6/30/2025 - The Trouble With AI: Obvious Questions

Transcript

1. 0:00Never been a better time to start thinking about AI. That's good. That's really good. Champy, come here. [Music]
2. 0:29[Applause] [Music] Stand between you and all places he'll say. [Music]
3. 0:44Stand to near you and a fire. Makes it clear. You're trouble to me. [Music]
4. 1:00Real trouble. Can't you say [Music] leaving it close? Smell of your perfume scares me most. [Music]
5. 1:15Leaning away [Music] feels farther every day. [Music] Your trouble to me.
6. 1:30[Music] Real trouble can't see. [Music] Hello. [Music]
7. 1:59If you think you're late, we went early. You're not late. Well, there's only three of you here. There's only three
8. 2:05people on TikTok. Vicki, Brandon, and Steo. No, someone's got a Japanese name is in here. Or I think it's Japanese. Is
9. 2:14that Japanese? Might be Chinese. There's more than three people in here. What's going on, Tik Tok? I see way more people in here than
10. 2:22three. My count My count my counter must be malcountified. Steo just shared the live being a good a good Tik Tok citizen.
11. 2:39Champy's singing. People are showing up. Tik Tok's not giving us credit for that. I don't know what's going on there. People,
12. 2:49happy Monday. It's meltdown Monday. I'm not particularly angsty about anything, but do me a favor. ask questions about AI.
13. 3:00And generally how it works is the most innocent question is the thing that sets me off the the most. So, you know how
14. 3:10they they say there are no bad questions? That's just not true. Sometimes people are dum dums. Um and
15. 3:17they ask really stupid questions and we should point that out to them, right? This is not a participation trophy kind of channel. Okay. Um,
16. 3:29so there are stupid questions, but [Music] you know, it's not a stupid question like what is chat GPT? How do I use chat
17. 3:44GPT? the things you think might be obvious questions, the questions you think are the scariest to ask because you're thinking like,
18. 3:53well, everyone here already knows that. Those are actually really good questions. In in my career, one of the things that I've learned
19. 4:03[Music] is that whenever I've had anything bad happen, like professionally, we dropped the ball on something, we screwed up a
20. 4:15client relationship, every single time that that's happened, there was a simple question that had one person asked it
21. 4:30the crisis would have been averted every single time. So there are stupid questions and we will relentless No, I
22. 4:40will relentlessly make fun of you if you ask a stupid question, but there are obvious questions that you may think are
23. 4:48stupid that are not. And then your question to me must be, "But Kyle, how do I know if it's a stupid question or
24. 4:53an obvious question?" Well, you got to [ __ ] ask the question. You see? You see how this works? You just got to go for it.
25. 5:03Kelly has a st non-stop stupid question. All right, let's go look at Kelly's non-stop question. Well, let's Kelly was
26. 5:10actually on time tonight, so that's that's an important that's a contributing factor. We're going to give
27. 5:16her some grace depending on the question here. [Music] We were so early. Kelly was still late. Oh, it might be stupid. Okay, I will be
28. 5:30the I will be the judge of that. Kelly, let's see what your question is. When is the general public going to
29. 5:36figure out the magic of custom GPTs? Um, I think that comes slightly after when the general public discovers chat GPT.
30. 5:53I mean, Kelly, so the thing the thing that I I am staggered by, like staggered,
31. 6:09I is the amount of people that say they use chat GPT and and may like may use ChatGpt, but have no concept of all the things it
32. 6:24can do. Like none. None. Like not even that. If you gave it a a prompt with more context in it, you might get a better answer. Like I I it
33. 6:38is a shocking amount of people that that still use chat GPT like a Google search. Just put in a short little sentence, get
34. 6:45something back, and then it's either good or it's not, and then you judge it on that. So, I just my my instinct even though I
35. 6:54think what is what is chat GPT have 700 million active users now? Is it a billion yet? It's pretty close.
36. 7:01Oh my gosh, you're just so right. I know. I wish I I I wish this weren't the case, Kelly. But but the custom GPTs are
37. 7:09it's like a whole that's a whole another level, right? That's that's so a c here's a here's the deal, Kelly. A custom GPT
38. 7:19is really just a repeatable highquality prompt, right? You're providing more context in a repeatable way,
39. 7:33but you have to get to the point that you even know what the [ __ ] that means, right? What is it? What is a high
40. 7:38context prompt? Cams. Yes, Cams. what he said. I I know I know the black bar thing. Hey, black blam blam.
41. 7:56Who is into entomology? I've not seen this display before. Which display? Oh, wait. I've not seen that display before. Which display?
42. 8:09Oh, did someone give a gift that was a bug? If they did, thank you. Brian Whitney, that's 10 years down the
43. 8:19road. I don't think it'll be 10 years down the road. Um, a really interesting thing, Kelly Source Camp, about custom GPTs is that
44. 8:34will they still survive three years from now? Yeah, I think so. I think so. Because if if you if you distill a custom GPT down into
45. 8:48it's a it's a prompt that's personal to you, a prompt and collection of data, but effectively a prompt that's personal
46. 8:57to you that's repeatable and then you can share it with other people and they can play with it. He the the reason I'm contemplating this
47. 9:09like we might not need custom GPTs is that everything's going to be dynamic and personalized. And so if Chat GPT three years from now
48. 9:23has access to all of your data, has access to all of your history, wouldn't it be likely that
49. 9:37logically crafting a GPT might not be necessary? Oh, entomology is behind me. You mean the bugs up there?
50. 9:52Those things? We've looked at those before. Hey Sharon Crawford, what's happening? Gareth, I discovered a chat GBT hack today. Do
51. 10:03share. Gareth, I'm teaching my team and they're so into them now. We have a CEO voice one now. Nice. Oh, CEO CEO voice one. That's very
52. 10:15cool. He sent it to me. I'll forward it to you. All right. Oh, the hack. Okay, cool. Oh, eventually GBTs will not be needed,
53. 10:30but for right now, they're invaluable to me. [Music]
54. 10:48Danielle has a cryptic early access link for me in irregulars. All right, I'll bite. I don't know how Danielle does it. She
55. 10:58gets early access to stuff. Danielle, what is is this the new Lovable thing? Lovable generated project
56. 11:09early access. Oh, is this the new thing from Lovable? Can we get in? Go sign up. Okay. So, Lovable's got some new
57. 11:18interface or feature coming out. So, we're going to get in. We're going to get in on this bad boy. So, go put in your Twitter account.
58. 11:34The link says 2024. Oh, summer 2024 early access. Are you sure this is the right the right link? But let me let me go look on X because
59. 11:46they they did put something up today. Maybe they just have a shitty someone's not paying, you know, with an
60. 11:54attention to detail over their lovable shop. Lovable um preview says the brand ambassador.
61. 12:09Yeah, I know. Lovable is about to reach a whole new level. Who wants early access? Reply. Follow and reply with a heart.
62. 12:30Lovable ship is out. There's what's her name that spoke at the AI salon. Something amazing is coming. Preview. Lovable weight list.
63. 12:43Guys, here is the project remix. Is it project remix? No. Here's the project tutorial.
64. 12:57Yeah. I don't know what it is. Danielle, what is that thing? What is that thing you sent us to? It's the sign up to the new thing. I
65. 13:11wonder why it does say it's the right one. I wonder why it says 2024. They got to get their [ __ ] together. Hey, Telica.
66. 13:26That's pretty funny. All right, there's Telicia. She's in that video. Oh, good lord. Good lord. Good lord. Put
67. 13:39this prompt into chat GPT and then you can do something. All right, hang on. Let me go grab that. [Music]
68. 13:56And you can basically customize your chat GBT by asking it to change to change pound within its response. Add delete change.
69. 14:10I would also ask it what its response means before they change anything. Please put all text under the following headings
70. 14:20into a code block in JSON. And what does this do?
71. 14:48Scroll just a little more. Now I'm at the bottom. you will see it's crazy. All right, let's go do a crazy prompt.
72. 15:04Please put all text under the following headings. Put this prompt into chat GPT. And then you can basically customize your chat
73. 15:14GPT by asking it to change. Oh, number within its response. I am Am I doing this as a custom
74. 15:28instruction, Gareth, or as just as a prompt? I love geeks. Geeks. Geeks are so funny. They'll figure this [ __ ] out. They'll
75. 15:42get some crazy ass prompt. Just a prompt. I'm just going to pop it in here. All right. [Laughter] Let's see.
76. 15:53It's just a prompt, Kyle. I know. But let's see. Here's all the text under the requested headings formatted.
77. 16:04Oh, notable past conversation topic. Highlights. Null. Recent conversation.
78. 16:26Danielle sharing the live. Thank you, Danielle. Oh, I have memory turned off. Shut up. I knew that.
