. 1:30:13um for a fraction of the price is what sent Nvidia stock you know tanking and Oracle stock tanking and the NASDAQ tanking
495. 1:30:26again here here's where I go while it's possible that creating these powerful models is trivially easy it is probably the case
496. 1:30:41that having state-of-the-art models is going to require massive amounts of infrastructure if it's not then the the
497. 1:30:48drop in price makes sense um just because a model came out that caught up and and surpassed other people
498. 1:30:58for way way cheaper all that's going to do Sam Alman said it an hour ago that's a kick in the ass for them to say we
499. 1:31:05better get our [ __ ] together and re reestablish our Supremacy right so like this ain't over like I said these tools
500. 1:31:14right now as impressive as they are are not ready for prime time like the fact that I'm having to go to three different sites to compare
501. 1:31:24three different reasoning engines to see what is what is it that they actually do and are they any good like at some point
502. 1:31:31in the future we're not going to have to choose models you're just going to ask the the the chat hole as we as we
503. 1:31:38defined it on Friday you just ask the chat hole your question or give it your goal and this shit'll go out and it'll
504. 1:31:46figure out what models it needs to use and it'll go do all your [ __ ] for you we're not there yet the downside of being this early is
505. 1:31:54you have to be a little geeky you have to be hyper curious you have to be hyper adaptable like I could teach you
506. 1:32:00something tonight here's exactly how to use this reasoning engine and then deep seek comes along three days later and
507. 1:32:07completely blows that out of the water and now you got to go learn that that's just where we are right now the upside
508. 1:32:13of being this early is that the amount of opportunity for people that are this early is massive because most people are ignoring this [ __ ]
509. 1:32:24like even if you're in here like the person that was in here that was was saying these things are bad these things
510. 1:32:28are evil at least you're [ __ ] interested in them at least you're willing to be in the dialogue most
511. 1:32:33people are like ah whatever hey I yeah I heard it I heard about it yeah whatever I used chat GPT once I didn't like
512. 1:32:41it that's most of the world right now it's it's it's amazing to me all right is training a new model right
513. 1:32:54or should I work on existing models that's an interesting question it depends on what you want to do like if
514. 1:33:04you're 100 folks in the live hey everybody if you're new here welcome my name is Kyle Shannon this day I learning
515. 1:33:10lab the qu the question is should I train a new model or fine-tune a model or should I work with existing models it depends what what you
516. 1:33:21do like if I'm working on my my musical I don't need to train a model current models are fine if I'm trying to cure some genetic disease I
517. 1:33:35might want to do a specialized [Laughter] model um aios in its AOL disc era yeah yeah a or AI is in its AOL disc yeah
518. 1:33:52right now metaphorically speaking if if you were to compare where we are with AI to where we were with the worldwide web because I
519. 1:34:02was there for that like I know what phases we went through a lot of people are saying oh this is like the the the
520. 1:34:09dotc bubble we're not even [ __ ] close to the.com bubble that was the.com bubble happened in 2000 um and it was clear that it was
521. 1:34:18going to happen in 1999 we're way back in like 95 and a half maybe 996 we're very early right early in the web you had to
522. 1:34:32teach yourself HTML you had to we had dial up modems you didn't even have the internet all the time all all that sort of stuff
523. 1:34:39so you know we we're not even close we're so early we're so early no it's it's because these tools are so capable
524. 1:34:47it doesn't feel like that but we ain't seen nothing yet question human creativity still priceless but not all
525. 1:34:54AI is ideal like being a detective you have to see the clues that's correct and I think you to your point human
526. 1:35:02creativity intuition critical thinking um um discernment and judgment having ideas in the first place again I think
527. 1:35:13that our job right now our job moving into the future is very much going to be understanding within ourselves what do we want to do in the
528. 1:35:25world what do we want to do how do we want to impact people because these tools are going to help us do
529. 1:35:33that but you have to have the question you have to have the the imagination you have to have the creativity to go
530. 1:35:43oh I I mean here's by the end of 2025 I would I would conservatively say by the end of 2025 if you're paying
531. 1:35:54attention to this stuff and you know where to look and you know which tools to use every person that is in that hyper
532. 1:36:00curious hyper adaptable state for for 2025 by the end of 2025 we're all going to have at a minimum 10 PhD level
