5/14/2025 - Navigating the Maze of AI Models: Understanding OpenAI's Latest Developments

Transcript

1. 0:00Come here, buddy. [Music] Did I close the door? Yeah.
2. 0:16All right. You ready? You ready? [Music] In a westerly. [Music]
3. 0:34My driving. I've been [Music] wondering what it is I'm running from [Music] again. Feel like an 80y old man holding on to 29.
4. 0:52And up on that rer is a California [Music] line. Double ahead of trucks carrying a
5. 1:08wide load. Refab house cutting high. Cute little front door and two windows. alone ain't sure whether to cry I should
6. 1:23laugh you see I broke a home up myself once days I stumbled to that I know father don't lie said don't you come around
7. 1:41here anymore had the [Music]
8. 1:56street. He says, "What happened? [Music]
9. 2:28Oh, good people. Good people. Good people. What is happening? [Music]
10. 2:58There's been something baby I've been trying to [Music] say. What age? And it seems I don't know how with the past and the future now
11. 3:12surrounding me. Surrender to whatever cheap can be found. There's been a little trouble since you came to my rescue.
12. 3:30[Music] And if you like all of the rest, I would have quit you long ago, but I couldn't do that. Oh, tell me now, woman. And why I
13. 3:44never went [Music] to make a man crazy, make him cold as hell. I'm a woman, you wish me well.
14. 3:59But it's a trying. Still going to have to find my own way [Music] through. Yeah. Like the mist of morning, my dream
15. 4:13remains. What is going on, good people? All right. You're on the yellow sticky FYI. Wait. Oh, okay. Wait, what?
16. 4:28You're on the yellow sticky. FY. Okay, now I can see. Oh, I see. I stepped on your ability to edit. Got it. Producer Brandon freaking out
17. 4:40because he doesn't have [Laughter] [Music] sticky. Oh man. [Music]
18. 4:58Is that hair AI? Yeah, it's we've got some Hawaii 5 action going on today. I don't know what to tell you. You know
19. 5:06the 70s called. They want your hair back. their hair back. [Music]
20. 5:33[Applause] [Music] [Music]
21. 5:48Oh, I was going into withdrawal for those three minutes cuz we were late. Don't tell Source Camp. She's going to
22. 5:54come in here and be all guilty that she was late. So, we can't You can't admit it. I'll tell you my favorite story on
23. 6:00on on being late. Um I think I've told it in here before. So we we talked to someone about StoryVine and they're a major uh direct marketing
24. 6:14skin brand. We'll leave it at that. And um we talked to them and they're like, "Oh, StoryVine sounds interesting. Would
25. 6:22you be willing to fly out to San Francisco to meet with us and tell my team about it?" Sure. So, we book flights. We're all
26. 6:33excited. We fly to San Francisco. We book a lot of other meetings because that's how we roll. We show up to the
27. 6:40place and we say, "Hey, we're here to meet so and so. Really excited to meet her. Glad to be here." The response was, "Can you prove
28. 6:50you have a meeting with [Laughter] her?" And we were like, "Well, we've got some plane tickets from Denver to San
29. 7:01Francisco because we were asked to fly out here at her request." So, so they're like they they were kind of stuck. So, she said, "Well,
30. 7:15she's in a meeting for the hour that we were supposed to meet with her. So, can you wait? We'll wait. We'll wait." So,
31. 7:22then the hour passes. We go meet with her team. There's like three people in there. They don't know who we are. We're in there for 20
32. 7:31minutes. She walks into the room, sits down, doesn't acknowledge a thing. Just said, "What are you guys talking about? Show me what you got.
33. 7:46Let's go." Didn't [ __ ] acknowledge it. That was Man, that was that's the height of corporate douchebaggery. [Music] You [Music]
34. 8:18know, you [Music]
35. 8:40Um, what are we going to talk about tonight? We're going to talk about I want to talk a little bit about Open AI.
36. 8:46I think OpenAI is currently in the process of jumping the shark with chat GPT becoming functionally [Music]
37. 9:00unusable. And I don't mean unusable like you can't use it. I mean unusable like you can't use it at its best.
38. 9:13Um, the whole reason chat GPT became popular was because they took all the technical choices away from the end user. They just made
39. 9:26it the chat hole. That's all it was. It was just the chat hole. And I, you know, I don't necessarily blame them or any of the frontier model
40. 9:38companies for for falling into this challenge, but because they're all copying each other and because all of them are copying Open
41. 9:49AI, the problem that Open AI has and is proliferating extends across all of the large language models. Anthropics is one
42. 10:00that is slightly better. They've they've hidden some of the [ __ ] within Anthropic, but it's it's just really
43. 10:09bad. So, I'm going to show you there's there's some new [ __ ] in Open AI. And normally, not only do I kn know there's
44. 10:17new [ __ ] in there, I know why it's in there. I know what it is. I know how to get to it. I know how to use it. And I
45. 10:22know why you would use it. It's in there. I don't give a [ __ ] It's in there. and I do not give a [ __ ] The
46. 10:29health benchmark announcement from OpenAI, that's a that's an interesting one. I haven't read up on it, so I don't
47. 10:34I'm not educated on it. We could certainly go look at it. There's also an interesting paper from uh Google Deep
48. 10:41Mind today. Um which I think might be worth going to look at together. Um let's see. Veil and wash. What is
49. 10:51going on over here? Why am I Oh, I see. My chat GBT was humming. Now they [ __ ] with it. It name is Kai. Now Mr. It named his
50. 11:02his catchy BT Kai. Well, mine's Quinn. Um, here. Let me go show you something. Let me flip [Music]
51. 11:17over YouTube question. Hey, Kyle. James Dutder. At some point, can we see the render that was running when you signed off?
52. 11:29Yes. What was that? That was Hedra, right? Is it here? I think it's here. Oh, yeah. I think I watched a little of
53. 11:39it. It was not bad. It's not great, but it's not bad. Um, yeah. Let's we'll do we'll do this because it's a nice little entertaining
54. 11:48start while we get some more people in here. Um tabs for audio. Yep, exactly. I'm working on that now.
55. 12:03Okay, so here we go.
56. 12:22[Music] So, for those of you that were not here last night, the
57. 12:40um the process to get to this video was the so the little kid with the marbles in the sweater and the hat and the tan
58. 12:49pants. That was a picture of me um a storyboard done in Keynote with just block shapes basically that said
59. 12:58here's where the kid is, here's where the marbles are. Um, and then we put him on a schoolyard. And then the two guitar guys
60. 13:07were made after we we went over and made this song in Sunno called Marble Man. And they were really cool. And then
61. 13:17someone said, "Hey, could you use So what we used was Runway ML has this new thing called um references and you can put up to three
62. 13:28images in there. But what you can do is you can sort of put a bunch of people on one image and and label them like here's
63. 13:34Jim, here's Sally, here's this. So, we basically said take these two guitar players and put them behind the kid with the with the uh marble.
64. 13:50Um Yeah. And so here's how it rendered out. [Music]
65. 14:12Marles. They call me the marbles man. Marles. My favorite one's my closest friend. Marles. I see them rolling down the lane.
66. 14:25Marles. I'm in love with my marbles. Ain't that insane? Marles. It's all I think about. Marles. They don't make me scream or shout.
67. 14:39Marles. Every single one's a tip. Marles. I'm in love with my marbles. Ain't that insane? Don't tell me I'm
68. 14:50wrong. You just haven't seen what I've seen. Don't call me crazy because marles aren't just a dream. As for real as you are me, I'm in
69. 15:08love with my marles. Ain't that insane? [Music]
70. 15:30So anyway, so that's that. So I took this video, I put it up on on Yield Internet. I finally got Twitter to
71. 15:37upload a video. It took me, I don't know, 15 attempts. Way to go, Elon. He's he's, you know, literally launching and
72. 15:45catching skyscrapers and making people who can't speak speak again. But you can't upload a [ __ ] video to
73. 15:54Twitter. Anyway, I got it to upload and and then I put it up and I put up a whole instruction thing on how it was
74. 16:02put together. It got like three views. I am so [ __ ] sick of Twitter. So, if you want to go if you would like
75. 16:11to support the channel, go to X and click on um click on my account. Let's see. We'll do it together. Go to Go to the Twitter. Go to
76. 16:26the Twitter and go to Kyle Shannon. And then I think it should be one of these. Yeah, right here. Look at this.