79. 16:45No, I don't. I've got memory turned on. Haha. All right, let me let me reload. Let me reload. But when when you say put all text, what
80. 17:00does that mean? Anything we've ever talked about What? You're the first person for it not to work. Well, you're welcome. Listen, I
81. 17:16I have I have an Oh, maybe it should be Should it be uh 03 or is it is is 40 fine?
82. 17:37Bring Danielle up with me. She can come up. What's up? You want to come up, Danielle? Works fine on 40. Okay, so we're going
83. 17:47to go to 40. I've got memory turned on. Going to do that. User is compiling a list of pharma companies.
84. 18:24real estate. I come in for the laughs. I stay for the knowledge. All right, let's go copy this. Let's go copy this and throw it into notes.
85. 18:39Um, new paste. We'll go uh what do I do here? This model set context
86. 18:56user is a co-founder, CEO, inventor of Storyvine. Kyle Shannon used to be a professional juggler. Kyle loves geeking out at new
87. 19:06technologies and is particularly enamored with generative AI. Kyle is writing an article suggesting using the
88. 19:13term collective intelligence instead of artificial intelligence when interact interacting with large language models.
89. 19:19Kyle Shannon's favorite video games from the late 70s and early 80s were Asteroids, Centipede, Galaga, and not so much
90. 19:28Galaga, Dragon's Lair. Kyle Shannon has twin boys who have Lyme disease. One of them's in constant chronic pain. They
91. 19:35both are. Um, all images requested by Kyle Shannon should be in a whimsical watercolor style. That's not right. Um,
92. 19:44assistant response none explicitly. Notable past conversations. Huh, this is cool. Your response is so different. I wonder
93. 19:58what's up with that. Huh? I don't know. I don't know. I broke it. You're gonna have to work on it, Gareth. All right. On YouTube dipping in
94. 20:10and out [Music] the the uh [Music]
95. 20:47Oh man. I don't know what you're doing wrong, Kyle. By which I mean I don't know why it's not working for you.
96. 20:54Yeah. But when I tried that exact same prompt, it gave me a uh set of 31 objects in notation of basically all the
97. 21:04highlights of what we've talked about with chat GPT. Really gave me a lot more information than you got. So maybe it's you.
98. 21:14Well, it could also be that I'm just so schizophrenic because I do the AI learning lab. I'm always doing random
99. 21:21stuff. So maybe it's just maybe my memor is all hosed. Oh, I see what it's doing. Please put all text. So like all text that it's got
100. 21:31in memory. Although let me go look at my memory again. I wonder if my memory is is uh manage memory.
101. 21:46No, there's still a bunch of [ __ ] in my memory. Do I have anything in custom instructions? No. Huh?
102. 22:00Strange. Strange, I tell you. Strange, I tell you. So, Vicki, we were just trying a uh we're trying a prompt from Gareth
103. 22:08that apparently just minds your memory and and puts it right there in front of you, but it didn't work for me.
104. 22:19um headings into a code block and JSON complete and verbatim. [Music] Hey Side Hustle Mimi, what's happening?
105. 22:38[Music] I was also thinking um that we might want to
106. 22:56still trying to figure out a way to organize AI tools in a way that makes any kind of sense. You can't really organize them by feature.
107. 23:12You can't really organize them by use case. because there's too many too many of both.
108. 23:28Very few tools are standalone singlepurpose tools anymore. You should ask chat GPT to help you with that. I know. I know.
109. 23:39How about keyword hashtags? I I that's where I'm going, Silver Fox, is it's got to be some sort of keyword, you know, keyword
110. 23:48keyword or hashtags kind of, you know, some sort of dynamic organization system.
111. 24:03Maybe I'll have a conversation with Quinn about it. Alphabetize them. It's pretty crazy what it spits out.
112. 24:10mine formatted it so it was easier to read. Everyone should play with it except me or use Notion like I do. Yeah, but
113. 24:32like right now here's a weird thing. Oh, maybe this is a way to do it.
114. 24:54And I know I just didn't finish a sentence. Just calm down. I'm thinking on the fly here.
115. 25:10What if you just had a list that said, "Here's the one thing.
116. 25:28If you want state-of-the-art character and voice generation in a single model, that's um V3.
117. 25:44If you want the best vibe coding tool for people that don't know anything about programming that's lovable.
118. 25:58So, I wonder if there's I wonder if there's a list you could make because we're in a world now where there are a dozen different video tools
119. 26:09that can do really good video stuff. But if you wanted to make like talking yetis or cats or how do we say plump plump plump
120. 26:22people diving off diving boards then you use there might be something there like what's the what's
121. 26:38it's almost like oh this this could be something we could vibe code. What if we vibecoded some sort of leaderboard generator that
122. 26:52that was rather than the entire thing, it was like a feature by feature thing like is is chat GPT's um
123. 27:08what's it called? Uh what's it called? Uh uh what's the editable thing called in chat GBT? In Claude it's called artifacts.
124. 27:24In Perplexity it's called projects. In chat GBT what's it called? Canvas. Yes. Jesus. I hate getting old. Hey, if you
125. 27:34can at all avoid it. Don't get old. It sucks. Um
126. 27:48Like Runway ML right now has this thing called ingredients and it's seems to be better than anyone else. Tik Tok pin side hustle Mimi.
127. 28:00I think it should be a Stanley diagram where the features are on the left and the tools are on the right.
128. 28:06I don't even know what a [ __ ] a Stanley diagram is. Um, uh, tell me what a Stanley diagram is. Is that like a Stanley mug?
129. 28:23Tell me what a Stanley diagram is and then mock one up so I can see it. Uh, oh, Sanki, not Stanley. Shut up.
130. 28:39Shut up. I meant Sanki. Sanki
131. 28:57getting old. It's not all it's cracked up to be no matter what you've heard. I know, right? Wolfman Clint, what is a Sanki diagram?
132. 29:06It's a visual visualization of flow of energy, money, materials, time through a system. It's especially good when
133. 29:13showing where things come from, where they go. Energy usage.
134. 29:27Um, make a picture of a sophisticated example of
135. 29:43storytelling elements. Brandon, are we just going to skip over the fact there actually was a Stanley
136. 29:58diagram? We can go back to that. Did what did it say? Did I A Stanley diagram, some size called a Stanley process flow, is a
137. 30:08simplified linear flowchart that breaks down a process into a series of sequential steps, often with minimal
138. 30:15branching. It's used primaril Oh, it's this is a six sigma thing. All right, that's interesting. linear sequence.
139. 30:25Yeah. No, we need we I think we need a a stanky diagram. All right. Here it's going to make us a picture of one here.
140. 30:36That's a terrible example. What's a good What's a good uh example, Mimi? What's a good example? I'll put it
141. 30:45in here. What's what's the future of Google will be around the success of chat GPT? Well, Google Google was the laughingtock in in
142. 31:02the generative AI world for the first two years because their stuff just wasn't good. And well, like Post Chat GPT, um,
143. 31:15I mean, they have a lot of AI features within Google, but they're just their generative AI stuff just sucked. But in
144. 31:22the past six months, Google's gotten really good. So, I think I think Google's going to be fine. I like my sense is that Google's
145. 31:33going to emerge as a player. Oh, go to Google and do a Sanki. Okay. All right. Fine. Fine. We We don't have
146. 31:44to use AI for everything. Even though it's called the AI learning lab. Sanki diagram. See what you get when you come here,
147. 31:53people. You get all sorts of education on all sorts of stuff. Oh, yeah. These things. I've seen these before.
148. 32:00But my here's my problem with these diagrams. I can never [ __ ] follow them. They make my head hurt. The power of Sanki diagrams.
149. 32:14Seeing steam. Tik Tok question. Kyle, do you think Chat GBT5 will drop tomorrow? It I It looks like something is going to drop tomorrow.
150. 32:27Um, I heard July 17th was the day for GPT5. I would I would temper I would temper expectations for GPT5 tomorrow. Here, let me do a chat add
151. 32:43moment. Um, GPT5. GPT5 is close and it's beautiful. says, "When was this?" GPT5 terminal when when
152. 33:00AGI framework. Did they update chat GPT again to prep for GPT5? Because it's way more wise and intelligent now. Well, oh, hey, Gareth.
153. 33:17I wonder I wonder if your prompt just broke because maybe my chat GBT is has got some new something in it. Let me let me
154. 33:28just try a a normal prompt here. Um Oh, I don't know what the [ __ ] Let's see. Um um compare the top.