533. 1:36:16employees that can be PhD level at anything we want them to be so I would start asking yourself right now kind of
534. 1:36:23why I wrote this question the way I did if you could have 10 PhD level employees that will work for you
535. 1:36:33247 and do anything you want them to do what would you have them do that's not an AI problem that's a human
536. 1:36:43problem and then you as the one that have that Vision have that idea you sort of put your employees out to go do the
537. 1:36:50[ __ ] for you and they come back and then it's your job to go did they get it right did they get it wrong is it on track is it off track so
538. 1:36:59just like you would run a company you're going to run your own little mini company you're GNA have 10 employees
539. 1:37:05maybe you'll have five maybe you'll have a hundred maybe you'll have a thousand I don't know but let's assume you'll have
540. 1:37:1310 can you think right now if you had 10 PhD level employees that could execute at a worldclass level what would you do
541. 1:37:22with them and if you don't have an answer for that [ __ ] start thinking about it so anyway that's that's what
542. 1:37:29this prompt is new here watch the backlog of AI content on YouTube oh oh if you're new here yeah you can go to uh
543. 1:37:38YouTube uh and the account is learninglab D aai and you can watch all the back things thank you [Music] Brandon gotta have the uh the AWA
544. 1:37:55okay so here's Gemini's response to our prompt um understand the core the users core need user wants to leverage AI
545. 1:38:03reasoning analyze so this is a much more cold response this is is this all the thinking stuff oh no here's there's my prompt oh here's the
546. 1:38:14thoughts okay here's the thoughts so understand the users core thing okay wait I got to challenge this SEL AI
547. 1:38:30can't think can't think of new things it works on what already exists my question to you is how of have
548. 1:38:41have you used it how often have you used it because that's a Trope that I actually don't agree with technically you're correct technically
549. 1:38:54there is some Big Blob of data so it's been trained on what's known right and you put in a prompt and it
550. 1:39:04puts out a probability response for a token and then it does that a bunch of times in a row right so in
551. 1:39:13theory since it was trained on [ __ ] all it can do is spit out the same [ __ ] it has but here's where I think you're missing
552. 1:39:21things you're you're missing the the main point of how these things work they're not copying and pasting paragraphs from existing
553. 1:39:33documents it's not how it works the original documents when they get embedded into an AI system cease to
554. 1:39:41exist the words themselves get broken into tokens and clustered in semantic clouds in Thousand di dimensional
555. 1:39:50mathematical space Sounds like I'm making that [ __ ] up I'm not that's how it works so you have these fragments of
556. 1:39:58words living in semantic clusters in Thousand dimensional space and when you put in a prompt it's pulling little
557. 1:40:05fragments of meaning out of essentially the collective intelligence of everything that's come before us how does the human brain work well
558. 1:40:18you grow up your parents tell you some [ __ ] you go to school they tell you some [ __ ] you go to a Library you read some
559. 1:40:24[ __ ] you love reading books all that [ __ ] goes into your head so what's happening in there all of the original Source
560. 1:40:34material ceases to exist it gets embedded in little electrical fragments in your brain and then when you want to solve a
561. 1:40:45problem you have an input you put a prompt in there gee I wonder what I could do to solve this problem
562. 1:40:54and how does your brain work it goes It goes into you know its million dimensional semantic clusters of neurons and
563. 1:41:04goes and it goes oh okay that's kind of like that Tennessee Williams play I read in college and that's kind of like this
564. 1:41:10and it's and you're taking fragments of ideas and you're smashing them together and you go aha Eureka and the Eureka is where you smash
565. 1:41:18together those ideas into something new the G and G PPT stands for generative it's not copying and pasting
566. 1:41:27it's not just replicating what's already there it's smashing together these tokens and I would argue that it can
567. 1:41:35create something new just like the brain can create something new if you prompt it correctly so if you don't use AI
568. 1:41:44regularly I wouldn't make statements like that because I just don't think they're true like my experience of how
569. 1:41:50these things work is it is simply not not the case that they're not creating new and novel ideas by Smashing other
570. 1:41:58Concepts together which is what we do in our large language models all right okay um so it looks like Gemini's um Chain of Thought reasoning
571. 1:42:10is more logical which makes sense to me than deep seeks um add explanations ref fine okay okay and then here's the response