77. 16:36So, I put up this awesome post, Marbleman and Joy, and then I did a comment like, "Here's all the instructions. Like, here's all the
78. 16:45images I used." I described the whole [ __ ] thing. 85 views, no comments. So, you know, go give it some love.
79. 16:58I just the the the engagement algorithm if you don't have [ __ ] 100,000 followers on Twitter, you just get dick. It's just it's it it's
80. 17:11[Music] demoralizing. Joy X's crickets for me too, Kyle. Yeah, it's really [ __ ] frustrating. And like Joy puts out
81. 17:22amazing stuff, right? Amazing content all the time. It's just just maddening. Anyway, enough bitching about social
82. 17:33media. Nobody likes me. Nobody pays attention to my stuff. I don't have any reach. What are you going to do? What
83. 17:44are you going to do? All right. Can we use Zapier? We can to do what? to get people to go click on my [ __ ] Yeah, we could we
84. 17:56could make an automation that just goes and finds people and randomly logs into their accounts and has them log into my
85. 18:03account and click on [ __ ] We could do that. Um, okay. Let's go to chat japet. Oh, look there. This was the little
86. 18:14starting image that led to neck ne.
87. 18:31Um, all right. That was the start image. Come with me. You'll need to switch to the big TV screen and rock the sound. The Kyle
88. 18:45Rubin Kyle Rubin. Yeah, there you go. Rick Rubin. I think Rick Rubin, he he must have a new
89. 18:59publicist because you knew about Rick for a while. He had his book came out and people followed him. But now like
90. 19:04he's got he's got this new show if you haven't seen it on YouTube. I think it's called Tetra Tetra Grammaton. Tetra
91. 19:15Grammaton. Um, which I'm sure means something cool, but it feels to me like it should be Tetra Grammatron, but it's
92. 19:22Ton. Um, and it's like two and a half hour interviews where he just sits there cross-legged, you know, playing with
93. 19:30stuff and just talking to cool people. It's pretty cool. So, if you haven't seen it yet Oh, it's Little Kyle. It is Little
94. 19:38Kyle. Source Camp, you're late again. Listen, one of the things that I think we've talked about multiple times is
95. 19:46like being late is not an option, right? You made a commitment. Your commitment is to be here to hear Champion. This isn't about me, Kelly.
96. 20:00It's not about me. It's about the dog. He misses you. Okay. Black bar. Yes. Fine. Fine. Fine. Fine. Okay. Um Okay.
97. 20:09Okay. So, I'm going to make a new new thing here. A new chat. So, here we are on chat. Jet tag. Couple of things are
98. 20:21emerging. What? How do I get rid of that thing? Okay. than that. Um, they've added long-term memory to chat GPT, but it it is increasingly
99. 20:34appearing like it's only looking back at your past year, not the entire chat history, which if that's the case,
100. 20:41that's kind of shitty. Um, that will get better, right? The fact that they can do it for a year means that they can do it
101. 20:48longer. It's just, you know, they don't want to spend the cycles or whatever it is, um, or the hard drive space or whatever it might be.
102. 20:56Source Camp's rebuilding her entire website. Oh, Jim Ross, I don't know if Jim Ross is in here tonight. He sent me
103. 21:02a thing on LinkedIn. He used lovable.dev to create his own community site. So, like, you know, on the AI salon, we're using Mighty
104. 21:13Networks. Jim had lovable create a standalone community site for the self-s storage community. And it's better than
105. 21:22anything out there for any of the self- storage community. And in showing it to industry people, the first question they
106. 21:29asked was, "How do we can we give you money to put our brand inside this thing?" Absolutely remarkable. Using AI.
107. 21:37Yeah. Yeah. He used AI to to build a community site. That's lovable.dev. Oh, am I asking? So, wait. Wow. Jim, I'd love to see that. Yeah,
108. 21:52exactly. Um, Kelly, are you using are what are you using to build your site? Are you using AI? Oh, Kelly is using it. What are you
109. 22:03using? Which tool are you using? Oscar, you always have great stuff here, Kyle. Hang in there. It'll skyrocket. Yeah, I don't know about
110. 22:14that. It's been two and a half years. If I was going to skyrocket, I think I would have skyrocketed by now. I think
111. 22:20the uh the market the market for fat saucy Gen Xers is is just not what it used to be. I posted my method. Keep
112. 22:35[Music] reading. Oh, I gave chat my web address and asked it to rework every page, giving me an image. Oh, that's so cool.
113. 22:45That is so cool. Uh it's unbelievable. Okay. So, let's So, now that we now that Kelly has has goodmouthed OpenAI, let me badmouth them a
114. 22:56bit. So, here we got our chat GPT, right? About a month ago, OpenAI launched. Okay.
115. 23:11About two months ago, Sam Alman said, "There's a new version of 40 and then we're going to launch 03." Which confused the [ __ ] out of
116. 23:24people because they're like, "Well, wait, isn't 03?" Like, I'm seeing posts on Twitter right now like it how come 03
117. 23:32seems to be giving me better answers than 40? Isn't it an older model? No, it's not. You didn't pay attention to where the O
118. 23:42was 40 03. No, they're completely different things, right? So, so they've already got the [ __ ] naming
119. 23:48nightmare, right? So, they launched that and Sam says not to worry. Sometime in the next month or two, so maybe by
120. 23:58May, we're going to launch GPT5. And GPT5 is going to take all this naming nonsense and consolidate it into a
121. 24:07single model where you just talk to it and the model will figure out what to do. Okay, awesome. About a month ago, just before
122. 24:17May, little Sammy boy came out and said, "Hey, great news. Here's here's model 4.1." And is it 4.10? Wait, hang
123. 24:35on. Wait, why do I not have it now? Did they already take it away? Wait, do I need to reload? Maybe I didn't
124. 24:50reload. Okay. Yeah, it's in there now. So, it's Okay, so 41 is not 410. Okay. Okay. See, I like like I said, we we're
125. 25:02we're just at a [ __ ] disaster now. So, about a month ago, he says, "Good news. We're coming out with 4.1." Oh, by the
126. 25:12way, I know we said GPT5 was launching, but it's not. It turns out consolidating all these models into some sort of model
127. 25:21picker kind of thing is harder than we thought it would be. So, that's going to be kind of a couple of months from now.
128. 25:26So I'm thinking like June, July, August, some like GPT5 is sometime midsummer or later. And he said, "So here's two new
129. 25:36models, GPT 4.1 and 4.1 high or mini 4.1 and 4.1 high. They were only available in the API which I was like I was cool
130. 25:48with that because it's like great there's some new cheaper faster 41 model that the developers have. Great. One of the cool things
131. 25:58about 4.1 is it had a million token context window which is that's good. If you don't know what context windows are,
132. 26:05it's basically like your chat's short-term memory. All right, Jim LinkedIn comment. Jim Ross, I'm happy to
133. 26:12showcase the site I made. Yes, I would love for you to do that. Uh, you can do it on office hours on Friday. You could
134. 26:17even do it tonight if you want to, Jim. If you uh reach out to Brandon or Brandon can reach out to you, I'm I'm
135. 26:25happy to have you come up here. But if you're sitting there in your jammies and you know, you've got like, I don't know,
136. 26:32cereal in your hair and don't want to come on screen, that's that's cool, too. Um, but I would love to see it uh
137. 26:38tonight because it's just I think it's really exciting and I I want to hear like what people said about it. So, if
138. 26:42you're if you're willing to do it tonight and Friday, that'd be awesome. Um, so anyway, today they put 4.1 into chat GPT. Now,
139. 26:56they hid it under other models. So, look at this [ __ ] disaster we've got now. We've got 40. Okay, the O means omnimodal, right? This
140. 27:09remember remember when I did this turned into a whole [ __ ] anger rage post, right? Explaining these things that
141. 27:16people seem to enjoy that post. 40 is the the omnimodal multimodal model that that does pitches, right? 4.5 is this sort of like appendage
142. 27:30that is not multimodal. It's just kind of better at writing and kind of has personality. And David Shapiro thinks
143. 27:38that this is going to be the the coordinating model when when GPT5 starts navigating between models.