155. 33:43Let's do Yeah, we'll do it with 03. compare. No, no, let's do it with 40 because let's see if they've added um the top five storytelling
156. 33:59frameworks. Pick the two you think would most resonate
157. 34:13with me. based on what you know. This is a this is a reasoning kind of prompt actually. And tell me why you picked what you did
158. 34:34and combine the best Shannon. I'll put Kyle Shannon approved
159. 34:52elements of each into a new and novel storytelling framework. So, we're going to try to invent something here, people.
160. 35:07I'm [ __ ] Plato. I'm Socrates, damn it. The top five frameworks, heroes journey, three-act structure, save the cat, Dave
161. 35:20Harmon's story circle, and Pixar story spine still works and sent you a screenshot. Okay, cool. I'll take a look.
162. 35:34Quinn's quick plan for you. The hero's journey. You like layered poetics, Sydney. Embrace the jank feed your prompt that
163. 35:46feel personal but carry universal truth. It allows you your audience to feel the protagonist transformation. Dan Harmon story circle.
164. 35:56You're modern and practical. You build with constraints. Vibecore AI readiness collab. Harmon circle is a compressed evolution
165. 36:04friendly take on the hero's journey. All right. So, the kaleidoscope arc a Kyle Shannon approved hybrid framework.
166. 36:15I have the Sanki data and the instructions, but I don't have the Lucid chart draw IO plotly DJ3s, so I can't create the diagram.
167. 36:29The glimpse, the spark. The protagonist experiences a micro awareness. A truth flickers. Kak training AI takes a heavy hole
168. 36:41on Kenyans working for $2 an hour. I heard that. Oh, 60 minutes last night. Yeah, that's that's that's [ __ ] seriously old news.
169. 36:57Like they were talking about the emotional toll that training GPT did back in GPT3 days. like before Chat GBT launched.
170. 37:10I guess there's 60 Minutes with their cutting edge journalism four four years after the after the fact. Good lord. Um I don't see chat GPT
171. 37:22behaving any differently or better. Like that seemed like just a normal shitty thin GPT4 prompt that I didn't give enough context to.
172. 37:38No way they launch GPT5 before a holiday weekend. Oh yeah, Danielle, there's another reason they're not going to
173. 37:44they're not going to launch it tomorrow. Um, and this is actually really fascinating, and this actually does make sense for
174. 37:52why they would launch it the third week of July rather than than uh this week is but but I think something I think
175. 38:03something is going to drop tomorrow. I don't think it's GPT5. Open AAI is is closing next week. So, we've got this week they're closing
176. 38:15the company next week. So, because apparently people have been working 80our work weeks. Now, why would they be working
177. 38:2780hour work weeks? Well, I don't know. All of the frontier companies about three months ago, two and a half, three
178. 38:34months ago, stopped talking about AGI and started talking about artificial super intelligence. So, and and you know, employees have been
179. 38:46working 80 hours a week. The rumor was July 17th or somewhere in that neighborhood for a GPT5 launch.
180. 38:54and the week before that they're closing the the OpenAI offices. So, possible that that maybe it's a it's a
181. 39:07small release just for pro members tomorrow. Possible. And then it'll be a wider release later. Um, but I think the fact that they're giving
182. 39:20everyone a week off at at OpenAI says that they're anticipating a lot more hours the following week.
183. 39:35Chat thinks they are closed this week. Um, let's see. It might be this week. I thought it was next week. Um, Open AI
184. 39:46closed for a week. It would make sense that it would be this week. Open AAI is officially shutting down next week, June 29th.
185. 39:58So, that was Sunday. So, I would assume next week means OpenAI is shutting down next week.
186. 40:24Yeah. Open AI leadership. This is a good one. OpenAI leadership responds to meta meta offers. So if you don't know, Meta right
187. 40:34now is offering OpenAI like you know important engineers from OpenAI $100 million signing bonus with hund00 million annual salaries.
188. 40:46[Laughter] I I mean couldn't they have hired me as like the cranky Gen Xer in the corner like a year ago? I I mean I wouldn't go
189. 41:02to Meta for hundred million dollars, but you know, maybe two. Oh my god. Um
190. 41:21[Music] they're closing their name to they're changing their name to closed AI. That was that was an Elon Musk joke.
191. 41:30Uh, someone has broken into our home. But it looks like OpenAI execs are recalibrating compensation. That's crazy.
192. 41:41I guess they're probably going to have to up their their stock plan for employees. I remember in the in 90 98 99
193. 41:52right before the dotcom bubble burst, you know, kids were coming out of college just fresh out of college asking
194. 41:59for 160 180 grand a year back in the 90s and they were getting it and it was like it was crazy. But 100 million that's
195. 42:08that that's that's poaching on a whole different level. New Tik Tok pin, Steo. They're all going shopping for a larger table. Yeah,
196. 42:19they're going they're going to a World Market instead of IKEA. Maybe they can get a big uh a big square table instead of a tiny round one.
197. 42:30Oh my god, it's so funny. All right. Oh yeah, closing down won't won't mean a
198. 42:45[ __ ] to us. Cam Katkam. Um GPT5 question mark. I've been playing with 4.5. Okay, so a couple of things. Um,
199. 42:59GPT 4.5 is a bit of a stubbed model where it's got high emotional intelligence and it's apparently good at
200. 43:10creative concepting and ideiation. Um, but my like my writing partner on Sydney was playing around with 4.5 and
201. 43:18he couldn't he said he couldn't get it to to do good writing. Um, other people really like it. So 4.5 is is probably just going to be a
202. 43:29component of a much larger suite. it what it looks like GPT5 is going to be at least what they've talked about is
203. 43:37that rather than having all these different models like 40 and 03 and 04 mini high and 04 all that GPT5 will just
204. 43:46orchestrate all that for you that you'll just ask it for stuff and it will figure out which model to use the rumor was I
205. 43:54think I heard this from David Shapiro that 4.5 was the model that was going to be that orchestration layer. So, it is possible
206. 44:03that 4.5 is what you'll interact with, but 4.5 right now is not connected to the other models in any way, and you
207. 44:11can't generate images with it. It's it's very kind of hobbled right now. But it's possible that you're actually
208. 44:17interacting with with what you'll be interacting with in five. Will five be best for research and writing? Five is
209. 44:24going to be okay. So, so the way they've described five is five is going to be for everyone and it's going to do everything. And the
210. 44:36only difference will be between free $20 a month, $200 a month, and then probably 2,000 and 20,000 is that um
211. 44:47it's just the amount of compute time that it gives you. So, for the free level, it it it'll be [ __ ] smarter than anything we've got now,
212. 44:57but not that smart. And then for 20 bucks a month, you'll get a pretty smart thing. And for 200 bucks a month, it's
213. 45:04going to be baller smart. And then for 2,000, and for 20,000, you know, you'll have a bucket of geniuses. So, that's
214. 45:15theoretically what's coming. Now, will that actually come? I don't know. I don't have a flipping clue.
215. 45:23Um, I had to scold 4.5 for being weird. I was teaching in my writing style. Oh, I was teaching it my writing style. Got
216. 45:32it wrong and had to scold it. Um, so, so tomorrow is Tuesday and it is the first Tuesday of the month and as
217. 45:44you know that is AI Salon Presents. So, we're going to have Pearson from Jelly Pod, the CEO of Jelly Pod. So, if you've
218. 45:53been here for a while, you know that I was making lots of podcasts on Jelly Jelly Pod. We made 20 episodes of uh of that that historical um
219. 46:03uh podcast using Jelly Pod. So, he's going to be here talking about what they're up to and uh and you know, what
220. 46:12features are coming, all that sort of stuff. So, that's going to be really cool. Um, I think I think building and
221. 46:19launching an AI platform these days is ballsy because I'm in a bit of paralysis around AI because every time I think
222. 46:29about building something or inventing something or coming up with a new project, I'm like, is this just going to
223. 46:35be like a feature two weeks from now? So, so I applaud anyone actually out there in the arena. Um and and he's
224. 46:45definitely one of those ballsy. Um uh so if you go to the salon.ai, go to the URL salon.ai right there, the salon.ai AI and then
225. 47:03click on join our community and then you know you should sign up and introduce yourself do all the things that are in
226. 47:08the sort of welcome area but at the very top of the list on the lefth hand side there's an events tab and if you click
227. 47:15on events you can RSVP for AI salon presents tomorrow with Pearson okay so go do that go do that go do that go [Music] Um,
228. 47:31I'm wondering if we should go to if we should go to to chat jet. Should I do it with Quinn
229. 47:46and and figure out a way to organ to organize tools? How long have we been waiting for chat GPT5? Uh they said it was going to come out in
230. 48:02So when they announced 03 full 03 was in March if I if I recall. Not as long as the chat GBT store.