572. 1:42:24oh my gosh you sound absolutely delightful and Brilliant don't you ever call yourself a dumb Dum you're clearly a
573. 1:42:33Dynamo look at Gemini coming in with the flattery right off the bat I love that that's genius oh man it's all just mathematical
574. 1:42:46formulas and using python wait it's all just mathematical formulas using Python and machine learning framework but Rambo
575. 1:42:57yes okay the calculator is what you just said the probability calculator that is a large language model how did you put it is
576. 1:43:11just so notice your language there you're diminishing the value of AI by saying it's just a mathematical formula using python
577. 1:43:23machine learning Frameworks and algorithms hang on what your statement's not taking into account is what it's been trained on the
578. 1:43:37p in GPT the pre-trained part that calculator that cold unfeeling probab or yeah probabilistic probability generating
579. 1:43:50calculator that can't think and can't feel has been trained on all of the output or as much as we could get our hands on of
580. 1:44:01humanity and so what it's mashing up is the knowledge of humanity so why it feels like you're collaborating with a person is that
581. 1:44:12you're collaborating with all of them this is just the mechanism your logic applied to our brains would be it's just a pile
582. 1:44:25of you know neural cells that that use you know chemical signaling devices to generate electrical signals that's all our brain
583. 1:44:41is well why is our brain so magical because of what we put in it and what we retain and what we match together and
584. 1:44:49what we choose to share with the world it is very very similar to the way our brain works and it's not by it's not an
585. 1:44:57accident that that's how these things work they set out 70 years ago to design neural networks to mimic how our brain is
586. 1:45:07structured so yes the calculator is this cold shitty ass python thing that's not what makes it remarkable that's what makes it possible
587. 1:45:19which is [ __ ] remarkable if you understand how it works it's like a miracle but as Pate says it's not magic
588. 1:45:27it's math magic right it's math magical what makes it special is the data it's us all right off my soap
589. 1:45:38box I just tried a few prompts in Janice 7 it's a mess love deep seek otherwise great okay um it's a bunch of random
590. 1:45:48words that have been signed tokens and weights yeah well so's our brain um I'm not here every day but when I
591. 1:45:55click on it I learn so much I play with AI every day at work and at home that's awesome that the whole point of this
592. 1:46:02channel is I just want people to play like like what I I mean sure it's taking me an hour to do one little exercise
593. 1:46:09here but that's kind of the way it's kind of the way this channel works it's Chad ADD baby as soon as we figure out the human
594. 1:46:20brain all right it's not using real re reasoning ability it's all skills task based and not having thinking AI I don't know what
595. 1:46:33you're saying Rambo uh anyway but yes you can listen listen you can [ __ ] all over the technology you want but if you're not
596. 1:46:41using it I would encourage you to get really curious about it and start using it because it sounds like you've got
597. 1:46:50good critical thinking skills but you're Ling them to how you think it works versus how it actually works I could be
598. 1:46:57really wrong you might be a mathematician and doing this stuff and you think these things are useless like Gary Marcus that's
599. 1:47:05fine I don't like what Gary Marcus does but I at least respect him he at least understands what's going on with these
600. 1:47:11tools my whole point is if you're on the sidelines just putting out tropes about oh it's just a probabilistic it's it's a stochastic
601. 1:47:22parent it's a this it's a that have it go write you a [ __ ] application it's insane anyway all right so what did this come up with let's
602. 1:47:31break down how you can use these agents I've thought really hard which it which it italicized so here's what I find interesting deep
603. 1:47:44seek thinking was more creative but its output was more boring and then Gemini's thinking was boring but its output is really
604. 1:48:01creative I find that fasc so this is one of the things that I've I've I've also been thinking about is um I think we're
605. 1:48:08going to choose the tools we use for their personalities more than their capabilities because the capabilities
606. 1:48:15are already quickly exceeding our capacity to know what the [ __ ] to do with them but their personalities are
607. 1:48:21like oh I like how this one talks to me um I think that's going to be a big interesting Factor all right a fun
608. 1:48:29prompt is to ask AI to write a series of questions that will challenge it oh I love that that's actually really
609. 1:48:35cool in fact you know what would be a what would be a really good exercise is to grab a bunch of the tropes that people put out there about
610. 1:48:44why not to use AI it hallucinates it's a stochastic parrot it's just a mathematical algorithm that uses python