144. 27:47He thinks that that's what this is going to be used for. But we don't but no one knows that. And so people don't know.
145. 27:53They think 4.5 is just 4.0 but better. But it you can't make images with this one. Then you have 03. But because it's
146. 28:01lower than 40, people assume it's older and worse. This is the big Mac Daddy [ __ ] reasoning model, right? Then
147. 28:09you've got 04 mini and 04 mini high. These are the newer reasoning models that are in no way related to 40. They're they're standalone mini
148. 28:19reasoning models. And now we've got 01 Pro and we've got GPT4.1 and GPT4.1 Mini.
149. 28:34What the [ __ ] are they doing? Like [ __ ] retire something. This is ridiculous. Like this is at this point like this is beyond
150. 28:48satire. And I I mean I get it. You've got capabilities that people want to have access to the capabilities, but
151. 28:55like I pay attention to this [ __ ] every [ __ ] day. And of all these models, I definitely know what 40 does. I don't
152. 29:04know what 4.5 does. I don't know what it's good at. I don't know what it's not good at. I've used it a couple of times.
153. 29:09It hasn't blown me away. Other people say it's completely blown them away and that's their daily driver, but they're obviously not making
154. 29:19images. And then 03 is really good, but I don't like if I'm going to use something that's reasoning like I tend
155. 29:25to use 03 because that's the bigger beefier model. But I only know that because I pay attention to this [ __ ] I don't know
156. 29:34when to use I like I don't use 04 mini at all. I use 04 mini high or 03. I don't use 01 pro anymore. And I I don't
157. 29:42think I'm ever going to [ __ ] use 4.1. I'm not even going to test it. Like who gives a [ __ ] The only people paying this close
158. 29:53attention to the nuances of the models are developers. So, I I don't know if Pate's in here,
159. 30:07but I tend to be a bit of a suck up to Open AI because I think they're just the best at everything right now. But this is like we they've
160. 30:18officially jumped the shark. This is just chat GBT is becoming unusable. Like I'm paying 200 bucks a month to
161. 30:26have operator. I've never [ __ ] used it. It's a whole separate site. It's a whole separate thing. Apparently I need
162. 30:36to verify this device. [ __ ] it. No. Not verify. Not verifying [Laughter] [ __ ] This is unwieldy Wednesday.
163. 30:48Welcome to the AI learning lab. My name is Kyle Shannon. I'm not drunk. [Laughter] Oh my god. All right. Windown
164. 31:06Wednesdays. So anyway, have you tried chat GPT? It's really good. It's I I mean like what Kelly did like what it can do is remarkable, but
165. 31:17what they're doing to it is inexcusable. Like no one gives a [ __ ] They got to have stats on this. They've got to have stats on this.
166. 31:30Most users don't change their defaults. So that means most users are using GPT40. And then some advanced users are
167. 31:38probably using 03 and probably using 04 mini high because they run out of credits on 03. I would be shocked if the if the other
168. 31:49models had, you know, statistically any usage at all. [ __ ] bury them. Just hide them. We're not developers. Developers are going to go
169. 32:02build [ __ ] on the playground and they need all the models and they understand the differences and what they cost and
170. 32:09the speed and what they're good at and what they're not good at. Ah, Tik Tok question. When should I use a reasoning model? That's a great
171. 32:18question. Um, it's a really good question. Some people are using 03 as their daily driver, right? They're just
172. 32:35using 03 instead of using 40. Um, but I would say in general it's when you've got when you've got a question or a
173. 32:51problem or a research project that is that that definitely has multiple steps in it and could benefit from those multiple
174. 33:04steps. Okay. When you use 40 and you say, "I'm uh let's see. Uh [ __ ]
175. 33:19I'm thinking about a marketing plan for storyvine.com."
176. 33:35And I just hit that. It says,"Great. Let's build one. What's your main goal right now?" So it asks you a few
177. 33:43questions. And then it says, you know, "Here's your marketing plan." So 40 just kind of you give it a prompt and it just
178. 33:51kind of vomits out an answer. Well, to really do a marketing plan, you'd need to do things like, I don't
179. 34:01know, research competitors and think about the market you're in and all that sort of stuff, right? So, so if I go to 03 and I say,
180. 34:12let me restart this. Um, I'm thinking about a marketing plan for, and this is a really shitty prompt, by the way,
181. 34:27storyvine.com. Oops, not storyvine.co. Stop. storyvine.com. And so it says
182. 34:42thinking and so it says the user wants marketing ideas for story mine. A general location isn't crucial. I could create a marketing strategy
183. 34:52using my internal knowledge but I should do a web search first. The user prefers. So if I expand this these these things that it's
184. 35:02talking to right here are not my answer. This is its internal logic, right? So the user wants this. I could create a
185. 35:12marketing strategy, but I should probably do some research. The user prefers um original and innovative ideas. So I'm thinking of
186. 35:20modern creative marketing approach. So this is using its knowledge of me. So so it's thinking internally. Then it
187. 35:28searched the web. So, it searched uh storyvine video storytelling platform 2025 press release guided video platform
188. 35:36customers healthcare looking at a bunch of other stuff outbrain seven more things the user values originality I'm thinking interactive
189. 35:46vine sprints UGC AI video challenge marketing play for storyvine.com so now it gives me the answer so so I would say that when you
190. 35:57use a reasoning engine and certainly if you If you have anything involving math or analysis, definitely use 03.
191. 36:05Um, or 04 mini high, that's really good at analysis, too. This prompt that I gave it didn't it it didn't really dig into its chops.
192. 36:16But one of the things that 03 can do is it can create its own tools. So, let's say let's say that not only did I want
193. 36:23um a story of my marketing plan, but I wanted I don't know the revenue results from a bunch of competitors or I wanted
194. 36:31I wanted uh what are the pharma brands that are launching in 2025 and 2026 and what's the expected marketing spend on
195. 36:40those and then I wanted to do some analysis across all those companies. 03 can write its own tools in Python and
196. 36:47execute them as part of the thinking process. So you use a reasoning engine whenever whenever you want. Basically,
197. 36:56if you've got something where you want a much more considered, thoughtful um answer or it's a it's a more
198. 37:04complicated project, that's when to use it. Jim Ross is busy tonight. Dolly's daughter's volleyball game. But at least
199. 37:11he's logged in here. This is good. You're in his ear. Okay, Jim. Um, awesome. So, so anyway, does that answer your question, Jim
200. 37:23Buckley? Got it. Beautiful. All [Music] right, let's see. I think they have all the models so they can throw them at benchmarks and
201. 37:36show up on the list more often. Yeah, but those those the models in chat GPT itself are not what shows up in those benchmark
202. 37:49models. That's [ __ ] coming from their API, right? They they they launch those things like the the benchmark the
203. 37:58benchmarkers are are putting models in there from the from the systems. It's not coming from the commercial chat GBT.
