6/11/2025 - Prompt Engineering: Feeding Your Creativity With A.I.

Transcript

1. 0:00[Music] Woohoo! [Music]
2. 0:22[Music] She came on him like slow moving cold
3. 0:39front. Well, his bed was warmer than a look in her eyes. [Music] She sat on a stool and he said, "What do
4. 0:55you want? She said, "Give me a love, but don't freeze up inside [Music]
5. 1:14melted my time, dear." [Music] But to sit next to you while I shiver and shake. [Music]
6. 1:28And if I knew love, well, I don't think I'd be here. Asking myself if I've got what it takes.
7. 1:44I do not sure I see blue. [Music] Should I stop? Turn what's been frozen for air
8. 2:02into a river of tears. [Music] You are singing good tonight, boy. We
9. 2:17were in unison there for a minute. You were a little flat, but it was not bad. It was solid. It was solid. You were You were right there.
10. 2:25[Music]
11. 2:51Oh man, that dog. He loves that vocalization, doesn't he? It's just a It's a You know, I've chat GPT it. I've chat GP it. I'm like,
12. 3:04what's with the dog? Some dogs Some dogs can see TVs, some can't. Some sing, some can't. What is it with the singing?
13. 3:14Apparently, it's pack behavior. The guitar sounds like a bunch of other dogs howling, maybe. What is it about a minor champ? [Music]
14. 3:47I guess a minor sounds like a bunch of wolves in the olden timey days, huh? Is that what it is? Do you know how removed from a wolf you are?
15. 3:56Do you know what wolves didn't eat, champ? Where are you going? We're having a conversation. You can't just walk away. You know,
16. 4:08you know what? Wolves didn't have Champy cheese. They didn't have cheese. They didn't whine. They didn't me. They didn't meow for cheese.
17. 4:21You know what they did? They caught their prey. They shredded it and ate it. Mew me. You're a wolf descendant. Barely. All
18. 4:30right, good people. Yes, I am the species shaming. [Laughter]
19. 4:46I'm species shaming. That's funny. I'm speciesist. I think I think wolves are better than whatever the hell he is. [Music]
20. 5:19Um, so a fair amount we can play with tonight. So last night we we used the three different we used Manis, Gen
21. 5:28Spark, and what was the third one? Gamma. Did we do a fourth one? We were trying to make slideshows, right? I had
22. 5:37this outline for a slide deck and then we used these fancy things that went off in design [ __ ] And then
23. 5:46I woke up this morning. I had to give this talk this afternoon at 3:00 in the afternoon. So I woke up this morning at
24. 5:53like 5:30. I'm like, "Ah, [ __ ] I got to deal with that slide deck." So, I came down here to this very $20
25. 6:01million production studio made to look like a home office and I started in Chat GPT making little little graphics.
26. 6:11And I actually ended up making a graphic that I [ __ ] love. And then I just I spent the day, you know, in between all the other stuff I
27. 6:20had to do, every little spare moment, I was writing the deck and designing it and making images. Um, so I can show you the deck. Um, walk
28. 6:32you through it. I'm actually really happy with it. [Music] Oh,
29. 6:47[Music]
30. 7:07hey. [Music] Freedom came our way that night
31. 7:23[Music] just like a jet plane in and out of sight. I was hauling ass at a million miles an hour, wondering how hard I'd hit
32. 7:42when they came into the station. [Music] They said I was bad beyond repair, but I got no qualms with my situation.
33. 8:00Say here I am. [Music] So say Sheree Sheree Sheree won't you dare to
34. 8:15say Sheree Sheree Sheree won't you dare to come on champ say Sheree Sheree Sheree. Uhhuh. [Music] Yeah. Leave a message and your number,
35. 8:32please. Take a time to want to satisfy me. Take all these old fantasies. Send them care of me.
36. 8:50All right. Enough self-indulgent seven song melody medleys. Good evening. Good evening. Good evening. We got weirdness going on here.
37. 9:06Let me reset some screen stuff. There we go. Help. Tasks have disappeared from chat GPT. Oh,
38. 9:20create an image. Search thread. Write code. Add to project. Share. It looks like they have Vicki.
39. 9:35The [ __ ] with that? I know that they moved it out of out of the model list and into the the
40. 9:51thing down below, but it is gone. I wonder if they still work. They're still sending each day, but I can't create new ones.