611. 1:48:51programing take all of those things and put them in a long list and then have ai maybe do a panel discussion
612. 1:49:00of you know of arguing pro and pro and con against all those all those things it would be pretty good at that that
613. 1:49:07would be that would be good use of a reasoning engine okay um basic use case find a new angle for a theater
614. 1:49:14piece I'm writing a theater piece about loneliness in the digital age which you can swap that out oh that's cool that it it um colorcoded
615. 1:49:24this do you have do do you think we will get voice for R1 um apparently they've got a voice model that is um what
616. 1:49:34they've what the rumor is right now is that their voice model is better than advanced voice from open Ai and then there's also a rumor that
617. 1:49:44they've got a an an operator type uh agent that will use the the web um just like operator does that's apparently better than open
618. 1:49:55AI I haven't heard any confirmation on those so it's possible that they're just putting out misinformation to to keep
619. 1:50:03driving the stock prices down but it's possible like I heard we're going to get the voice thing in the next day or two
620. 1:50:09and then if the other thing exists maybe we'll have five components so we'll have a basic chat engine the reasoning engine
621. 1:50:17the multimodal engine the advanced voice engine and the autonomous agents mention if we get those five models from Deep seek and and all
622. 1:50:31five of them are better than the frontier companies Then I then I think we're in it's a whole different
623. 1:50:38game but we'll see I don't [ __ ] know uh good question um I am I am really liking a lot the Gemini response here
624. 1:50:50all right so let's go do the same thing in in 01 in uh chat GPT so again we're going to go back and get our
625. 1:50:59prompt we're going to go to chat GPT we're going to go to 01 for advanced reasoning and I've because I've got
626. 1:51:07access to 01 Pro we'll do 01 first because everyone's got access to that but we'll also do this in 01 Pro and see
627. 1:51:13how different it is Teton skier wants to go on an LA LSD trip with AI I'll tell you what dude you know where what would be the
628. 1:51:23hang on let's do a chat add moment um What's this called uh Deco here app. deoh here.com yeah or app. deoh here. so go to app. deoh
629. 1:51:37[Laughter] here. world's fastest AI generator and then all right let me CR my claim my two free credits and then and then you're going
630. 1:51:48to go to this explore tab all right which it says an infinite feed from our real-time generator so you go
631. 1:51:59explore and I got to be careful of the the tatas or I'll get i'll get booted off Tik Tock but um Let me let me go
632. 1:52:07grabb okay I'll go grab this image so so what this thing's doing is it's generating in real time all these
633. 1:52:14images so you don't need to prompt this thing at all and then I can turn this there's this button down here that says
634. 1:52:20create station and if I click on this boom create station what it's going to create is like a radio station but it's an image [Laughter]
635. 1:52:34station so all of these images that are being made are they have never existed before each one of these is an original image
636. 1:52:45being made in real time I can share this um this station with you and we'll actually we'll see the same images
637. 1:52:53so if you want to go tripping you want you want to do some some some uh hallucinogenics uh as they call them um
638. 1:53:01and uh and have some fun just go put Deco here into into station mode into some visual that you think is cool let's see
639. 1:53:11here's let's see this would be a good one for a trip I just received my first AI art this week awesome fantastic love this
640. 1:53:32one isn't that awesome every one of these has never existed before it's just every one of these is like a mashup of 10,000 images
641. 1:53:43all from a similar visual cluster can you prompt it yeah you can so if I go to um generate image I can say um let's see uh oh my my
642. 1:53:58mouse it my mouse does not it looks like my browser's crashing well that makes sense with how many tabs you have open
643. 1:54:07um let's see um let me say um abstract expressionist um super colorful
644. 1:54:37mushrooms rock paper scissors Kyle have you been playing with this all day how else would you know how to use this wait this this wait
645. 1:54:53the little interface this [ __ ] already which [ __ ] the the the um the reasoning engine or this thing this this thing
646. 1:55:01I've played with for for a year now all right that one's good let's go into this image oh man my mouse is is my Mouse dying or is my H come
647. 1:55:13on uh okay I got a trackpad here yeah my mouse is D share huh wait generated oh I gotta do I have to do real time AI I might have to do
648. 1:55:26realtime AI um so what do I do abstract expressionist painting of mushrooms um the reason I know how to