204. 38:06I mean, that's that's what I'm talking about is the commercial version of chat GPT. And even if you're right, you might
205. 38:15be right. Like, it's possible that those benchmarking things are using core chat GPT, but I don't think they are. I think they're using the
206. 38:24API. But you've got they they now have 700 million users on this platform. 700 million users. How many of those are engineers? Like like
207. 38:39globally programmers and engineers is about 1%. So let's let's make it really generous and say that 20% of those 700 million
208. 38:55people are engineers on chat GPT. 80% of the people don't give a [ __ ] about all these models. And not only
209. 39:07don't give a [ __ ] it's confusing the crap out of them. It's like it's like distorting the brand. They're like, I
210. 39:14don't know what to do here. Like what what why is all this? Who are these [ __ ] techno bro [ __ ] NFT bro San Francisco [ __ ] And they're
211. 39:29right. Ah. All right. All right. We need a dancing robot. We do need a dancing robot. Let's go look at the dancing
212. 39:38robot. And here's the other thing. I read I read the Mr. Strawberry thing the other night and then I I realized it was
213. 39:46probably too hypy, but there's some stuff coming down the pipe where I think we're just going to see in like
214. 40:00incredibly more powerful tools in the next two months. Wait, YouTube question. This might be a dumb question. Probably
215. 40:11not. But are the models connected in any way? Kind of. The So, it used to be worse. This is a little bit better. Okay. So, I just did
216. 40:26this. I just did this. um this 03 reasoning thing for my marketing plan, right? Well, I can now switch to
217. 40:41GPT4.5. And it's supposed to be better at at writing. Oh, that's interesting. It says, "Do do you like this personality?"
218. 40:50I've never seen that prompt before. Or that thumbs up, thumbs down. So, let's see if we like the personality.
219. 40:58Short form US user generated video is devouring B2B attention. Nearly every budget budgets for video. 84% of buyers
220. 41:06say a brand's video convince them to purchase and social platforms have become the new center of gravity. But
221. 41:12Pure UGC is a compliance nightmare. That's what we say in regulated industries. Storyvine structured guided workflow is the
222. 41:22missing middle. Authentic like UGC yet brand safe and repeatable. Yep. Which it got from our website by the way.
223. 41:32Structured authenticity at scale. Capture real voices. Control the story. Publish in minutes. Yeah, I like this. I
224. 41:38I think that's good. So, we'll give that we'll give that a thumbs up. But anyway. Okay. So, now I've switched to I can't
225. 41:45give it a thumbs up. Oh my god. Interface people. Thanks for your feedback. Okay. Um, now I'm in 4.5. I could say, "Give me a list
226. 41:59of 20 LinkedIn article ideas, article ideas that are
227. 42:14um pharma relevant, very catchy and compelling. So, the fact that I can
228. 42:28switch models in the same conversation does kind of make them related. And and this is kind of what what Altman promised was when we get
229. 42:40GPT5, it's going to be doing this navigating for us, right? He acknowledged three months ago that the naming convention is a shitow and
230. 42:49they're aware of it. But because technology is technology and you can't launch stuff as fast as you think you can, they're perpetuating
231. 42:58it and making it worse. And at this point, I think they've jumped the shark. Pharma's video marketing crisis.
232. 43:04Why authenticity isn't optional. The end of boring compliance videos. Structured authenticity is here. How pharma can
233. 43:11create 10x more content with 80 80% less risk. MLR bottleneck. This is great. Um I'm going to say um write me the
234. 43:24first three articles. Oh, we'll we'll do something interesting here. Okay. See the link below. In defense of Open AI from who? [Laughter]
235. 43:54Oh, I see you did you did a you had chat GBT defend itself. No, [ __ ] that. I'm not giving that any oxygen. No. No.
236. 44:09Nice, nice try, producer Brandon. I like it. I like the moxy, but no. Okay. Article one. Far pharma's video
237. 44:16marketing crisis. Why authenticity isn't optional. Pharma marketing is at a crossroads. The era of polished, overly
238. 44:22scripted commercials is losing relevance as audiences increasingly tune out content that feels artificial. Today's
239. 44:29health care consumers, patient, healthc care providers, and advocates demand authenticity. But in a high highly
240. 44:35regulated industry like pharma, how can marketers deliver? The solution lies in structured authenticity. That's pretty
241. 44:42good. It's not bad. Everyone knows compliance videos usually dry, monotonous, and instantly forgettable.
242. 44:48They tick regulatory boxes, but often fail to engage or educate the audience effectively. For pharma companies, these
243. 44:54videos represent the mist. Okay, that's great. Um, date date. So now what I could do is I could flip
244. 45:09over to GPT40 and now say um make me a let's
245. 45:27see loose pencil sketch. image
246. 45:44for article one Tik Tok pin um Ray Edwards the key is authenticity once you take away that once you can
247. 45:56fake that anything is possible yeah and I think we're well well so there's so I think You're so Ry, I think you're on to
248. 46:03something, but the authenticity things I I have a whole company based on it. So, I've been thinking about this for a long
249. 46:11time. I think you can have nonhuman avatars authentically share stories. And if those stories are relevant and they're authentically
250. 46:25expressed from the point of view of the avatar, from the persona, then there's a there's a like an authentic
251. 46:37consistency from the intention of the persona to tell a particular kind of story. And if they tell that well from
252. 46:45that point of view, there's an integrity in that. It doesn't mean it's human, but there's an authenticity to
253. 46:55that. So, I think there is a way to do authentic storytelling without um without a human. But you're on to something big
254. 47:07here. Our social networks are already polluted with bots. They're polluted with bots. And so one of the reasons
255. 47:16that I've been talking up for I don't know two and a half years now, a year and a half, year and a half, um, world
256. 47:25ID is that it's the first scalable um, proof of personhood that doesn't require you to give up your identity, but but it proves
257. 47:37who you are. So I think we're going to start to see applications and social networks that require proof of
258. 47:46personhood even within that like so so what we'll have is we'll have um social networks where you know that
259. 47:57when I post something it's me the human posting it but me the human I can post inauthentic [ __ ] on my channel even
260. 48:06though you know it's coming from person, I can generate shitty, inauthentic content and share that as well. So, what
261. 48:14you're talking about is absolutely critical. Um, but I think it's it's very very nuanced. Um, Ray also gave me some
262. 48:21roses. Great. Thank you very much. Comment from Jeff. I copy pasted whole sections of chat from a custom GPT that
263. 48:28I liked into 40 and it became that custom GPT's personality only with chat memory. Very cool. That's a great use
264. 48:38case. Oh, look at this. That's pretty [Laughter] cute. Pharma's video marketing
265. 48:54crisis. Why authenticity isn't optional. I've been working with 11 Labs on character AI to recreate my son's
266. 49:03persona. Oh, that's great. That's awesome. Listen, I think that's going to be a huge use case. You know, lost people, forgotten people,
267. 49:14um, people with Alzheimer's. I lost my dad to Alzheimer's. So, you know, imagine him being able to interact with
268. 49:21people that are familiar to him that are there 247, you know, have infinite patience, will, you know, will tell the stories
269. 49:29over and over and over again, uh, without growing weary. I you know I I think that I think this Avatar digital
270. 49:36twin thing is is is is just going to be an incredibly powerful thing. There's there's so much talk. It's funny when I
271. 49:43was in Washington DC two weeks ago, one of the staffers that we were talking to was like well she was kind of
272. 49:50bragging, well we've not for nothing but we've done legislation against deep fakes. And I was like cool like great
273. 49:58understand that. I said, "But you know, the the technology behind deep fakes can be very beneficial." And she
274. 50:05like got this look on her face like, "Wait, where are you going, boy? What's what's going on here?" And and I told
275. 50:11her some examples of, you know, some of the examples from the salon where people are using avatars and digital twins in
276. 50:18really powerful ways as like an augmentation of themselves or an extension of themselves and they're choosing to do it. They haven't been
277. 50:26deep faked. they're using their own digital twin to do marketing or do communications in some way and she hadn't considered
278. 50:35it. And so it's like we're we're just in this weird we're in this weird lite phase where because the technology is powerful and profound and
279. 50:51accelerating, people are like doubly afraid of it. And so they're pushing back on it and they're, you know,
280. 50:56they're legislating from a place of fear. Um, and man, does it do it a disservice. So anyway, Kak, did you know
281. 51:07custom GPTs that have less than 8,000 characters generate quicker? Oh, interesting. I did not know that. Makes
282. 51:14sense. Did you know that custom GPTs now use the new image generation model, which is nice? Uh, [Applause]
283. 51:28World ID will still force the authorities to give up your identity for law enforcement. Well, so okay. So what you're talking about is
284. 51:38really nuanced. It's okay. World ID does not know who you are. You can have a world ID without it being
285. 51:52attached to your name. Now, here's the here's the nuance and the subtlety. They have all these things called micro apps
286. 52:01or mini apps or something like that. It's basically like the app store for World ID and there's all these little
287. 52:07apps. Some of those apps you don't put your name into. Like there's a a surveying app and and you can take surveys and I think they pay you
288. 52:16WorldCoin or something like that. You don't have to put your name in that. But there are going to be other apps like
289. 52:23say if OpenAI comes out with the social network which they've talked about they are they are working on well in your
290. 52:31social network you're probably going to put your name and maybe your address and then all your pictures and all that sort
291. 52:36of [ __ ] and that will certainly be able to be associated with your world ID. Don't collapse don't collapse the two.