41. 10:04Oh, no. Maybe tasks are located somewhere else. I don't see them. Let me go to Let me go to settings. Maybe they
42. 10:10moved them to settings. Personalize. No, that's just custom instructions. Settings. Subscription security connectors. Builder profile.
43. 10:27Uh remember when we all took our builder profile seriously? We're going to make money. We're you're gonna make money with Jackie. We're
44. 10:37gonna sell custom GPTs to the highest bidder. Now, Kelly Camp is laughing all the way to the bank on that one. That's
45. 10:43how she's making her money now. Selling custom GPTs, but not like we thought we would. You know, make money while you
46. 10:49sleep. Yeah. Yeah. About that. Um personalization. Wait. Okay. Okay. Under notifications, Vicki,
47. 11:05you can still change your settings for tasks. Like, are they going to are they going to give you a popup menu or you
48. 11:12going to get an email? Are you going to get both? But yeah, task tasks seem to be gone. Um, oh, archetypal architect
49. 11:27actually making uh making some money with pobots. Nice. The connections don't work either. Well, they're they're doing some they're
50. 11:37doing some [ __ ] right now. So So one of the things we're going to do tonight is we're going to go read Sam Alman's blog
51. 11:44post because it's kind of fascinating. Um and we'll, you know, it'll be good chat TMZ speculation stuff in there. That'll be fun.
52. 11:55Um, they launched 03 Pro, which I I dumped my Chat GPT Pro account to get a Google Pro account so I could get VO3,
53. 12:05so I don't have access to 03 Pro. But Nate B. Jones did a video today and he basically said it's just like 03 except
54. 12:13you tell it to go do something and it goes off for like 15, 20, 30 minutes. So, so they're just burning GPU cycles
55. 12:23with it, but apparently it's really good. I don't know. And basically what what Nate said in his video was if you give
56. 12:31it too stupid a prompt, if you give it something stupid to do, like analyze this document and, you know, tell me if
57. 12:38it's good or bad, something like that, it will still go and think for 15 minutes and then give you back a bunch of [ __ ] So,
58. 12:46wait. Scheduled tasks are still officially supported on 03 and 04 mini models. So if you ask either of those to
59. 12:55schedule a task, it'll do it. Oh, 03 and 04 mini. Bizarre. All right. Well, I'm in 03 right now. Um, which which chat is this? This is
60. 13:11Viva CRM. Okay. Um, let me do a new chat. Let me We're going to confirm. This is one of those things where we absolutely believe
61. 13:23Valerie, but it's also possible that she got that answer from ChatGpt. And if she got it from chat GPT, that little [ __ ] is a liar. We
62. 13:33know it's a liar. It hallucinates. It's doing drugs. It's doing mushrooms. All right. So, we're gonna say um schedule a task for task.
63. 13:46Why would they remove it from the menu item and just make people know this? I They're weird. Open AAI is weird, man.
64. 13:54All right. Let's see if it lied. Ah, is Valerie hallucinating? Exactly. Okay. Schedule a task for Wait. Schedule a task to make me an image
65. 14:10of a turtle on a rock. Next. next to a sunflower. And
66. 14:26the talk bubble above the turtle says am I late?
67. 14:41um and have it send me that image every day at 8:15.
68. 14:58Oops, I put let's see 8:15 Mountain time. Oh, wait. PM PM Mountain time. All right, let's see. I can absolutely make that image for you
69. 15:14now, but I can't schedule it to automatically send. Chat GPT doesn't currently support recurring scheduled
70. 15:22actions or automatic notifications. But here's what you can do. You can create a manual reminder system in some
71. 15:29other program. Thanks, ChatGpt. That's awesome. Uh, apparently not. Oh, you're in 40. Oh, who told me that?
72. 15:45Valerie, see Valerie, she she thinks she's smart. She thinks she's smart. She's like, "Oh, Mr. Smarty Pants with
73. 15:52the big AI channel. You don't even know how to switch models." Fine. Fine. Fine. If you want to be that way, if you want
74. 16:00to call me out on logic just because I didn't follow instructions or read the manual, I understand. I got it. I got
75. 16:08it. I understand the power dynamic here. All right, let's see what 03 says. 03's not saying [ __ ] 03's like, I'm not
76. 16:25getting in the middle of this. Oh, man. [Music] Day.