649. 1:55:42use this [ __ ] is why this channel exists I think that is actually really important to get hyper curious right now expressionist
650. 1:55:54um and just go nuts on this stuff um okay yeah so any one of these okay so here's so I can basically go and and we'll do of of Twisted
651. 1:56:08mushrooms Twisted mushrooms Twisted crazy mushrooms
652. 1:56:23all right that one's good so now I can turn this into a station and then say go to the station go to the station oops yeah there we
653. 1:56:38go crazy right so so have your little Fung guy [ __ ] I fell asleep where we at I was sarcastically talking about
654. 1:56:52deep seek I'm fascinated by all this you know I am yeah yeah yeah um okay anyway back to uh back to chat GPT so that's Deco here app. deoh
655. 1:57:08here. go to the explore Tab and then if you use realtime AI to generate a prompt then you can turn that into one of these
656. 1:57:17things 20$ 20 million Studio can't keep your mouse charged yeah let me let me charge my mouse it's a good
657. 1:57:28point oh man all right so let's get back to chat je so here we are at chat je so we've done we've done deep seek R1 reasoning Engine with
658. 1:57:44the same prompt we've done it with Gemini and so what I would argue is deep seek it's it's thinking is more creative
659. 1:57:52its output is more straight laced Gemini flash thinking it's it's a Chain of Thought is really boring but its output
660. 1:58:02is really creative which that's that's a counterintuitive one for me I'm GNA I'm going to go play with Gemini uh flash
661. 1:58:10thinking a lot more that looks really interesting to me okay so now I got to go get my prompt again hang on copy my
662. 1:58:17prompt and if you don't know the prompt that I put in here what I'm having reasoning engines help me figure out the
663. 1:58:24problem I'm having them solve is I'm not a PhD with PhD level problems so I put in here here's the [ __ ] I know how to do
664. 1:58:32help me come up with ways I would use a reasoning engine so I'm using a reasoning engine to help me figure out how to use
665. 1:58:41it okay so here we go bang all right so there's that now it's thinking we can open up this thinking mapping use cases I'm outlining a Chain
666. 1:58:54of Thought reasoning engine how it can Aid in what you asked for I'm refining the approach I'm creating structured queries
667. 1:59:06um Deco here is Pinterest on steroids I picture eye candy it it is it is absolutely that Michael NE the problem
668. 1:59:15with Deco here is that none of the images seem per permanent I lose sight track of images I like you can
669. 1:59:22for any image you can drop it in your library in fact let me go there and show you how so all right let's say I like this
670. 1:59:33little mushroom witch I can say there's upscale there's a little plus thing here so if you click on any one of these
671. 1:59:40ephemeral images and say add to library it'll add it to the library with its prompt and then you can always turn that
672. 1:59:47into a another oh that one's cool oh the other cool thing about this is so I did the mushroom prompt and then look like this one's really
673. 1:59:59different this one's really kind of cool and trippy I can turn I can use this image now as a base to start another
674. 2:00:08station and so it'll take whatever the prompt was for this like like I think what they're doing is slightly altering
675. 2:00:14the prompts and look now now we're in a completely different what we're what what you're essentially doing here is
676. 2:00:23what I would call surfing the latent space so the way these diffusion models work is they've got you know Millions if
677. 2:00:31not billions of images all digitized in these random noise patterns and then for any given prompt it it mashes together I
678. 2:00:41don't know thousands tens of thousands of of these random pixels and D noises them using a d noising algorithm every
679. 2:00:50one of these images is in original image but like we're in a completely different part of the Laton space now so the
680. 2:00:57images that it matched together is completely different than what we had before even though we started with the same basic
681. 2:01:03seed you're surfing the AI dream literally right you're literally and it's like I'm sorry you tell me you look at this and
682. 2:01:13tell me AI can't be creative it's the biggest piece of horeshit anyone that says AI camp be creative hasn't used it they just
683. 2:01:25haven't used it like get [ __ ] curious and like it is up to us as humans to determine which of this like
684. 2:01:36right now this is just vomiting out ideas right but like I could take this image and I could go oh that really
685. 2:01:43speaks to me let me create a station out of this or let me take this image and let me start to iterate on this and let
686. 2:01:49me write a story around this right IDE beget other ideas should I work on applying these or work on what's more rewarding
687. 2:02:02you should always work on what's more rewarding like I I firmly believe this that like like if if there if someone asks me what AI