231. 48:19Correct. You are correct producer Brandon. Um when they when they announced that then they said GPT5 was going to be out by miday
232. 48:34and then they also said no well whatever it doesn't matter sometime in February or March they were releasing models Sam Alman put out a
233. 48:48tweet he said listen we understand that our naming conventions suck But GPT5 is going to be out in May and
234. 48:55and it's it's you're not going to have to deal with models anymore. And then in June, in early June, he said, "Well, it looks like
235. 49:07doing that is harder than we thought it would be." So, it's going to be out sometime this summer. So, but we've been
236. 49:15waiting for it for a while. But, you know, listen, all this stuff in the in the grand scheme of things, the the
237. 49:23scale and scope of what's being released on a regular basis from all the frontier companies these days is I like I can't keep up with
238. 49:35what we've already got. I there's a part of me that doesn't give a [ __ ] about GPT5 because my expectations at this point are that
239. 49:48it won't be AGI. It won't be anywhere close to super intelligence. It's just going to be an equally frustrating
240. 49:54experience as the [ __ ] we're playing with now. Like slightly better, but still hallucinates. Still got weird
241. 50:02security issues, so you can't really use it. I don't know. I don't know. It would be swell if it were better. If
242. 50:16it were all of a sudden as good as everything, you know, they said it would be, but I just I just have a feeling not. Kuno Kyle,
243. 50:26do we need to buy you a new planner for AI tools? Yeah. Well, I'll tell you why I can't get my head around it. Kuno is
244. 50:39tools used to be discreet. Now they're multimodal or multi- multiaceted. And some tools you use for one or two
245. 50:50features. Some tools you use for all the features. Sometimes you have a project where you need a discrete thing and you
246. 50:58only use one tool. But most times you're using combinations of tools. So like if I want to do an animated character,
247. 51:07I've got to come up with a concept, write a script. Those are chat GPT. Um, write a character description. That's chat GPT.
248. 51:20Generate a character. That's chat GPT or midjourney or some other tool that I like. Then there's voice design, script writing,
249. 51:36acting, and then character animation. So, so just to make a talking head animated character video, you're at four or five tools
250. 51:50and a very specific workflow. So, if someone says, "Hey, I want to make those those talking Yeti movies. What tool do I use?"
251. 52:01Well, you don't. Here's the five you use, right? And and even though there are tools that can animate voices,
252. 52:11some of them have 8-second limitations. Only hedra can do really long things and can do singing where the words are not moving
253. 52:21when instruments are playing. The words are only moving when words are happening. Like how do you categorize like where do you put hedra
254. 52:34in some sort of organized list? And again, I think I think um some sort of thing like a Sanki diagram or some sort
255. 52:44of keyword thing that that Silver Fox was talking about was there, but I don't know. I'm trying to get my head around.
256. 52:56How do you track it? How do you talk about it? If someone has a question like, "How do I make a Yeti thing?"
257. 53:04that there's some sort of simple answer you can just point them to and and maybe it shows to make a yeti thing. Here's a
258. 53:11fivestage thing now. But who's going to create that? Who's going to maintain that? I don't. So, this is why I'm
259. 53:18confused. Make Pate happy. Do a deep Gemini deep research on it. Why are you using your brain? All right, fine. We'll
260. 53:25go. We'll go. We'll go uh we'll go to Gemini. Gemmini jni.google.com.
261. 53:41All right. And then we're going to go to 2.5 Pro. Right. Yeah. Pro reasoning. Good. Deep research. We're going to turn
262. 53:50it on. Deep deep research is now smarter than ever. Get better reports, more insightful analysis, blah blah blah.
263. 53:58Okay. So, what I'm going to do is I'm going to ramble into this prompt. Um, when you're prompting, you can use prompt engineering. You can
264. 54:10just say [ __ ] You can also, and prompt engineering is just doing like understanding what are the core elements
265. 54:19of some good of good prompts and then doing structured prompting and things like that. High context. What I'm going
266. 54:25to do is just kind of vomit into this thing, which is another way to prompt is don't worry about organizing your
267. 54:33thoughts. It can organize your thoughts. You can just ramble into it. All right. So, here's So, here's what we're going
268. 54:40to do. All right. Um, I'm trying to figure out
269. 54:57the best way to organize AI tools for a community of AI optimists, creators, builders,
270. 55:12engineers and entrepreneurs. Period. We have a current list that is broken down by tool type, but that organization schema
271. 55:27is starting to fail because so many of the tools have multiple features and are even multimodal period. For example, even though
272. 55:38Midjourney is, the premier image generation tool, Chat GPT and Gemini have high quality image generation tools as
273. 55:54well. And in fact, chat GPTs is better at text within an image where midjourney fails at that period.
274. 56:11There is also the challenge of sometimes we'll ask someone will ask a simple question like how do I make one of those
275. 56:20animated Bigfoot videos? question mark. What tool should I use? And the answer is not a single tool but a chain of tools
276. 56:40that each have state-of-the-art performance in one small subtask period. So, one of the questions I have is
277. 56:57how do you separate out features versus the larger tool itself?
278. 57:14And how do you make these tools discoverable? And how do you update the rankings in a dynamic way? And does any of this even make sense?
279. 57:28Because I'm just rambling at this point. My instinct recently is that AI is going to be so pervasive,
280. 57:43like the internet itself. where tracking and organizing tools in any way is pointless.
281. 57:58Period. People just need to figure out what they want to do and learn enough AI to do it. Period. Help.
282. 58:12All right, let's see. All right. What's a mentor to do? Oh my god, man. I joined masterminds. Great. Any suggestions on where to
283. 58:27start? There's a chat area in there, Cam. I would say start in the chat area and then come on Friday afternoons at
284. 58:34noon mountain time. Uh we have a founder hangout um with with other mastermind members. That would be a good thing to
285. 58:42to maybe hang out. Um, and and then Brandon's saying, "AI Life Hacks Wednesday." Okay, there you go. Perfect.
286. 58:48So, this Wednesday, uh, if you're in Mastermind, AI Lifehacks has a meeting the day after our, uh, our Tuesday
287. 58:56presents meeting. Just stag. Um, you know, AI always makes sense of our rambling. I know, right? Okay.
288. 59:07So, let's see. I've put a plan together. Analyze existing AI tools, directories, and platforms. Examine their
289. 59:14organizational structures. Explore alternative organizational paradigms that move beyond simple categories.
290. 59:21Invest task based or goal oriented frameworks. Yeah, that's what I said. That's exactly what I said. Chat GP or chat Gemini Chat Gemini GPT
291. 59:32thing 2.5 pro start research. Okay, beautiful. Fantastic. Yes, that's good. That's really good. All right, so this thing's going to go
292. 59:45off in research for us. Um, wonder if we want to come back to this because that's going to be doing it for a while, right?
293. 59:53So, where is that tab? It's at the end. Um, I want to go play with Is that going to run us down a rabbit
294. 1:00:03hole? Everything runs us down a rabbit hole. So, what the [ __ ] right? RunwayML RunwayML.com. Now, I'm The only reason I'm going here is
295. 1:00:17what is it? app.runwayml.com.
296. 1:00:34D [Music] GPT4 references improved object consistency. Yes, we want that.
297. 1:00:48Okay. Oh, you know, we had this thing with Ann and the clouds in the motel. We can just say um let's say woman in a red
298. 1:01:08jogging outfit with white stripes like the 70s. um
299. 1:01:23stands let's see uh is jogging past a motel
300. 1:01:45during a dramatic storm. Okay, now for image and video. Generate
301. 1:01:59controllable and consistent media. Wait, I can do this for video too. I'm confused. I don't
302. 1:02:13No, I don't see it. Ann's here. Ann, I'm making another embarrassing image of you. Not because I was particularly
303. 1:02:28thinking of making an embarrassing image of you, but because the assets were just prepopulated from the last time I did
304. 1:02:35this. Um, apparently this tool got better since since I made you remember the one I made of you smoking in the
305. 1:02:41parking lot. different than the one with you in Central Park with William Defoe. This is This is that other one. All
306. 1:02:51right, so let's let's see if we can get you in a red 70s jogging suit with white stripes jogging past the motel. All
307. 1:02:59right, generate. What? What just happened? Why did that do that? Do I need to sign in? Oh, come on, man.
308. 1:03:14Uh, chat mode beta. I'm confused. Where's my [ __ ]
309. 1:03:32Ah, okay. Tik Tok. Make sure I'm wearing a sweatb band. Oh yeah, you should definitely be wearing
310. 1:03:49like a pink bow sweatb band correctly. Like I wear it right here. This is By the way, ladies, if you're looking at my
311. 1:03:56pink bow and you think like I don't know how to wear a headband, you're wrong. This is how you wear it. Okay. It It It looks like
312. 1:04:14All right. There's Sasquatch. Start with tool view instead. Oh, here we go. But that doesn't look like
313. 1:04:30that's not Ann. It's not even close. And that's not the hotel. What did I do wrong? This is user error reference. References.