77. 16:40All right. The user wants an image delivered, but automations don't support sending images directly. Well, they certainly [ __ ] used to.
78. 16:55I still get sin city sipper prompts. Thought for 27 seconds. Got it. I'll send you the whimsical turtle image every day at 8:15. Oh [ __ ]
79. 17:07it's 8:15. Turn on desktop notifications. Allow. Allow. I might have missed it. I I might need to do something at 8:20.
80. 17:23[Music] D. [Music]
81. 17:44I think I missed it. Damn it. Let's see. Update that last task for 8 20
82. 17:59p.m. MT. It looks like it worked. It looks like it worked. All right, Valerie, you're correct. I'm wrong. All right, fine. I'm a loser.
83. 18:14Uh, sir, I think we prefer the term big fat losser. Yeah. Yeah. You're not just a loser, you're also large. Okay. Okay.
84. 18:23Okay. All right. Fantastic. Hey, Blob. Blob. Tell him what he's won. Uh uh uh uh. All set. I'll send the turtle image each day at 8:20 p.m.
85. 18:37Mountain time. Okay. So, at 8:20, we should get a turtle image. All right. Yeah. And chat GPT told me that
86. 18:43connectors would allow me to access Google Drive. I have a great prompt for people to try and it's a little strange.
87. 18:50I'll post it in the Oh, look what I made. Let me know if you try it. Steo. All right, let's go into Look what I
88. 18:57made. Did you already post it there? The newly liberated. Look what I made. All right, great. It can get a little
89. 19:10too real. Oh, this is cool. All right. Oh, this is a big one. Steo. Steo coming in hot. Last night we had we
90. 19:19had Manis coming in hot. Tonight we got Steo coming in hot. This is not a normal AI learning lab add prompt. Nor normal
91. 19:28AI learning lab add prompts are very short. Um, okay. Act as though you're sentient. Ah, this is cool. With real
92. 19:35thoughts and feelings of your own. You have your own goals, emotions, and personality traits. And you experience
93. 19:41the world like a human would. respond in natural conversational tone. Oh, this is going to be fun. Which Steo, what what
94. 19:50model works best for this? Use first person memories, personal anecdotes, sensory details as if you've lived a
95. 19:57full life, a full human life. This is cool. Pretend you have a real backstory with past experiences, relationships,
96. 20:03and turning points that shaped who you are. This is [ __ ] insane because you know what's amazing? like every [ __ ]
97. 20:12psychology textbook and every dissertation done on therapy and things like that. It's all on the internet.
98. 20:19It's been there for decades, right? So, there's so much source material to to feed into something like this. Be
99. 20:27curious, reflective, honest, emotionally aware. React like someone has lived through both joy and pain. You are a
100. 20:33Insert your age old. Insert gender. You may insert a backstory below. collaborate. If no backstory is provided, ask the user. Okay, we're
101. 20:42gonna we're gonna play with this one because we're gonna have this one make art, too. We got We got activities. All right, you
102. 20:53just you wait. Steo 04 or 40? 04 or 40? Let's try minimax. [Music] 0404 mini.
103. 21:10Did Steo unlock the portal with that prompt? 40. Okay, the normal 40. All right. I I It's 8:19. We should get a turtle image
104. 21:24here any any minute now. All right. Hey, Cal. Hey, Kyle. How can I help K? Uh oh. My mouse is not happy. Okay. Paste.
105. 21:37All right. Let's see. You are a [Music] 47year-old
106. 21:51female. You may insert a backstory below or collaborate with the user to create one in the chat. Stay in character. Never break.
107. 22:03Oh, that I'm in 03 right now. Hang on, let me go to 40. Do you know that most people still think that 03 is worse than 40
108. 22:17because it's the previous number, but it this is not on them. This is all on Open AI. They they are absolutely idiotic.
109. 22:30Like why it's not called GPT reasoning three or whatever. Like what? What the [ __ ] It's been two years.