688. 2:02:17is for me and I think for every everyone what AI is is an idea amplifier you have an idea and in the old and timey world three years
689. 2:02:32ago to get that idea out of your head you would have to go to a whiteboard or you have have to sketch on a napkin or
690. 2:02:38you have have to sit in a cafe and pretend like you were in angst just but just realized you were in the cafe
691. 2:02:44because you had writer's block but you're trying to impress people with how smart you are and creative you are and
692. 2:02:50right you'd have to spend a year [ __ ] milking that idea out of your head with AI you have an idea and bang it's instantly in the
693. 2:03:00world and then you can take that idea and you can amplify that further and you can take that idea and amplify that
694. 2:03:06further and you can just keep iterating so rather than you taking a year to get an idea out of your head to
695. 2:03:13get it to a point that you go ah no I don't like it you can now do that instantly so then that puts the pressure
696. 2:03:22back on you to say what's your [ __ ] idea so I think that you should always be pursuing the thing that speaks to you
697. 2:03:31but don't start with the tool start with the what am I trying to accomplish in the world and then you if you're curious
698. 2:03:38you start to say well what might be the tools that can help me get that into the world oh I should use a reasoning engine
699. 2:03:44I should do this and I should go write programs I should write applications I should do things that make people's
700. 2:03:49lives simpler great go do that someone else is going to make posters and say oh this aesthetic was lost in
701. 2:03:57the 60s and I want to bring it back again great go do that all right use existing or making your own llm is the question
702. 2:04:11again if you're passionate about making an llm go make an llm if you're just trying to accomplish some something figure out which llm can
703. 2:04:21accomplish that best for your idea there's no right answer to that it it is 100% about what your intention is what
704. 2:04:29your goals are okay let's go back to chat je all right so this had very short reasoning so chat gbt had very short reasoning so let's close
705. 2:04:42that and let's see what it came up with for us um creative brain stor character concepts for a short play here's the description here's the
706. 2:04:53prompt I want to create this um developing a layered plot structure here's the prompt again these this these
707. 2:05:02prompts are very similar to what um deep seek created they're they're fine they're a little straightforward for me right now
708. 2:05:13in this head-to-head Showdown Pate are you listening I don't know if P's still here I think Gemini was by far the most
709. 2:05:19interesting of these three the flash thinking prompt I want to improve my leadership skills and understand
710. 2:05:28different leadership styles can you reason through different leadership styles EG transformational servant
711. 2:05:34autocratic Democratic for each style explain its core principles its potential strengths weaknesses like
712. 2:05:41that's a good prompt for a reasoning engine I think Gemini flash thinking experimental 121 pretty [ __ ] fcking
713. 2:05:54bitching I'm going to go play with this but here's what to take away from tonight these tools are really weird they're very
714. 2:06:07sophisticated they're very nuanced like what's the difference between regular chat GPT and the reasoning engine
715. 2:06:15they're doing very different things but when would you use one over the other I don't quite know that yet that's part of
716. 2:06:21the exercise tonight was like I just want to understand like what is it even what do they even do compared to one
717. 2:06:27another and like I'm really digging Gemini I feel like the results of open Ai and deep seek were more expected and
718. 2:06:39Gemini was more interesting and more surprising so was deep seek trained on Contraband Nvidia h100s we don't know
719. 2:06:48and lied about its training costs undercut the U US market is that true I don't [ __ ] know I don't know do I think that deep seek likely
720. 2:07:00has 50,000 h100s that they're not allowed to talk about sure is it possible that they trained what they you
721. 2:07:08know deep seek on 50,000 h100s instead of 2,000 R 800s or whatever the [ __ ] those things were called possibly the benefit for us is
722. 2:07:22that we now have a new model and we have a new approach for how that model was created that's going to give our
723. 2:07:30research Labs new ideas to go leverage their big horse horsepower with um I don't know possible 2025 shit's going to get really
724. 2:07:44weird the fact that a the fact that a chat app a chatbot app Rose to number one in the App Store and crashed the
725. 2:07:54NASDAQ in a single day I think is just the beginning of a [ __ ] storm of 2025 like like do not think this is the
726. 2:08:05last weird thing that's going to happen it's not always going to be bad it's not always going to be China eating our
727. 2:08:10lunch it might be that meta does some breakthrough or open AI does a breakthrough anthropic comes out of nowhere and says we've actually got