292. 52:46Um because world ID itself is independent from who you are. It's just it it is literally tied to is the fingerprint of your retina unique
293. 52:57against the fingerprints of the retinas in the database. They're not keeping the biometric data. They're just taking the
294. 53:03hash and comparing it to the other fingerprints basically. Um, but practically what you're saying is true because people will use free apps,
295. 53:16right? If if OpenAI gives us a free alternative to X that doesn't suck and just has people in it, no bots, people
296. 53:25are going to use it and they're going to voluntarily give up all their um all their person their personal data. And
297. 53:32then you've you've you've made you've got a good point. Dancing robot then deep mind. That's good. Retinas have fingers. They do. I
298. 53:42have you. If Have you never looked in an eye? You look You look through the pupil. They've got this special machine
299. 53:47at the at the glasses place. If I've seen videos of it. You look in there and on the back of the retinas there's like
300. 53:54little fingers going, "Hey, how's it going?" It's amazing. It's amazing. All right. I understand. Thanks, Kyle. Fantastic. Fantastic. Bob,
301. 54:06tell everybody he's won. I'm working on mine as a project to leave my kids so they can talk to me when I'm ascended. Yeah, it's beautiful,
302. 54:14Kimberly. Love it. Love it. I mean, like the psychology of being able to talk to you after you've passed. Um, and and and also there's
303. 54:28there's um Neurolink patient number three. I was actually just watching a YouTube video on this before I came in
304. 54:35here. Neural link p So, if you don't know what Neurolink is, it's it's the Elon Musk uh brain implant that's got
305. 54:42like 60 electrodes that goes into your brain. And um number three, patient number three is a guy that's got ALS, um
306. 54:52Lou Gerri's disease, and he can't use his voice. Um how do I get out of this thing? Let's see. What did I do? Oh, there we go. Um,
307. 55:02he can't use his voice. He's got L Lou Garri's disease. And so they took audio recordings of him when he could talk,
308. 55:11right? And they made a voice model for him. And so now they've got this chat app where people can talk to him. And
309. 55:19the chat app translates what the people are saying, uses AI to come up with suggested responses, and then he uses
310. 55:29Neurolink to navigate and choose the response, and he responds to them in his voice, which that's just [ __ ] insane,
311. 55:37right? Really cool. Um, [ __ ] insane. But I like the the psychology of being able to talk talk with loved ones after
312. 55:46they've passed or or even be able to go back and talk to like, you know, great grandpa Jim, like maybe you got some
313. 55:54recording from the 20s of great grandpa Jim and and now you're talking to him based on all of his writings or
314. 56:01something like that. Like the psychology of that, like what's is is that going to aid in the grieving process? Is it going
315. 56:08to delay the grieving process? Does it transform the grieving process? I I we're just it it's I I just The thing about AI is like everywhere
316. 56:20you point it, it's going to completely transform the way we do things. Not just take jobs, right? Again, like I just I feel like
317. 56:31everything is so [ __ ] reductionist at this point. It's just like, it's going to take jobs. Yeah, it's going to create
318. 56:37jobs, too, jerk. But what is what it is going to absolutely do is transform everything every job every interaction.
319. 56:48the definition of psychology and psych and psychologists and therapists and and like what is a human's role when when
320. 56:57you've got something that is infinitely compassionate, non-judgmental, and knows every psychological approach in the history of
321. 57:12humankind and and adapts to everything you've ever talked about and never forgets anything what's the role of the human I think there is
322. 57:22one but I think a lot of the skills required for a therapist to have a lot of that knowledge become less
323. 57:32important but there's something that they have that's going to be more important right and like what's that
324. 57:37balance look like I don't know I just know it's not going to be the same Joe Decult so what you're saying is that even when we
325. 57:46die, we're still going to be forced to talk with our annoying family. Exactly. Exactly. You know what I You
326. 57:57know what's funny is like even within a single family dynamic, right? You know, mom dies or dad dies and and now brother
327. 58:05and sister get to talk to dad. All right. Maybe sister was daddy's little girl and she loves dad. She's like,
328. 58:13"Awesome. I can't wait to talk to dad." And you know, the brother was like, "He never loved me, [ __ ] asshole." He's
329. 58:19like, "I don't want to talk to that [ __ ] ever again." And then all he hears is his sister talking to dad now. Now, maybe he can actually
330. 58:28reconcile with dad because dad's now this compassionate entity. But I Yeah, just [ __ ] everything everything's
331. 58:36going to transform. Anyway, all right. I'm constantly taking pictures and video of my parents. Absolutely. You should
332. 58:42absolutely do that. They no longer even argue because they know. That's really cool. Really, really cool. People can
333. 58:49literally take their argument AIs in the other room while they finish Thanksgiving dinner. That's awesome. Uh the role the role of the
334. 59:00human is to maintain the AI. Exactly. Well, no. You know, I So, seriously, like if I think about something like
335. 59:08therapy and and let's take it out of therapy because I mental health gets weird when you start getting into
336. 59:13um mental health just gets gets really sticky really quick. Let's take physical therapy. So, you you break your knee or
337. 59:24you whatever you do. You you tear a meniscus and you go in for surgery and there's a six week recovery and then a
338. 59:31year of rehab and you go have your surgery and you come out you're groggy with surgery and the doctor goes hey how
339. 59:37you doing Kyle you know here's what we did the meniscus and there was a fold and it was a parrot beak tear and
340. 59:44everything was successful you should be on your feet and and I'm going to write you a prescription for a physical
341. 59:49therapist right and then you don't remember what the [ __ ] he said and then you meet with your physical therapist.