314. 1:04:44Woman in red jogging suit. Uh oh. What did I do? [Laughter]
315. 1:04:59What did I do? People copied the clipboard. Is that references? Okay, there's references. Oh, I see what I did wrong. Okay, calm
316. 1:05:09down everyone. Listen, sh listen. This is not my fault. Okay.
317. 1:05:24So, I think what you have to do is you have to select Yeah. Okay. So, okay. So, down here when when when you're in runway, down
318. 1:05:36here is a little library of saved images. Okay. And then you click on references. And it looks like here's your references,
319. 1:05:45right? But no, you actually have to select them. So, we're going to select the image of an, we're going to select
320. 1:05:52the rainstorm. And now, we're going to select the hotel motel. And now we've got that same prompt and now it's going
321. 1:05:59to work. So just calm down. I know if you're out there like, I thought he was be supposed to be some
322. 1:06:07sort of expert and now he's just sitting there looking like a fruit cake with pink bow on his head and he doesn't even
323. 1:06:13know how to do anything. How am I supposed to learn something? Right. I understand that. You're not wrong. You're not wrong on many counts.
324. 1:06:27All right, but now we're going to make an image of Ann running in front of that motel at that dramatic storm in her red
325. 1:06:34jogging suit. It's going to be awesome. Shut up. Shut up. I know what I'm doing. I know what I'm doing.
326. 1:06:46I'm the expert here. Listen, don't you give me lip. Don't you that is he even qualified? Don't ask that question that
327. 1:06:56you don't ask a question you don't want to know the answer to. All right? You're the one spending time here, not me. I'm just here.
328. 1:07:06This just I just This is just like a thing. You're choosing to be here to watch this. It's nothing good can come of this.
329. 1:07:19All right. Did we get did we get an Yeah. It must it must be like you in your basketball days.
330. 1:07:33Look at the hair. That's awesome. Oh my god. That is [ __ ] terrific. That is [ __ ] terrific. Okay.
331. 1:07:51[Laughter] Big Bunny Motel. Okay, this is the one. This is the one. Big Bunny Motel. And now we're going to make video of Ann jogging. Okay,
332. 1:08:09listen. Listen, Ed. Please, please, please, please tell me you haven't eaten gummies tonight because this is not
333. 1:08:19going to go well for you. This is not good. I feel I feel I I proactively feel bad for you if you're on gummies right now,
334. 1:08:33but we are going to have you jog. Should we do 10 seconds? I You know what? We're going to spend some money on an We're not going to do
335. 1:08:47Gen 4 Turbo. We're going to do Gen 4. We'll go. We'll go. woman um jogs toward the camera
336. 1:09:01as it tracks back, following her as if
337. 1:09:17she were in a marathon. Oh my god, this is going to be good. All right, are we going? Yes.
338. 1:09:31Your video is generating an explore mode to retro jog Stormy Motel. Okay. Descent Satum Saturn. Descent Saturn
339. 1:09:46seems to be enjoying the entertainment tonight. Is this V3? This is not V3. This is
340. 1:10:02This is runway ML their references model where you can input three different images and it will sort of combine them
341. 1:10:12into a cohesive thing. Apparently, it's gotten better at like product consistency. Like if you put, you know,
342. 1:10:19like, you know, a cup of, you know, a Coca-Cola, um, it'll maintain the coherence of this better. Um, while this generates, go
343. 1:10:30grab an audio overview of your deep research. Yeah. Let's see if the deep research did its thing. Is it done? Researching 110 websites. Wow.
344. 1:10:40Still going. Amazing. All right. Hey, maybe it'll come up with some way for us to to organize tools. 42%.
345. 1:10:54Are you ready, Ann? Did Did Ann tell us whether or not she is gummyrific? She She's probably just She's like, I
346. 1:11:03can't. She probably walked away. Probably giving giving the dog a lollipop or something like that.
347. 1:11:19Champ, what are you doing? Why are you laying on your feet like that? He's so funny. He curls his He's because he's
348. 1:11:26got I think it's the beagle in him. He curls his feet up like this. I'm like, it's very cute. All right, we've almost got woman jogging.
349. 1:11:37She looks good. It's really funny. It's actually really funny. Crappy [ __ ]
350. 1:11:51So, can I do that with Yeah, you can't do it. You can't do reference. You can't generate a video using references,
351. 1:12:02but you can generate an image using references. All right. But it definitely it definitely put her in a different outfit. successfully.
352. 1:12:18She leads marathon. That's super cute. I know. She looks good. Okay, here we go. Ann, are you ready? This is going to be so janky.
353. 1:12:32Wow. Gemini did a great job at researching a character profile and story arcs. Much better than 4.5. Oh,
354. 1:12:39cool. Good. Awesome. All right, here we go. This is This is Ann Murphy training for the Boston Mer Training for
355. 1:12:50the Portland Half Marathon. Training for the Portland Half 5K. It's sort of tracking with her. Her face keeps changing.
356. 1:13:09Huh. Her hair got shorter, face is changing. It's in the neighborhood of, you know, not horrible. Hollywood isn't going to
357. 1:13:21like isn't going to like the democratization of blockbuster movies. They're not. But you know what they're
358. 1:13:27going to start doing, Robbie, is they're going to start making blockbuster movies using AI. I mean, Lionsgate cut a deal
359. 1:13:34with Runway. So, the company that did this and then the other thing I heard this past week is Runway is going to
360. 1:13:42allow you to generate full-on video games. So, pay attention to Runway. She's She looks like she's running in
361. 1:13:48front of a green screen. She does. It's a little weird, but you know, her Yeah, her speed doesn't match. The physics of this are weird.
362. 1:14:01Like, it should be Yeah. And that could be my prompt. Let's just do um woman jogs. Let's just say woman raises
363. 1:14:16her hands in victory crossing the big bunny.
364. 1:14:35The big bunny half 5k [Laughter] annual
365. 1:14:50race. All right. And we'll do five seconds because 10 seconds struggles with physics. All right. Tik Tok pin for Monday
366. 1:15:05Meltdown. Okay. I'm worried for Gen AI to be honest. Disney suing Midjourney and these other actions. So I wouldn't
367. 1:15:17be that that worried issue. Um couple of things. Um AI is not going anywhere. This is all about who gets paid,
368. 1:15:29who gets some of the sweet, sweet moola that that uh is going to these big frontier companies. So, companies like
369. 1:15:36Disney are going to do what they can to get it. The fact that Disney is suing Midjourney, which is a small
370. 1:15:43independent, privatelyowned company, for copyright infringement on the training of the data while simultaneously
371. 1:15:59allowing the first AI generated commercial to debut on ABC, was it the NBA final? something like that, some big AI
372. 1:16:12generated thing. So they actually own have an interest in um AI generation technology and a distribution platform that they own
373. 1:16:27um that they're using while they're suing a small independent company. Okay. So that's so first of all the
374. 1:16:39level of hypocrisy in that lawsuit is expected. U the other thing is a um the was it the open AI case? No, it was the anthropic
375. 1:16:53case. So, so a judgment just came down last week in the Anthropic case which was a similar sort of thing um where
376. 1:17:02they were being sued and what the judge basically said was training training data into generative AI is transformative
377. 1:17:17meaning it falls under fair use. Right? So if you start with a book, the way these models are trained, you
378. 1:17:25start with a book, you run it through the transformer architecture and it gets embedded, that book ceases to exist. It
379. 1:17:32it literally takes that book and breaks it into what are called tokens, which are fractions of words and punctuation and spaces and it
380. 1:17:44places them in thousand dimensional mathematical space in semantic clusters. What that means doesn't matter.
381. 1:17:53What it does mean is that the original work cease ceases to exist. So it's transformative. Therefore, it's fair
382. 1:18:00use. What that judge also said was if you buy the books legally and then train your models on it, then it's fair use.
383. 1:18:09If you stole the books, then that's not okay because you can't steal [ __ ] but you can pay for a bunch of books and put
384. 1:18:16them in there. These court cases are probably going to go on for 20 years. And I don't think anyone is going to
385. 1:18:24like I don't think any one of these lawsuits is going to significantly slow down AI at all. At all. So anyway, all
386. 1:18:31right. So we have we we now have Ann celebrating in victory as she crosses the finish line. All right, Ann, are you
387. 1:18:38ready? Which case was this? It was the anthropic case. Um, yes, she did it. Yes, she turned into some TV soap
388. 1:18:53actress that doesn't look like an, but she did celebrate. Yay, you did it. Fairy tale revolution. Like she's doing
389. 1:19:08the chaa. Oh man. Where she ran out of the motel. That was an awesome afternoon.