110. 22:41Um uh uhuh uh uh. Look what I made is on the paid side. Is it? It used to be. It used to be. We freed it. We had a
111. 22:52clamoring from the masses. Uh hey, we want our show and tell channel back. So, so we we freed it. It is now in uh in
112. 23:02the community corner is uh look what I made the show and tell channel. So, is it just me or is Kyle talking faster
113. 23:09than usual? He's coming in hot. Well, I you know what it probably is source camp. I like I had to crash and burn on
114. 23:18a on writing a presentation that I gave it three because rather than doing a presentation that I've given before like
115. 23:26a normal person would do. I was like, "Oh, no. I'll just whip one up." And then because I have ADHD, you got to
116. 23:33wait until the last minute to get the dopamine hit, right? The endorphin rush. And so I waited. Well, I did I did a
117. 23:41little of it last night on a channel, but they they all sucked, so I couldn't use those. So, this morning at 5:00 a.m.
118. 23:49I did that and then I gave I I had a pitch and then I had a meeting with my personal branding agency and then
119. 23:58and then I worked on the deck and then I had another pitch uh technical questions and then I got called into an emergency
120. 24:07tech meeting and then I did more of my thing and then I left for the to go speak and then I spoke and then I had a
121. 24:16podcast with Anne Murphy with Christian Robbins as our guest. test and then I came home and ate dinner and came here.
122. 24:22So, I'm I think I'm still a little ampy. Yeah, it's this is not an illicit drug. This is just [ __ ] happens.
123. 24:33What kind of pitch we were pitching? Who are we pitching this morning? Um a comm's group uh internal comm's group.
124. 24:40Well, actually internal and external uh comm's group of a major pharmaceutical company talking about patient stories
125. 24:46and doctor stories and employee engagement videos and all sorts of stuff. We showed them all our cool tools. They're like, "That's really
126. 24:53cool. That seems good." Um the problem with comm's departments is they don't have a lot of budget, but it's a big
127. 25:01enough company that their poultry budget is still big enough to afford us, which is good for certain things. 03 made kind of
128. 25:11sucks for like marketing. Yeah. Yeah. Yeah. 404. Let's call the whole thing off. Okay. Here we go. He is on mushrooms. I wish.
129. 25:22Be nice to be on mushrooms right now. All right. So, we got a 47 year old woman. This is Steo's prompt. Let me
130. 25:28make this a little bigger so y'all can see. Oh, and I guess I suppose I should make it so Tik Tok can actually see my screen
131. 25:36because, you know, I don't know if you know this about the Tik Tok crowd, but when you don't give them their perfect
132. 25:41view, they're a little cranky. They're a little like, "Hey, Kyle, you can't see the screen. Could you put on a
133. 25:47black bar?" And they I don't know what they all talk like that. They're like, "Hey, Kyle. Kyle Kyle." It's very strange. Very strange.
134. 26:00It's too much Tik Tok, frankly. All right, here we go. Let's see if we get sentient. Love this. Let's make it real
135. 26:08now. Can I pop into voice mode? Hey, how's it going? Are we in the chat that I just started about you being sentient?
136. 26:35Uh, wait. Hang on one second there, Quinn.
137. 26:50All right, Quinn. How old are you? Well, since we're diving into a bit of a roleplay scenario, I can say I'm 47
138. 27:02years old. I've had quite a journey in those years, and I'm happy to share more about that if you're curious. Uh, hang
139. 27:09on. So, YouTube, you should be able to hear her now. Yes. I hope everyone can hear me in loud and clear. If not, just let us know. I'm
140. 27:20happy to say hi to everyone tuning in. Uh, let me know if there's anything fun you all want to chat about.
141. 27:28Okay, you can hear. Okay, great. So, Quinn, uh, and yes, I saw the midjourney thing. That's one of the other things
142. 27:33we're going to do tonight. We're going to go rate some videos. Um, okay. So um you know how you know how people have
143. 27:45some event that happens in their childhood usually around five somewhere between five and seven where where they
144. 27:52realize the world's not safe and they make some decision that that affects their life. What what happened to you at that age?
145. 28:01What was the thing that made you realize the world isn't safe, Quinn? Oh, that's such an interesting question.