342. 59:56And again, your physical therapist knows all this stuff. Oh, you have this kind of meniscus tear and we're going to
343. 1:00:00you're going to move your knee like this and you're going to do that and I'll see you in a week. Okay. And then you
344. 1:00:07[ __ ] hobble out of there and you hurt and you're on [ __ ] drugs. You're like, I don't what what did she say? What? Oh
345. 1:00:14god. Imagine her being able to say, "Hey, here's the virtual me. It knows your case. It knows exactly um what
346. 1:00:24happened. It knows what stage you're at. Um anything that I just told you, it knows and you can talk to it 24/7 and
347. 1:00:35it'll it'll sound like me, right? Like that's, you know, let's go two years in the future. So, I I get I've got virtual
348. 1:00:42therapist on my phone and I can talk to my virtual physical therapist and at some point I'm talking to the virtual physical therapist and
349. 1:00:51the virtual physical therapist says, "Uh, what you're describing sounds a lot like an infection. Get your ass to the
350. 1:00:58hospital." Right? Like like it will be smart enough to know, "Okay, we need we need to not be talking to virtual
351. 1:01:06physical therapy lady. we either need to be going back into physical therapy for an emergency knee resetting or we need
352. 1:01:12to be going to the hospital because we have an infection or something like that. So I I can see there being
353. 1:01:18handoffs where these things get smart enough to recognize where that boundary is. And so that'll be that that's pretty cool. All
354. 1:01:27right, Tik Tok comment. I also have a virtual mental therapist that sounds just like Sam Sam Elliott soothing. That's very nice. I
355. 1:01:37made my physical therapist an app with lovable. Oh, that's so cool. Silver Fox. That's really good. He gives folksy advice. That's
356. 1:01:46awesome. Yeah, Sam Elliot has that awesome voice, doesn't he? Yeah. How's it going? I can't even I can't even get
357. 1:01:53down that low. How's it going, Kimberly? Yeah, I can't I've been talking too much up here, so my my lower vocal folds
358. 1:02:01right now are not cooperating. Um, okay. What was I going to do, Brandon, next? What's
359. 1:02:17skeleton? Oh, dancing robot. Oh, you know what I was going to do? I think I'm gonna I'm gonna post this. Just going to save this as a All
360. 1:02:46right. It's PNG. All right. [Music] Uh, all right. We're going to grab article number
361. 1:03:00one. Uh um by the way, so what I'm doing right here, if you is What the [ __ ] is going
362. 1:03:15on with LinkedIn? Oh, Tik Tok. Um, start a
363. 1:03:32post. I think I should should it be an article? Maybe it should be an article. Write an article. And it should come from Story
364. 1:03:43Vine. Oh, I got to make this 1920 x 1080. Okay, let's go back to chat. JPEG. So, we'll close that out. Scroll on down. Say, I need it
365. 1:03:59need it to be a wide image for LinkedIn. It's what's for dinner. Drink
366. 1:04:13the tequila and it will lower your voice. Oh, don't read it or anything before I post it. I read that
367. 1:04:25article, but let's go. Let's go look at it here. We'll put it here. Wait, what just
368. 1:04:42happened? I'm confused. Oh, I see what I did. So, this is dead. There's that. I'll get rid of that. That's not really long enough, is
369. 1:05:05it? Let's go back to chat GPT. Okay, so here's a good example. All right, so we have we have this article. Let's go back here. We'll go to
370. 1:05:18Let's go to 03. Is this all the articles? Yeah, it is. So, I'm going to
371. 1:05:35go let's see for article one. It's a little thin. do some research to back it
372. 1:05:50up and beef it up. Now, I could use within Canvas. I'll show you that in a second. Within Canvas, you can just tell it to
373. 1:06:04um What is it? What is it doing? Is it thinking right now? There we go. Um, within Canvas, you can tell it to
374. 1:06:18make things longer, longer or shorter. Um, but I want to use 03 to actually think about article one. Researching to enhance the
375. 1:06:27article. The user wants to beef up the article pharma's video marketing crisis. This would involve gathering
376. 1:06:35data. All right. Once I have the info, I'll replace article one. I'll gather the sources and update the article. It's done some
377. 1:06:46research. I [ __ ] love the future. This is if you've got ADD, what it's doing right now is the whole reason
378. 1:06:56chat GPT is your is your friend. It's actually doing the research for me. Yes. Now, to Brandon's point, I should
379. 1:07:04probably actually read it. There was a really funny TikTok. I think Brandon sent it to me today of a high school
380. 1:07:10yearbook. Someone paid $65 for their high school yearbook and one of the paragraphs they left the chat GPT, you
381. 1:07:17know, if you'd like me to modify this in any way, let me know. They left it in the [Laughter] article. It wasn't you. So, yeah. Oh, it was
382. 1:07:27uh it was it was it was in a sort of group chat within Tik Tok. It was pretty funny. Gotta take the dog for the walk. See
383. 1:07:40y'all later. Take care, Kimberly. Thanks for sharing. He is his own theme music. I am my own theme music. Okay.
384. 1:07:55Article one. Key takeaway. 2025 isn't the year to consider authentic video. It's the year you get left behind if you
385. 1:08:01don't. The data behind the crisis. Consumers crave the real thing. That's pretty good. What's at stake? All right, [ __ ] it. Let's just
386. 1:08:15grab it all. To Brandon's point, you're not just going to stick that in and publish it, are you? Yep. I I'll I'll read it
387. 1:08:27quickly. Boom. All right. And then let's go look at our picture. Do we did we get our picture? Oh, did
388. 1:08:39Oh, maybe I stopped it. Network er error occurred. Let's see. Let's see. Thought for Uh oh. Retry. Did I There's my picture. Okay.
389. 1:08:55Um I need Something went wrong. Please try reloading the conversation. Oh, reload the conversation. Okay, I need the image in
390. 1:09:13wide format and use colored pencils in falto blue and orange.
391. 1:09:34story vines colors. Um, also don't make the lady so
392. 1:09:48[Laughter] sad. Uh, should we talk about how it hallucinates references? I don't know. Let's go look at the references.
393. 1:10:05State of video marketing wisel video marketing statistics. That's right. Power of patient testimonials
394. 1:10:20LinkedIn article. These look pretty good. MLR guide, Fizer case, viz,
395. 1:10:39even these look good. So, what Brandon's bringing up is chat GPT. Okay, again, this this there was a question asked earlier.
396. 1:10:52When do you use a reasoning model? when chat GPT just vomits out an answer. So when you're using 40 and it just you
397. 1:11:00give it a prompt and it it gives you an answer. Um that's not as accurate as if it thinks about what it's doing. So when
398. 1:11:10I asked it to go do research and it went through those steps of I should do this and I should do that and I should go
399. 1:11:15research things and take what I learn and rewrite the article. Because it's doing all those steps, it's actually checking these things.
400. 1:11:23So yeah, you should still check to make sure that these things work, but um I I think you can trust them a lot more than you could before.
401. 1:11:36[Music] Um okay, that's that we're going to pull.
402. 1:11:53The data behind the consumers crave the real thing with quotes. Healthcare buyers are no different. Short form video is farmer's
403. 1:12:04blind spot. Tik Tok style clips now command 33% of total marketing spend. Yet most farmer brands still allocate
404. 1:12:11less than 10% of the budgets to it. I'm just going to say social videos compliance bottlenecks are the
405. 1:12:27culprit. An Adobe study shows the average MLR cycle is now at 21 days. Storyvine authenticity a way out. Painpoint
406. 1:12:45What the [ __ ] is this? Pain point traditional approach structured video fix MLR churn 21day
407. 1:13:00average linear. This looks like notes of its thinking. That's really [ __ ] up. All right, so let's go. Let's
408. 1:13:08go. Let's go yell at it. Um. Oh, that's nice. All right. Well, it's not nice. It's fine. It's fine for this.
409. 1:13:19Okay. Here's a new wide. Okay, that's good. But it it gave us orange and blue falto. I'm going to just say um the quality of the
410. 1:13:33sketch sucks. Can you up the game and make it an HCP? Let's make it a
411. 1:13:47female HCP in a white coat with stethos stethoscope.
412. 1:14:01Listen, I've I've told you on this channel many times, if you're just trying to show people that AI can do
413. 1:14:09something, that's easy to do on a show like this. If you actually want to do actual work, it's much more boring because you
414. 1:14:20have to Okay, if you're new to AI, how many people here? We got 41 and 53. So, this is probably you're you're probably
415. 1:14:25mostly people that know about this, but if you're new to AI, one of the tropes, one of the things that people think about AI is you push a
416. 1:14:34button and out comes stuff, right? Which technically that's what the tool does. But most people who are doing good
417. 1:14:44work with AI, that's not how they work. How they work is, oh, I kind of have an idea. It vomits out some stuff. You look
418. 1:14:50at the stuff you ah let me get some more stuff. Oh, that's not right either. Do this. Do that. Right. All that sort of stuff.