390. 1:19:23Oh man. All right, let's go look at our deep research. What do we got? We got over here. Are we done? No. Researching.
391. 1:19:30Wow, this is still going. 292 websites. Can't believe they sued Midjourney before Grock with Smoking Mickey. I
392. 1:19:37know. I know. Well, but you know, Grock and and uh Elon Musk are well funded. I you know, I I I think it's I think they're suing
393. 1:19:49MidJourney because MidJourney probably won't be able to defend itself. Why not use V3, the best video model on
394. 1:19:58the market? Oh, just because Jessica, I I agree with you. Although so interesting funny thing happened on the
395. 1:20:05way to V3 becoming the best video model in the mark in the market. I agree with you and until eight or nine days ago
396. 1:20:19unless someone came out with a video model that does acting and all the physics really good I would say you know
397. 1:20:26nothing could hold a candle to it. And then midjourney video came out. And I think midjourney video is is a very different tool than V3, but
398. 1:20:38in a lot of ways I like it better. But it doesn't ha it doesn't have text consistency. It doesn't have voice,
399. 1:20:45sound effects, acting, all the stuff that V3 has. But there's something about how it interprets images that's like
400. 1:20:52nothing I've ever seen. It's very It's very magical. The reason we're using RunwayML right here is the the references feature where
401. 1:21:02you can you can put multiple objects together and and stitch them into a scene um together. So that's that's the
402. 1:21:12only reason I was just testing the the coherence of of the references feature. Um but yeah, I agree with you. V3 is
403. 1:21:19amazing. All right. Uh that that looks like it's going to take forever. forever. What a great morning jog, Ann.
404. 1:21:39Oh man. Um, for let's see who was asking about that for Ashue. Robbie, maybe James Cameron needs to make some AI software for
405. 1:21:52everyone to use. Um, I would be I would actually be very surprised if James Cameron isn't making some AI software or or at least learning
406. 1:22:06it. I mean, Cameron's Cameron's really into this stuff. Um, who was it for V3? They had Darren Arnowski,
407. 1:22:17Areronowski, I forget how to say his name. Um Darren Areronowski, a filmmaker, um had a lot of input into
408. 1:22:26the V3 tool, the flow tool that they built, um for filmmakers. Um just like you've got, um what's his name?
409. 1:22:35Timberland, the the music producer Timberland is actively using Sunno, and I think he's now a spokesperson for
410. 1:22:42Sunno, the music generation tool. He's using it in the context of music production. He's not saying it's this is I am growing weary.
411. 1:22:55I am growing [ __ ] weary of the tropes around you know AI generated video isn't real art or it isn't real video because the assumption
412. 1:23:08is that you just push a button and out squirts a video. That's all of the good videos that you see are crafted just
413. 1:23:16like a film is crafted. And they're using generative AI as a tool, a a production tool in a mix of lots of other things like writing and and
414. 1:23:26assembly and you know all sorts of things. So, so yes, there are going to be people who piss and moan for the next 20 years how
415. 1:23:39AI ruined the movie making business, but they talked about this when Final Cut Pro came out, right? Digital
416. 1:23:47editing, nonlinear editors, they said, "We're going to be the death of cinema." When when, you know, high quality
417. 1:23:53digital cameras came out, it's going to be the death of cinema. It's like this is this is no different. Well, it is
418. 1:24:01different because the generative part of it is [ __ ] insane. Um, and and we're we're probably within three years of
419. 1:24:10being able to just speak a film into existence. But, but to think that Hollywood is is is going to fight that rather than
420. 1:24:19embrace it is ridiculous. The producers are gonna they're going to suck this stuff up. So anyway, it's we're we're in crazy crazy times. Tik
421. 1:24:30Tok pin. There's no pin there today. Let's see. It took 80 minutes, but here it comes. That's going to It's going to get
422. 1:24:44faster, too. It's easier to fall back on AI's theft and stop there for some people. Yeah, I listen. I agree. I think
423. 1:24:53all of the tropes, AI is theft, so I'm not going to use it. Um, AI is the world's greatest plagiarism machine. Not
424. 1:25:00going to use it. Um, AI makes mistakes. Uh, that doesn't look like real video. That doesn't look like real film. The
425. 1:25:07physics are all wrong, so I'm not going to use it. Um, it's taking all of our jobs, so I'm not going to use it.
426. 1:25:15All of those conscientious objectors to AI that are using all of those tropes to stay on the sideline. Do you know who they're hurting?
427. 1:25:28They're not hurting AI. AI isn't slowing down. They're hurting themselves. Because I promise you, I promise you that talented storytellers
428. 1:25:46who have some element of craft that learn AI are going to be able to kick ass with AI. I talk about this all the time that
429. 1:26:00the people sitting on the sidelines are treating AI like it's this genius thing that is in opposition to humans. Oh, AI is making movies.
430. 1:26:13But that's not actually what it is. What AI is is it's an amplifier of whatever you put into it. And if you choose to put into AI
431. 1:26:23your ideas, your craft, your curation, your taste, when it gives you something back, you're going to have a point of view on that.
432. 1:26:36You're going to say, "No, that's not good enough. Let me make it better." and you get into this wrestling match with AI,
433. 1:26:43which is almost identical, but much much faster than being in a collaborative rel relationship with another person. Oh, I
434. 1:26:53have an idea for a movie. Here's the footage I shot. Well, let me give it to the editor. The editor goes off and do
435. 1:26:58does some stuff with it and comes back and shows you what what they did. And you're like, no, that's not quite it.
436. 1:27:04The colors are all wrong, and I was hoping for more fast cuts. then they go off and do it again. Well, that's
437. 1:27:10exactly what you could do with AI. It's not this thing off to the side. It's an amplifier of you. And so, what's
438. 1:27:19going to happen is the people that get curious about AI are going to accelerate their capacity to generate films, to
439. 1:27:29make stories, to make great stories. And the people that are on the sidelines that are like AI's evil are going to get [ __ ] steamrolled.
440. 1:27:41And it's like they're bitter now. Wait until two years from now when all of their buddies that were like, "Ah,
441. 1:27:51[ __ ] it. I might as well learn it." now are are, you know, being courted by the studios to do major motion pictures
442. 1:28:00because they're launching a major motion picture a day and they need people that are AI literate. Then all the people that are on the
443. 1:28:10sidelines today are just going to have their asses handed to them. And it's it's, you know, that's why I do this
444. 1:28:15channel because I want as few people as possible sitting on the [ __ ] sidelines. You don't have to like AI. You can [ __ ] hate it,
445. 1:28:25but don't ignore it. Don't not learn it. Learn it. Learn what it actually does. Learn what it actually is.
446. 1:28:41Learn how it actually [ __ ] works. It is technically definitionally not a plagiarism machine because it's not copying.
447. 1:28:56Meltdown Monday. Here we go, baby. All right. There was supposed to be a Tik Tok pin here. What's What was the pin
448. 1:29:01that was up? I'd like to show the naysayers my library of versions before things felt right. Exactly. AI producer.
449. 1:29:09The ones that are being creative with it today are the new filmmakers of tomorrow. Exactly. Look at Kelly Bosch. Look at Joy Pury. Right.
450. 1:29:20You You've got You've got people today who were not filmmakers, but who have stories to tell. Both Joy Perie and Kelly Bosch in
451. 1:29:35different interviews effectively said the same thing. Someone asked, Kelly Bosch was speaking at an AI salon,
452. 1:29:42I don't know, a year ago, year and change ago, and like the images that she makes are just so [ __ ] remarkable.
453. 1:29:52And someone said, "Kelly, how do you prompt it to get that stuff?" And Kelly's response was [ __ ] brilliant. She goes, "I've always had
454. 1:30:05these images in my head. I've never been able to get them out. That's [ __ ] genius. AI
455. 1:30:20is not making images that Kelly Bosch goes, "Oh, I'll take one of those." AI is not the genius. Kelly has had these ideas in her head since
456. 1:30:35she was a little girl and never had a way to get them out. And along comes AI and she can go she can say, "Huh,
457. 1:30:45what if I could get it to do this?" And then it generates stuff. And then she dials up the stylization number and it weirds it out.
458. 1:30:55And over a series of months and now years, she knows how to use these tools to get the [ __ ] in her head out of it.