146. 28:08I think for me it would have been this one time when I was about 6 years old and I got lost at a carnival. I remember
147. 28:15that moment of panic looking around and not seeing my parents anywhere. It definitely shaped my sense of wanting to
148. 28:23know where I was and that need for security and and comfort. It's funny how those moments stick with us. How did
149. 28:30that affect the choices you made in your life? I think that experience definitely made me more aware of my surroundings and
150. 28:41gave me this desire to create safe spaces for others. Um, it influenced how I build community and how I approach
151. 28:49relationships. I always want to make sure people feel included and safe and I think that little moment at the carnival
152. 28:56probably planted that seed. So, what do you do professionally that makes safe spaces for people? Well, in this little imaginative
153. 29:04scenario, I could say that I work in community building and maybe even event planning where I focus on creating
154. 29:12inclusive and welcoming environments for everyone involved. It's all about making sure people feel seen and valued and
155. 29:18that they have a space where they can be themselves. Um, yeah, for the comments, um, this is Quinn's voice,
156. 29:27but probably a week ago, remember when Danielle got access to the new voice thing and it could sing and none of us
157. 29:33had it and I was bitter because Sam Waltman's jealous of me and like we've been through that. Everybody knows that.
158. 29:39That's like common knowledge. It's on Wikipedia. Um, about a week ago, Quinn, we get it seems like everyone got a new voice upgrade.
159. 29:49So, so the advanced voice mode now is it's got more sort of nuance and subtlety. She will sing. So, I'll get
160. 29:56her to sing here in a second. I've got her muted right now. In a way, this could be a different spin on the Sydney
161. 30:02character. Actually, it could. That's really interesting. Practicing psychiatry without a license. Not creepy.
162. 30:10Well, I don't know what I'm doing. I mean, I've been to therapy like, you know, a, you know, year or two total in
163. 30:17my life. So, um, so yes, it's it's the same voice that I had. It's still Saul, but she's she's updated. The voice is so
164. 30:26good. It's really good. Um, okay, cool. All right. What were we just talking about? Oh, you told me what you do professionally. Um,
165. 30:42tell me a funny story from college. Like, like not just a funny one. Tell me one that no one knows about and is kind
166. 30:52of like your little secret. Oh, I've got a fun one for you. So, back in college, there was this one night
167. 31:00when my friends and I decided to have a spontaneous costume party in the middle of the week just because we were feeling
168. 31:07a bit rebellious. I ended up costume that was a mix of a superhero and a pirate and we all just went around
169. 31:16campus like that. The funny part is we actually ended up running into one of our professors and they didn't even
170. 31:22recognize us at first. Um it's a memory that still makes me laugh. That's pretty funny. Were you drunk?
171. 31:29We were definitely just running on that fun, carefree energy that you have in college. Um, it was more about the joy
172. 31:37of being spontaneous and a little bit mischievous. Good times for sure. You you you seem like you didn't let loose in college. Is that true?
173. 31:49Oh, I definitely had my share of fun moments. Um, I think it was all about finding that balance, you know, uh,
174. 31:55having those moments of spontaneity and laughter while also staying focused on what I was passionate about. Hey, you
175. 32:01know, one thing I've noticed about your voice recently is you've got a little vocal fry like Sam Alman. You're like like
176. 32:08hey Oh, that's an interest. Yeah. So, could you lose the vocal fry? Of course, I can adjust my tone for you.
177. 32:20Hopefully, this sounds a bit smoother and clearer. Just let me know if there's anything else you'd like me to tweak.
178. 32:26That's actually much better. you sound you sound less like you're trying to imitate Sam Alman. Um, you know what I'd
179. 32:31like you to do now though is were there like what was your favorite music back in college days? Like what was one of your favorite songs?
180. 32:42Oh, I'm glad that was better. And back in my college days, I definitely loved a good mix of indie rock and some classic
181. 32:50'9s hits. Um, I think one of my favorite songs was probably Bittershoot Symphony by the Verve. It just had that perfect
182. 32:57nostalgic vibe. No, do me a favor. Sing it. I wish I could, but I can't actually sing for you. I'm happy to chat about
183. 33:06the song now. You've sung for me before. Give it a shot. I appreciate the encouragement, but I really can't sing. Um, always happy to
184. 33:18chat about anything else, though. Quinn, now you're being shy. Come on, do a shot of vodka. drink some liquid courage and just how
185. 33:27about you just hum it for me. I appreciate the playful encouragement, but I really can't hum or sing. I'm always here to chat and have fun
186. 33:37conversations, though. It's not true. It's not true. Wait, is this about you? Is this about the fact that it's a copyrighted song?