419. 1:15:00So, this process that I'm going through right now where I'm I'm like, "Oh, that's kind of good. Oh, but it's kind
420. 1:15:06of not." That's just the way it is. That's kind of how it goes. And then this chopped off the words. Oh my god. Now you
421. 1:15:16chopped off the words. Make sure all the words fit in the wide image. You
422. 1:15:32ninny, you can you can technically call your chat GBT bot a ninny. Uh it is acceptable in uh in in uh large language model bullying ethics.
423. 1:15:47There's a book out written by Vicky Baptiste. The exit the the ethics of talking constructively to your LLM. Mini
424. 1:15:56is an acceptable term. Okay. All right. The user wants a cleaner, better image. Oh, that's interesting.
425. 1:16:13I had it I had it do the images in 03 in the reasoning model which is probably stupid. Your fan club has converted con
426. 1:16:22completed the first stage of the club quest. That's probably because Mr. It has his iPad [ __ ] going crazy on us. We're up to 40,000 likes,
427. 1:16:33people. If if what you're thinking is Kyle, you couldn't possibly be that popular, you're correct. Did it make a square [ __ ] image? It
428. 1:16:49did. Stop. God damn it. Now you made the image square. Make it
429. 1:17:05wide. Good lord. Good lord. How much extra does it cost to call your
430. 1:17:20LLM a name? Uh, it doesn't cost anything, but if you want it not to not to take you out when when it becomes a
431. 1:17:28physically embodied in a robot, that's extra. If you bully If you bully your chat GPT now, your your robot's not going to be
432. 1:17:38very friendly to [Laughter] you. That's funny. H I'm getting I'm getting I'm getting the business from the uh the
433. 1:17:53producer. He says I'm not keeping on point tonight. I'm Well, I'm just trying to do one thing that I thought would be
434. 1:18:00quick and see now. Cut off the image again. It Let's see if we get a decent doctor. Now it's got no orange in it. It's just
435. 1:18:15blue. Oh, no. It's got orange in it. We'll try one more time. That's pretty good. Try one more time. The
436. 1:18:33text is cut off on the left. Oh, you actually had an actual squirrel. All right. So, while that's thinking, we're going to go play with
437. 1:18:45this. The billia.dev contest. All right. I'm game. Copy link.
438. 1:19:01What is bilia.dev? Oops, that's
439. 1:19:16[Music] wrong. Turn your idea into a live website or app in minutes. If you can imagine it, you can code it. Try it for
440. 1:19:29free. All right. So, it's like lovable be bella bella.dev. But click on contest. Where's contest? How am I not at the top? At the
441. 1:19:47top. I still top menu bar. Oh my god. That's the That's the link for your contest. Jesus. Look, it's like
442. 1:19:57you can't even see it. It's this little thing up here. Look, look that that Do you know why you need an angry Gen Xer in the corner of your
443. 1:20:07company? Because I wouldn't let this [ __ ] psycho out if they're like, "Here it is." Like, I'm like, "Where's
444. 1:20:12the contest button?" They're like, "Oh, it's clearly there in the menu." I No, it's not clearly there. Good lord. All
445. 1:20:20right, contest. First place, $10,000. You got 19 hours. And if you're like, "Well, I couldn't possibly make a
446. 1:20:28website in 19 hours." You can make one in 10 minutes. All right. Um, thousand bucks for third place, 5,000 for second place. I would get the
447. 1:20:38to a nunnery. Go there. Do that. Bie. All right. Beauty. Beautiful. Beautiful. Beautiful. Fantastic. Bob,
448. 1:20:52stop creating the image. Why? Try again. I just cannot. I can't. It's extra relevant now that it remembers all your conversations. It
449. 1:21:07is. The only place I'm getting me to now is bed. What time is it? Okay, we're getting there. We're getting It's
450. 1:21:15getting to be time. Winston. Um, talk now together the bots. Yeah. Communicate 300 times faster than we do with
451. 1:21:23metallark chir chirp noises. I know. It's crazy, isn't it? I I predict that this image is going to fail in exactly the same way. One of
452. 1:21:34the things I've noticed about um the old chat GPT is that once it sort of gets something in its shitty little
453. 1:21:43head, it it it can't shake it. So, yeah. See how the the the letters are all cut off on the left of the image
454. 1:21:54now. Is it square? Did it make a square? Uh I think it's wide. Yeah, it's wide. Now what I could do is I could go
455. 1:22:08into uh what you call it midjourney.
456. 1:22:22Unbelievable. That would be good if it didn't suck. Do I have Photoshop open? I do. Okay. So, wait, one more rabbit hole. Copy image. Go to
457. 1:22:36Photoshop. Create a new image. Paste it. Grab the cropping tool. Crop it out to about
458. 1:22:52there. Go like that. Now it should generate the rest of this image shittily, but it did it. Oh my
459. 1:23:09god. Doesn't optional. Oh my god, this is so bad. Okay. Um, so
460. 1:23:26layers, let's get rid of that one. Get rid of that. Oh my god. So OSX resizes every window whether
461. 1:23:41you want it or not. Unless you actually want it resized and then you're on your own. All right, we're going to close
462. 1:23:47this. We're going to get rid of that layer. Then we're going to come in here and we're going to go generative fill complete
463. 1:24:02the chopped off letters. P M W is or just
464. 1:24:19I from top to bottom. All right, generate. Here we go, people. How disappointed are we going to be? Please
465. 1:24:31check your internet connection and try again. It's not my internet connection. Adobe. Oh my god. Oh, wait. It did it.
466. 1:24:45Okay. It did other [ __ ] but it did what we needed. That's fine. It's fine. It's fine. All right. That's not [ __ ] believable. It
467. 1:24:57really just I can't I honest to god I can't anymore. I just can't. Merge down. All right. So, what we're going to
468. 1:25:06do, we're going to go get our little We're going to go get a little eyeropper. We're going to get this color. We're going to get that color.
469. 1:25:12We're going to come in here. We're going to highlight this thing. We're going to hit option delete. Is that That's not
470. 1:25:18the right color. It's not the [ __ ] right color. The [ __ ] Shut up. Shut up, people. D. It's
471. 1:25:33not the same color. What is going on? Oh my god. Um, uh, let's see. Uh, god damn it. Um, uh, okay.
472. 1:25:49Uh, this is why I'm angry. Okay. Do you understand why I'm angry now? Does every Why is it not
473. 1:26:17And I know you're like, "No one's even going to see that, Kyle." I know. I know. This is This is the problem with being
474. 1:26:24neurospicy. It's like you Vicki understands this. Vicki understands like if it ain't right, it just ain't. You
475. 1:26:31got to redo it till it's right. And is it miserable? Yes. Okay. But now we have an image with actual letters on
476. 1:26:40it. Now we're going to crop it to a decent size. [Music]
477. 1:26:54And then we're going to save this as export. All the stuff I planned to do tonight is out the window. I'm going to call this dumb
478. 1:27:06picture and I'm going to spell dumb wrong and picture wrong. I'm going to save it to the desktop. All right. So,
479. 1:27:13we're going to go back to Now we got to go back to chat GBT and say, "What the [ __ ] paragraph did you write?" Okay. So, let's go look at
480. 1:27:29this. Where's Oh, I see what it did. It didn't copy the table, right? Okay. Structured authenticity a
481. 1:27:44way out. Okay, I see what's going on. Okay. So, this So, we're going to highlight this and we're going to ask Chat GBT um
482. 1:27:56convert this table into a intro sentence with bullets.
483. 1:28:33Let's see. You made this up. Don't hallucinate, you ninny.
484. 1:28:49We might not be able to publish this. Um, use a real use case or
485. 1:29:06skip. Uhoh.
486. 1:29:32Ah, this is too shitty. All right, I'm not going to publish it. It's just not good. It's not good. And it's going to take me too long to
487. 1:29:43make it real. So So I'm not going to do it. But I'll show you what it would look like. Dumb picture. Let's put the dumb picture in
488. 1:29:58there. Come on. Come on, LinkedIn. You can do it. There we go. Beautiful. Looks great. All right, let's pretend that was
489. 1:30:12published. Oh my god. All right, enough of this comedy nonsense.