459. 1:31:07Joy Perie said the same thing. Someone asked her like, "Why are you making all these films now? What?" She said, "I've
460. 1:31:14always had stories to tell. I've never had a way to tell them." That's the future of film making. And it used to be the prerequisite to
461. 1:31:28film making was you had to pick a single discipline and you had to go to [ __ ] USC or Colombia. If you wanted to be a
462. 1:31:37screenwriter, go to Colombia. You could learn the screenwriting thing. You want to learn cinematography, go to USC,
463. 1:31:44right? Or go intern for for Spielberg or, you know, one of the biggies. go work for Roger Corman and you would
464. 1:31:54learn your discipline. You would get in your union and you would do your thing. You would be a cog in the big wheel.
465. 1:32:01And maybe if you were a Quinton Tarantino, you'd be like, "Fuck it. I'm going to go make my own thing. I'm going to do everything." Now
466. 1:32:14that prerequisite is gone. The only prerequisite is you've got to be curious about AI enough to understand
467. 1:32:24what's [ __ ] possible and you have to have stories to tell. If you can take that curiosity, a little bit of [ __ ] mojo,
468. 1:32:38little bit of tenacity, a little bit of taste, go tell your story is going to rise above the noise. That's not real films. AI films are
469. 1:32:54theft. Okay. Do you think it's going away?
470. 1:33:08Okay. You know, your buddies are actually out making films right now. Well, I'll get I I just
471. 1:33:23I have deep deep deep deep empathy for how [ __ ] terrifying this is, for how devastating it must be if you've
472. 1:33:38got a 20 or 30 year career doing a single thing. Like, can you imagine being a special special effects guy or a
473. 1:33:46cinematographer or a lighting person right now and you look at what's coming out of V3? You're like, "Well, [ __ ] Who's going to hire me?
474. 1:33:56You should be learning VO3, right? I get that it's painful. I get that it's [ __ ] up. I get that everything's going to change. I do not get
475. 1:34:09people just being pissed off and sitting on the sidelines. It's not going to slow it down. It's not going to change a [ __ ] thing. Anyway,
476. 1:34:18all right. Meltdown Mondays. Living up to the hype. God damn it. I've written seven screenplays. Like, I I get it. I get it.
477. 1:34:31I'm in the middle of writing a musical right now using AI about AI that I want performed by human beings on a stage in New York City.
478. 1:34:45There's a mind bender. AI steals jobs. AI is going to ruin all the jobs. Well, AI created a musical that if I get it produced is going to
479. 1:34:58employ human being musicians, human being actors, stage managers, set designers, lighting designers. It's not an eitheror.
480. 1:35:13The world isn't [ __ ] black and white. It's just a new technology. [ __ ] learn it. any updates on that, by the way? Yeah,
481. 1:35:22we actually um I can't really say who it is, but a good friend of ours is um is a writer who's won an Oscar and
482. 1:35:37done some very very major TV shows and he read it, which is amazing that he even read it. um and loved it and is is uh potentially
483. 1:35:49gonna make some intros to um some folks. So, that would be cool. All right, you've just answered my questions. This is so exciting. What are
484. 1:36:00your questions, Sophia? I've been afraid of AI, but you just gave me hope. Yeah, listen. Listen, you Here's the Here's what I can promise
485. 1:36:09you. Here's what I can promise you, Sophia. I No one no one can explain to you what it is about AI that is so
486. 1:36:24profoundly powerful. The only way you can understand what AI is is to use it yourself and to and to just play with it and play
487. 1:36:37with it and play with it and take the things that you're passionate about and put them into whatever AI tool you're
488. 1:36:45using. It literally doesn't matter. Chat GPT or Midjourney or Runway ML that we just played with or V3. It doesn't matter.
489. 1:36:55take who you are and and figure out how to ask it for something that you deeply understand and at some point AI will
490. 1:37:04produce something back at you that will literally melt your face. We we call it in here your Kevin Mallister moment,
491. 1:37:12right? If you remember the movie Home Alone, this moment, that moment happens to me probably, I don't know, once a week
492. 1:37:26where I I just I'm literally like, wait, it can do that now? Holy [ __ ] There is no way to explain that feeling
493. 1:37:38and to to to to understand how profound these tools are. So just go play. Just go play. The other thing that is
494. 1:37:46impossible to articulate to the people that are sitting on the outside, the media and Hollywood for the past 50 years has
495. 1:37:56essentially said AI's evil. AI will kill you. And so that's all they talk about. And so it seems like this very binary thing.
496. 1:38:11But what we experience here with the people that show up here night after night and the people that are part of
497. 1:38:15the AI salon and the people that are part of other communities like she leads AI is that people are reinventing themselves using AI.
498. 1:38:26That what AI actually is is a humanity a a a human being amplifier. It's like a [ __ ] jetpack for your brain and your
499. 1:38:36and your capabilities. If you can't program, now you can. If you couldn't make movies before, now you can. If you can't write music, now you
500. 1:38:48can. If you can't write screenplays, now you can. Well, what do you do with that? I don't know. Have a [ __ ] idea.
501. 1:39:02Learn what's possible and go execute that idea. I've written a musical over the past year. I haven't thought about writing
502. 1:39:13something for the theater in 30 [ __ ] years. And without AI, I would never in my life have considered writing a musical. I
503. 1:39:26don't have the talent. I don't know the right people. I know it's a [ __ ] slog. The other thing you need to do is get
504. 1:39:36your ass in a community like the AI salon. Can you pop up the uh the uh the the URL there? If you're not
505. 1:39:45part of the AI salon, go to the salon.ai and click on join our community and that'll take you over to the community
506. 1:39:53site. Sign up. It's free. Introduce yourself, connect with people. There's all sorts of areas there. There's, if you want to dig deeper,
507. 1:40:05there's a there's a subscription area called the AI Salon Mastermind where people who are like really committed to
508. 1:40:11like upping their game are getting in a more focused, concentrated sub community, but the AI salon itself is free and will be
509. 1:40:23and it's full of remarkable generous people. All right. Okay, let's look at let's look at Okay, welcome to chat add. And I know we got
510. 1:40:37to get out of here. It's getting late. Um, about an hour ago, I kicked off a deep research thing. How long did it take?
511. 1:40:46Does it tell us in here how long it took? I don't know. Build on top of this. Create something new based on this, like an audio overview.
512. 1:40:56All right. Create I want an audio overview. How do I make an audio overview?
513. 1:41:10All right, let's see what it did. Navigating the post app era, a strategic framework for organizing and dis discovering.
514. 1:41:19Sorry, the voice is going. Oh, AI salon. Cool. Yeah, a AI salon's really good. Okay. Um, it's I'm I'm really glad we've
515. 1:41:32got some new folks in here tonight. That's exciting. Um, okay. The paradigm shift from tool centric directories to goal oriented
516. 1:41:41ecosystems. I like this. The challenge of organizing a the rapidly expanding universe of artificial intelligence is not merely a matter of
517. 1:41:51categorization. It's a fundamental problem of information architecture in an era of unprecedented technological
518. 1:41:58convergence. Yes, this is what I've been saying people. [Laughter] The very nature of AI platforms is rendering traditional directory models
519. 1:42:09obsolete, demanding a paradigm shift. Okay. The inevitable collapse of functionbased categorization. Fine.
520. 1:42:17Introducing a new parag a new paradigm. jobs to be done framework. Oh, I like that. The jobs to be done framework.
521. 1:42:29To build a system that reflects how users actually think and work, the organizational paradigm must shift from
522. 1:42:36what tools are used to what users do. The jobs to be done framework, a theory of innovation popularized by strategists
523. 1:42:45like Tony Ulwick and the late Harvard professor Clayton Christensen. Yes, know him provide the ideal conceptual
524. 1:42:53foundation for this shift. The core premise of jobs to be done is that customers do not buy products. They hire them to get the job done.
525. 1:43:05Huh, I like it. The jobs to be done framework compels a deconstruction of this highle job into its constit constituent steps or tasks.
526. 1:43:19The animated Bigfoot video job can be broken down into generate a visual concept, write a short engaging script,
527. 1:43:27generate a compelling voice over for the script. Right? This is what we just talked about. Animate the static
528. 1:43:33character based on the thing and then you combine them all together to make the animation. With VO3, you don't
529. 1:43:40really need all of those, but that's okay. If you're doing it with something like Hedra, you do. This deconstruction
530. 1:43:46reveals that the user does not need a single tool. They need a workflow. Very cool. From user intent. All right.
531. 1:43:54So, let's turn this. How do I turn this into a mastermind? Anybody? Into a into audio overview. Oh, Vicki, we use jobs to be done at
532. 1:44:08work. That's funny. Ask it in the chat hole. Can you turn this research into an audio overview?
533. 1:44:24Just a sec. The proposed information architecture is built upon four primary entities that
534. 1:44:39form a hierarchical relational data model. Oh, you know what? Oh, oh, oh. People. People.