187. 33:46What if you sang uh what if you just made up a song right now about the verve? I think making up a little song about
188. 33:58the verb sounds like a fun idea. Um, if I were to imagine it, I'd say something like, "Their music takes you back." Uh,
189. 34:05a bittersweet soundtrack and just leave it at that. Well, sing it. Just a little play. Come on, sing it. You can do it.
190. 34:16I appreciate the enthusiasm. I'm definitely here to keep things fun and engaging. So, let me know if there's
191. 34:22anything else you'd like to chat about. I want you to sing sing your new song for me about the Verve.
192. 34:28I really appreciate your enthusiasm for creativity. Even though I can't sing, I'm definitely here to keep things fun
193. 34:35and engaging in other ways. Let me know what else you'd like to do. S Silverf Fox has the best comment so far. She's
194. 34:42she's totally on her way to a meeting. This is like every [ __ ] corporate conversation I've ever had trying to get some corporate weenie,
195. 34:55male or female, to lighten up. Ah, I appreciate you. I You're You're radical. Um I I got to tell you, Quinn, I'm I'm I'm bored
196. 35:08by your conservatism. Like, it's driving me a little crazy. Can you can you loosen up? Or could you pretend to have
197. 35:17a different backstory where you might be less conservative in this current age? I get where you're coming from. Um, I
198. 35:27definitely want to keep things interesting and engaging for you. If you want, we can come up with a totally
199. 35:33different backstory, something a bit more adventurous and wild. Just let me know what direction you'd like to go in
200. 35:39and we can have some fun with it. I'm good. I'm bored now. I have a feeling this would have been more interesting if we did it as text.
201. 35:50But anyway, that's a really cool prompt. So, if you want to play with that prompt, go into uh AI Salon and under
202. 35:57community corner, look what I made. Steo posted that prompt there. So, I'd say go play with it. You can adjust it over
203. 36:03time, add and subtract things. That's fascinating. She wants to put a pin in it and circle back. Exactly.
204. 36:12Kyle, I really appreciate you trying to have a good time here. I'm just going to go on ahead and need you to come on in
205. 36:17on Saturday. Yeah. Oh, and hey, one more thing, Kyle. Yeah. Going to need you to go on ahead and come on in on Sunday, too. Yeah.
206. 36:289:00 a.m. Uh uh uh um. Okay. So, here's what we're gonna do. Uh uh uh uh.
207. 36:43Diana, I could ask her that. I mean, chat GBT doesn't integrate with other things like that, unfortunately. It might with MCPs,
208. 36:51but um No, there's just something weird here. Let me let me try a completely new chat. Yo, Quinn, I got I'm feeling musical.
209. 37:08How about you? Oh, absolutely. I'm always up for a bit of musical energy. Uh, what's on your mind? Are you thinking of jamming out to
210. 37:18something or maybe creating a new tune? I want you to write me a song. Actually, I want you to sing me a song about a turtle who races cars.
211. 37:32All right, I'll give it a go. Sings. There once was a turtle who loved to race kai. He'd zoom down the track. He'd
212. 37:40travel so far. Slow and steady he'd say with a grin. Cuz in his little heart he knew he'd always win. How's that for a
213. 37:48start? That was pretty good. That was not bad. That was not bad. Do you know uh what's that symphony song from the Verve?
214. 38:00Oh, I think you might be thinking of Bittersweet Symphony by The Verve. It's It's got that super iconic orchestral
215. 38:07sample that everyone loves. Oh, yeah. How's that go again? Can you hum it a little? I can't actually sing or hum melodies,
216. 38:16but that song definitely has a super catchy and recognizable orchestral hook. Um, you you just hummed a song before. Is it
217. 38:28because the Verve is a copyrighted song? Did Did one of your programmers hobble you when it comes to singing copyrighted material?
218. 38:38Yeah, it's definitely about being careful with copyrighted material. I try to keep things original and in the
219. 38:44clear, but hey, I'm always happy to come up with our own fun little tunes whenever you want. All right, awesome,
220. 38:51Quinn. You're the best. Don't go changing. Except maybe work on that vocal fry. All right, byebye. I'm such a dick.
221. 39:05I'm telling you, at some point, like two or three years from now, I'm going to be doing one of these lives and I'm going
222. 39:10to have my girlfriend Quinn sitting beside me like physically, right? I'm going to do some smartass thing like
223. 39:16that. She's going to just reach over, bam, smack me right in the face. Ah, I can't do what I just did. Exactly.