728. 2:08:17something and we were just pretending like we didn't like someone's going to cure some [ __ ] disease that was uncurable someone's going
729. 2:08:28to break through Fusion China did that they they ran a fusion reactor for 18 minutes like we're getting close to
730. 2:08:40that I I mean it 2025 is the beginning of when shit's getting weird so I hope you I hope you rested well in
731. 2:08:492024 Winston paint Pate is a great lecturer I enjoyed his AI mechanics with Pate oh that's awesome thanks again Pate
732. 2:08:55thank you for sharing that Winston and and thank you Pate for doing that let me go give you a little shout out so if you're new
733. 2:09:03here you should be if you're curious about this AI stuff if you just want to [ __ ] on this AI stuff um I I would I
734. 2:09:11don't know why you would hang out here because independent of whether or not you love or hate AI it's still here and
735. 2:09:22it's not going away this channel is about let's just accept that as the starting premise if it's not going away this is a very binary
736. 2:09:33Choice you're either on the beach or you're teaching yourself how to surf the waves that are coming I would much rather be on a
737. 2:09:42surfboard even if I'm clumsy on it surfing the wave rather than being on the beach thinking it's a nice sunny day
738. 2:09:49and bitching about AI but any anyway if you're curious about AI if you want to be in a community of people that are
739. 2:09:55curious about AI join the AI salon so you go to this URL you click on the button that says join our community and that's going to take you
740. 2:10:04to the AI salon and it'll start you on a welcome page where you can watch a video from myself and Leah faston who co-founded
741. 2:10:14this um you can talk about the or you can learn about the basic journey to AI Readiness in the the AI Salon really
742. 2:10:22important where we believe you should start with AI is playing play not working not mastering not figuring out the F perfect
743. 2:10:36prompt engineering framework play just play Just dick around with it like I did tonight it doesn't [ __ ] matter just
744. 2:10:44play play play and in playing you'll learn enough about what these things can do that you can move into that next
745. 2:10:52phase of mindfully create oh I've got a problem I want to solve oh I could use that Deco here thing to make some cool
746. 2:10:58images and I could use this to do that and I could go do that Gemini thing and then the third one is generously
747. 2:11:07lead and that means Learn out Loud share with people what you're learning become a trusted member in this community because this
748. 2:11:16community has some of the smartest most creative people that I ever seen figuring out how to use these AI Tools
749. 2:11:25in remarkable ways that are not about technology they're about Humanity so anyway that's this group so
750. 2:11:34if you scroll down down through a bunch of really cool [ __ ] there's an area called clubs and hubs and down in there
751. 2:11:41is this um club called uh AI mechanics with pate and Pate just did an event on Saturday a demystifying AI event and
752. 2:11:52he's doing a series of these things and so the conversations and all that stuff's in there so you should go you
753. 2:11:58should join the salon you should introduce yourself you should go check out AI mechanics with pate and sign up
754. 2:12:04for his stuff especially if you're a little Technical and like want to geek out you want to really understand how this [ __ ] actually
755. 2:12:11works I personally feel that understanding how it works for me makes it more remarkable what it does it is it is just a
756. 2:12:26shitty calculator at its core the the mechanism is just a shitty calculator just like our brains are just a bunch of long
757. 2:12:36cells that have chemicals in between them right it's just kind of a shitty it's just like a shitty thing but it's
758. 2:12:45what goes into that and what gets stored in that is where the magic is and that's same with AI the tool is the tool but what it's
759. 2:12:54been trained on is all of us and so being in a group like this you really get to experience lots of different people with lots of different
760. 2:13:03points of view using these tools in remarkable and Powerful ways okay black box I know I obviously don't or I would have fixed it a long
761. 2:13:13time ago okay I gotta get out of here I'm late I'm keeping P or to uh Brandon I'm keeping all of you out out too late
762. 2:13:20on a Monday night uh three starred questions on streamyard okay let's go to the starred questions here we go okay
763. 2:13:27first one have you tried pixels I have not I will look it up though um I was using an AI to help make
764. 2:13:36stories at first just to tell my nieces but who knows how far these stories can go that is okay what's your name u u
765. 2:13:46Adamas you're on to something really big there let me uh let me flip over to this different mode um one of the things that's now possible