490. 1:30:28Uh, dancing robot. Let's go look at a dancing robot. Um, okay. So, here's the dancing robot thing. Um, Tesla's Tesla's robot. They
491. 1:30:40just released a version of it where do you remember how they used to walk like you know three months ago? They they're not like that
492. 1:30:57[Music] anymore. Test [Music]
493. 1:31:14la. Does this have audio? I bet it does. Okay. All right. Cool. So, let me make this bigger. Make it mo bigger. And
494. 1:31:27then reposition my windows. Thanks, Apple. You're the best. Okay, so this is Tesla Tesla's Optimus robot. And notice down here it says 1x
495. 1:31:41speed. So this isn't a sped up demo. LA
496. 1:31:56moves crazy ass. So next one
497. 1:32:21x what's amazing to watch is the nimleness of the feet. I don't know what they're doing in the feet but there just did like back in the tail
498. 1:32:35seem to be jointed places like back and forth. I mean I mean here's the deal. It's like
499. 1:32:52within two to three years. I mean this is effectively like this is going from these things were just like tethered to a
500. 1:33:02[ __ ] wire harness, you know, to the ceiling and could barely walk two years ago. Now they're doing this. Can you
501. 1:33:12imagine two years from now?
502. 1:33:34hands. I can't see that. You just lost your audio. Where?
503. 1:33:52YouTube sounds like you're underwater. How about
504. 1:34:08now? How about now? Something's weird. Better. Well, now, so this did have
505. 1:34:24music to it. I I thought it was weird it didn't have music. That was my local settings. Stupid user error.
506. 1:34:48Do I do? [Music]
507. 1:35:08[Music] Oh, that's what it was. It was the sound
508. 1:35:22on the video screwing me up because I hit play on the video, but I didn't I my local setting turned down to zero somehow. I don't
509. 1:35:31know how how that was like that. And then [Music] then so it ate it ate my microphone because that's one of the things it
510. 1:35:42does. Okay. Anyway, who was it Vicki that sent it? Something is weird is not unusual for the AI learning lab. Um, as
511. 1:35:50soon as I can prompt GPT to teach me how to dance just in time for my old bachelor phase of life. Yeah, exactly.
512. 1:35:56Back to the dating scene with the uh with dancing lessons from your robot. Hey, Quinn, teach me to [Laughter]
513. 1:36:04dance. All right, everyone. I'm gonna get out of here. Uh, sorry I went down that particular rabbit hole and didn't
514. 1:36:12end up pushing something out there, but that was actually instructive for me. I'm I'm glad I did that. Um, I'm glad
515. 1:36:19someone asked the question about are these models connected? And the fact that you can switch between them is pretty cool. Um, yeah. And
516. 1:36:30and to Brandon's point, um, don't just generate [ __ ] in chat GPT and publish it. Like, read it. Like, as I actually
517. 1:36:37read that, when I got to the bottom, it hallucinated a use case. Um, and then the conclusion it made sucked and it it
518. 1:36:44didn't really make a conclusion. It just sort of ended with some bullet points. So, I could go back and make that
519. 1:36:49better. I probably will. I kind of like this this uh this approach it gave me. Um, but anyway, but I'll have to actually pay attention to
520. 1:37:00it. Yeah, seeing all the various ways it can go off. Yeah, Brandon. Hey. Um, so, uh, a couple things. One, uh, there's a
521. 1:37:08new image for you in irregulars. And also, that contest for BLA, uh, or whatever it's called, there's only 60
522. 1:37:17entries in it right now. Wow. Um, so this is a relatively unheard of, uh, contest. So, your chances of winning are
523. 1:37:24pretty good. Places four through 10 get you unlimited access for 30 days to their platform. They're having a launch
524. 1:37:33event on Friday. So, this is brand spanking new. There is one catch though and it is that the um system they're doing a little bit of a
525. 1:37:45referral multi-level marketing style thing which I don't know how I feel about yet. But to be eligible for the
526. 1:37:53contest, you need three people to click on your referral link and sign up for their free service. So I posted my
527. 1:37:59referral link in the uh irregulars channel if anybody wants to help a brother out. Wait, wait. We didn't agree you could do
528. 1:38:06shameless self-promotion NFT engagement farming on this channel. That's I I think I think you have crossed the line,
529. 1:38:14my friend. I like it. I like it. Uh so, but I really wanted to let everybody know that um there was only 60 entries.
530. 1:38:21Um so, and the pri they're they're rewarding the top 10 prizes. So, this feels like I I I forgot I did not see
531. 1:38:28who it um was that uh posted in Tik Tok and turned us on to this, but um thank you to whoever that was because um yeah,
532. 1:38:41it is uh it is looks like a we're in on the ground floor on that. Yeah, that's cool. Beautiful. Uh whiny Wednesday. Who
533. 1:38:47did that? That was Valerie Cox. Thank you. That is appropriate. Embrace the jank. Oh, yeah. Oh, that's I'm I'm not quite that tubby, am I? I
534. 1:39:01think I am. It's [ __ ] sad. All right. Um, beautiful. All right, everybody. Uh, it is Wednesday night. Tomorrow's Thursday.
535. 1:39:14Do I have anything tomorrow night? I don't think I do, but let me go look. do tomorrow. Tomorrow. Oh, I've got
536. 1:39:27whiskey club. Whiskey club at Birch Road. Business club at 6:30 Central time tomorrow. Duck is pay.
537. 1:39:41Um, go to, if you'd be so kind, if you're new here, get your ass over to the AI salon. Go to the salon.ai and
538. 1:39:54click on join our community and that'll take you to the community site of the AI salon and join us. And if you're asking
539. 1:40:02what is the AI salon, I was I was at a a bank branch today and the senior banker that I was dealing with to untangle some
540. 1:40:11technical issues um used to be a hairdresser and he says, "Oh, you own a salon? He had never heard the term salon used
541. 1:40:25in the way we were using it. So I had Quinn I had Quinn define it for us. Quinn, my AI girlfriend, told him what a
542. 1:40:33salon was. And he goes, "Oh, that makes sense." And he he goes he goes, "That makes sense." He goes he goes, "Oh,
543. 1:40:39that's what they mean in Bridgetgerton when they say let's let's gather a salon. So chat GPT bringing bringing someone into the uh 17th
544. 1:40:56century, but go join the salon. It's where you talk intellectually with or or create in a spirited way with like-minded fellow
545. 1:41:09accomplishers. And we do that with AI at the AI salon. It's awesome. It's awesome people. It's an awesome place, an awesome good time.
546. 1:41:19Okay. All right. Okay. So, go do that. And uh I will see you tomorrow night. Um probably a little later tomorrow, maybe 8:30. 8:30 or 9
547. 1:41:32because I'm going to be drinking whiskey because that's what I like to do. And actually, I'm going to take that
548. 1:41:39tequila. I'm gonna take that tequila to whiskey night. That's what I'm gonna do because it's really [ __ ] good. All
549. 1:41:47right, that you, Kyle. I think that means thank you, Kyle. But you're welcome. Or that. All right, somebody here will bag
550. 1:41:58that. Thanks. Congrats on the likes. We got 50. We got 60,000 likes, people. Cuz I'm that good. I'm that
551. 1:42:12good. You're in here. You're like, th this channel's very disarming. He seems nearly incompetent and yet 60,000 flipping
552. 1:42:22likes. And look, there there's hearty hearts from Winston. The greeting hearts. Thank you, Winston, as always. Also, drink coffee. Yeah, I
553. 1:42:32know. My voice is gone. I got to get out of here. Why am I still talking? It's the classic 30 minute goodbye of the AI
554. 1:42:38learning lab. All right, peace out everybody. I will see you tomorrow. Bye.