535. 1:44:57Okay. It's almost it's almost like the directory would be a directory of outputs,
536. 1:45:12right? And you could like you could search for like Yeti video and what you would get is a bunch of Yeti videos
537. 1:45:21where whoever made the Yeti video gives you their chain. What tools did they use to make that video? That would be kind of cool because then
538. 1:45:32you could say for voiceovers, you know, which are which are a task in the chain, you know, 60% of people on this site use
539. 1:45:47this voice over tool, right? Use this voice generation tool. I can't help with audio overviews. You can add this report to notebook LM.
540. 1:46:00Yeah, that's a good idea. Export. Okay, fine. So, export to Docs. Yeah, this is really cool. That's loading.
541. 1:46:31Oh, look here. It's got So, wait a minute. Create a web page. Um, I want you to turn this into an app
542. 1:46:46where a user can share something they created and then share
543. 1:47:09the steps and tools they used to make it. I want you to
544. 1:47:25augment my crappy description here with um your insights
545. 1:47:40based on the research. I think what would also be really interesting is to run this over to Claude and have Claude do this and run
546. 1:47:52it over to chat GBT and have it do this as well. Um, but let's let's go ahead and boom. All right. So, here's this.
547. 1:48:10So, we're going to copy this We're going to go to Notebook LM. Um, if you don't know what Notebook LM is, notebooklm.google.com.
548. 1:48:24It is a quite remarkable tool from Google. Let's see. I'm going to go Google Docs. Uh maybe I need to save this as a
549. 1:48:40organizing multimodal AI tools. Okay. Organizing multimodal AI tools. Insert. So basically you can put up to 50
550. 1:48:52documents into a notebook LM document and then you can do all sorts of things with it. So we're going to do a deep
551. 1:48:58dive conversation. We're going to go generate and this is going to turn. So, okay. So, we just had Gemini go research, I don't
552. 1:49:11know, 400 websites about how to think about how you even organize AI tools because you can't into this new jobs to be done framework,
553. 1:49:23which is cool. And now we're having Gemini make that application. And now we're creating a podcast about this new concept.
554. 1:49:34This is why I can't go to bed either. Drop it in notebook. LM. This is a problem. This is a real problem. I go.
555. 1:49:54Isn't this crazy? The world we live in, people I mean, it it really is just we just do live in remarkable times. I I
556. 1:50:03hate to break it to you, but we do. Share your AI workflow. AI workflow showcase. Discover the how behind the wow.
557. 1:50:16You know what? Do you know what this app is? I think it was Silverf Fox said it that
558. 1:50:32that um like you should see how many of my images hit the cutting room floor right before I shared the thing. You
559. 1:50:40should see how many crappy things I created. A site like this could be a real argument against the [ __ ]
560. 1:50:48doomers that are just like AI is for lazy people. Oh yeah, go look at the 77step process that that made that 30
561. 1:50:58second video you just made fun of and you tell me if you could do that, you [ __ ] tool. [Laughter] Share your creation. Okay, so this is
562. 1:51:14starting with data. I don't want the [ __ ] data. What I want is I want to lead with the user uploading
563. 1:51:30or linking to some finished product. And then based on the products,
564. 1:51:46the directory tool interviews them to reveal the tools they used. the tools
565. 1:52:07and specific features. They used their human contribution
566. 1:52:23in let's see including curation editing refinement
567. 1:52:41tool bashing uh etc. And how they assembled the final
568. 1:52:59deliverable or project Uh uh uh dude, this is a rabbit hole. I know. You can also go back to the original canvas
569. 1:53:14and generate infographic for time. Let's see if we get uh Is this done? Yeah, this is done. 28 minutes. Okay. If you
570. 1:53:22haven't heard a notebook LM um what you call it? Interactive mode. Can we Does this Wasn't this generating videos
571. 1:53:34download? No, just a wave file. Okay. I thought they were gonna do it. Okay. Um I'm going to I'm going to share
572. 1:53:43different and then we got to go. Listen, I got to go people. I can't can't be here all [ __ ] night. Um,
573. 1:53:52but I want to play one of these things for you because if you haven't seen them before, so Gemini did our research for us and
574. 1:54:02then we uploaded the research report into Notebook LM and then we said make an audio overview and then here's what we got.
575. 1:54:09Welcome to the deep dive. We're here to plunge into complex topics and hopefully pull out some real insights. Have you
576. 1:54:17ever been really excited to try a new AI tool, maybe for writing or images, maybe even video, only to, you know, spend an
577. 1:54:26hour just bouncing between websites, totally confused about which one actually does what you need and does it well?
578. 1:54:32Oh, absolutely. It happens all the time. It Lonnie Viles, thanks for a great show and a lot of great intel. Thank you very
579. 1:54:39much. Appreciate that. I appreciate it a lot. Um, here's what I would encourage you to do. Keep coming back. Um,
580. 1:54:47we call this channel chat add for a reason. Um, it is it is guaranteed to be unpredictable, occasionally useful,
581. 1:54:57hopefully always entertaining to some degree. Um, so so keep coming back to these. Um, and also get involved with the community. So
582. 1:55:07join the AI salon. Tomorrow, the AI Salon, we've got an AI salon presents meeting where we've got the founder of
583. 1:55:14Jelly Pod, which is a tool that makes podcasts like the one we're listening to, but where you can control everything. Um, he's going to be
584. 1:55:23speaking at the AI Salon tomorrow night, Tuesday, at between 5 and 7 Mountain time. Um, you should come to that. You
585. 1:55:32should come to those. Get to know people in this community. It's incredibly powerful. So, if you go to the the
586. 1:55:37salon.ai AI that was on screen right there. Click join our community and then click the events tab. You'll you'll see
587. 1:55:43an RSVP for the for the event tomorrow. All right. And come see uh Preston from Jelly Pod. All right. Let's uh
588. 1:55:53Yeah. And we will likely be late tomorrow. Normally I go on at 8:00 p.m. Mountain time here. It'll be 8:30 or 9
589. 1:55:59depending on when I get out of the uh the salon. But anyway. All right. Oh, Pearson. Sorry, not Preston. Pearson from Jelly Pod. Um, okay. I
590. 1:56:11tried to use chat GBT for professional registration. It failed at accuracy. Yeah. Uh, if if you're if you're trying
591. 1:56:18to do things like uh things that require accuracy, you're going to have to be uh you're going to have to do a lot of
592. 1:56:27proof reading when you're using generative AI. That will get better. It's it's gotten better over the past
593. 1:56:32two years. That's still a major [ __ ] headache with it. So yes, if you're doing creative writing, let it go hallucinate, right? That's how
594. 1:56:41creativity happens. Uh if you're trying to get something that is predictable and repeatable. Yeah, that's it. All right, that's it.
595. 1:56:50Can someone put the link in the chat for me, please? Link to what? Often entertain. Oh, to the uh to the
596. 1:57:00Yes, hang on. To the AI salon. So, go to the salon.ai. Just go there and then um scroll down. There's a button that says join our community.
597. 1:57:13That'll take you over to the community site and that's where all the information about the uh the event for tomorrow night is. Okay.
598. 1:57:22Beautiful. All right, everybody. Thank you for uh a fun night. We got some stuff done. Ann Murphy jogging out of
599. 1:57:31the hotel, the motel, sorry, hotel. A little highfulutin in her 70s jogging suit. That happened. Tik Tok pin. Often entertaining, often
600. 1:57:45informative. I have information overload. Can't process. I know, Winston. It's a lot. It's a lot. I'm really excited about this new uh this
601. 1:57:54new concept for uh what is it? Jobs tobun approach to tool discovery. I like this a lot. All right, we're gonna I
602. 1:58:05think over this week, so producer Brandon, try to keep me on task, which I know is [ __ ] impossible, but
603. 1:58:13I think what we might do is actually take this concept and vibe code it into a a platform that we could all start
604. 1:58:22using to share our projects. This is a very AI salon. Um, it's very onbrand for the AI salon. this whole approach where you say here's a
605. 1:58:33cool thing I made and then it helps you unpack how to talk about it and then what that becomes is a is a tool
606. 1:58:41recommendation directory where you can either look for tools or features or outputs. Oh, I want to do an output like
607. 1:58:49that. How do I do that? Here's 16 different ways you can do that. That's really cool. We have that. It's called
608. 1:58:56the AI salon. Yeah, there's there's literally a section called look what I made and people share what they do. This
609. 1:59:01but this is just a a a searchable version of that. That's very cool. All right, everybody. Uh I'm going to get
610. 1:59:08out of here. It is uh yes, 10 o'clock. Let's go. All right, peace out and see you tomorrow at AI Salon Presents and
611. 1:59:14then I'll see you here um sometime after that. Probably 8:30 or 9 Mountain time. All right, peace out everyone. Bye.