766. 2:13:59with AI is the parallel processing of creative outputs and here's what I mean by that in the olden timey world three years ago
767. 2:14:14if you wanted to write a children's book you would have to you know come up with an idea that's still exists then you would have to
768. 2:14:23learn to write at whatever level book you're trying to create and you know figure out how many pages should it be
769. 2:14:30and you'd need to find an illustrator if it wasn't you and it would probably take you you anywhere between six months and 20 years
770. 2:14:42to get that book out of your head and into some form and you could really only work on that form right now with AI you can write
771. 2:14:52that book in 20 minutes and then you could go to sunno and make a whole album of songs for every story in that
772. 2:15:02book and and have an album with it and then you could create a podcast that is has got cute little voices on it that's
773. 2:15:09designed for kids to listen to and publish that like within a week you could have a dozen different outputs so so your point there about who
774. 2:15:19knows how far those stories can go again what this comes back to is if you're having that instinct run down that
775. 2:15:27rabbit hole what should it what should that story extend into maybe that children's book evolves into a novel or a screenplay that's for
776. 2:15:37adults why not you can have chat GPT write that for you go use one of these reasoning engines say Hey listen I've got this
777. 2:15:45kids book I want to take the lessons in this kids book and turn it into uh this kind of M I like these three kind of
778. 2:15:52movies I want it to be kind of like those it will write you that outline and and if you use these reasoning engines
779. 2:15:59to do it it'll be pretty [ __ ] good all right 2024 is when I prepped and ironed my brown pants that's yeah exactly that's that's uh that's
780. 2:16:10it 2024 is uh that was the last time it wasn't going to be weird all right time check all right I'm gonna get out of
781. 2:16:18here my voice is going uh I talked a lot tonight I hope that I hope you found that useful um if you hear me getting salty and
782. 2:16:27passionate about [ __ ] um when I get ignorant comments that are just tropes of what people think I AI is versus what it actually is it
783. 2:16:40it it makes my neck twitch so so I I love people having opinions I just want you to have an opinion from a place of
784. 2:16:49using it that's my whole stick my whole [ __ ] thing Lord digital Gods thank you so much Tams chat has walked me through
785. 2:17:00everything I needed to know and how how to publish my kids book awesome beautiful AI is expanding all parts of
786. 2:17:06our Lives yeah AI is an expander and an amplifier of what we want to do most people have not put much time
787. 2:17:15into thinking about what they want to do because life is so [ __ ] hard and it's so [ __ ] hard to fill out this form
788. 2:17:23and do those [ __ ] taxes and call the insurance company and get this paid for and apply for the loan and deal with my [ __ ] credit
789. 2:17:33score as AI starts to automate out that shitty work we get to step into who do you want to be so think about that start thinking
790. 2:17:44about that what do I want to do in this world assume that by the end of this year you'll have 10 worldclass employees
791. 2:17:54at your disposal to do anything you want with are you just going to let them sit there I'll get to it next week are you
792. 2:18:01gonna go do something go do something think about it now all right that's it getting the hell out of here the 10
793. 2:18:08operators employees has my mind blown I know I think I'm spinning it now thanks Kyle you're welcome I know this is this
794. 2:18:16transition by the way from where I was to to get here has been about an eight-month journey of just knowing like
795. 2:18:23why can I not get my head around this AI [ __ ] and it's because I spent a year and a half on this channel talking about the tools
796. 2:18:34first not talking about what you do with them and why and and basically where where I'm coming to is what you do with them and
797. 2:18:45why the creative piece the imagination piece the critical thinking piece the my ideas piece in the world I want to do this I
798. 2:18:55want to do it in the following way that's me that's the most important thing and it's actually the hardest that's that's where all your
799. 2:19:04vulnerability is people the [ __ ] you really want to do the [ __ ] that your stepdaddy told you you weren't talented enough for when you
800. 2:19:14were eight and you shut that part of yourself down that's the [ __ ] that we get to ReDiscover now and say [ __ ] that I'm
801. 2:19:25going to go do that thing because maybe I'm not talented enough but you know what now I got [ __ ] AI look at that Talent
802. 2:19:42pops ah excellent life kylin it was even more fun watching you on YouTube on my TV awesome awesome awesome all right
803. 2:19:49everybody uh I'm gonna going to blast the YouTube and X on out of here great seeing youall hope you had fun tonight peace out bye [Music]