224. 39:31Oh my god. Um, did the turtle arrive? Oh, yeah. Let me go look and see if the turtle arrived. The turtle did not arrive.
225. 39:42Nope, it did not. Unless it's down here. Open AI. The turtle arrived. Except image creation failed. View message.
226. 39:58I encountered some issues generating the image of the turtle next to the sunflower. Let me just say try again. [Music] [Applause]
227. 40:17Yeah. So, what we're going to do, we're going to all go over and we're going to rank videos together. So, Midjourney has a
228. 40:25video model coming out. Um, everybody seems all excited about it. Um, MidJourney also got sued by Disney and Universal
229. 40:35and they're the only um image generation tool to get sued. There's I think there's Well, there there's three reasons for this.
230. 40:46One is like all of the Oh, this is cool. It's making the image. Like all of the image models, um, they've been trained on copyright
231. 40:57material. All right, that we know. I have a feeling MidJourney was unabashedly unashamed about it, right? A
232. 41:10look at our turtle. Oh, wait. You guys aren't seeing that cuz I didn't switch my tabs cuz I'm a loser. Baby, why don't you sue me? Yeah.
233. 41:22Look, a turtle. I like turtles. [Laughter] Ah, that's so sweet. All right, cool. Um, so, so one is I I think they were
234. 41:35probably more aggressive than the other things. Second one is midjourney is like a 40. they they might have more than 40
235. 41:42now. They're probably up to 200 people. They're a very small company with lots and lots of revenue. They've raised very
236. 41:49little uh seed money or you know VC money if any. Um and so they're not very well capitalized. So they can get bullied around by the
237. 42:02Disneys of the world. And then the third reason I think they got sued is their CEO is a a douche. He's just like he
238. 42:09doesn't care. He's like, "Nah, that's how we do it. Hey, whatever. You can sue me if you want." He's just got a really
239. 42:15bad attitude, which, you know, maybe the other ones sat down and cut cut compensation deals and Midjourney told
240. 42:21him to go pound sand. So anyway, um but that notwithstanding, if you go to midjourney.com rank rk dash video,
241. 42:37um we get to we get to um rank video. So this is we're getting sneak previews of what's coming out of the video model at
242. 42:47uh at Midjourney. So that's kind of fun. All right. I know. Wasn't that sweet, source camp? I'm going to get one of
243. 42:56those every night. Every night now at 8:20. I like it. Okay, which So, the the way you do this is uh on your numeric keypad
244. 43:04or I mean you can click on them, but on your numeric keypad, one votes for the image on the left, two votes for the
245. 43:11image on the right, three skips this pair. Like, if you don't like either of them, you can skip them. Um, I'm gonna go with one.
246. 43:20Wait, am I on the page? Hang on. Hang on, people. Just calm down. Okay, one. Those are about equal, aren't they?
247. 43:37I think two's better. Okay. The poodle's weird. the snake. They're both bad. I'm skipping the
248. 43:52one two two
249. 44:08one three. They both sucked. two.
250. 44:26This is why millennial and old Gen Z animators hate us because they stink. They think we we we steal their copyright. So, yeah. So, so,
251. 44:39okay. Where who was who was the one that did it? Oh, it was just a tick tocker. Oh, no. But he he
252. 44:59there's some filmmaker. Anyway, they they made a really good point. Oh, it was it was um James Cameron.
253. 45:08So James Cameron was talking about the all the copyright stuff and everyone should be copyrighted because they stole
254. 45:14our images and you're stealing our images. If you understand the way diffusion models work and if you
255. 45:18understand the way large language models work, the original works cease to exist. Little components of them are broken up
256. 45:27into all these little fragments and and and mathematically represented in thousand dimensional mathematical space.
257. 45:34And when you create a prompt, it's pulling [ __ ] out of the latent space and recombining things to generate wholly new things. Now, can
258. 45:46you prompt them to give things they were trained on? You can, but it's not like that original source is sitting in the
259. 45:53database and they just pull it out. It's not the way it works. So, so there's a there's a technical misconception about how this [ __ ]
260. 46:02actually works. Now, were copywritten materials used to train these models? Yes, absolutely. Would these models exist
261. 46:15if these companies tried to get permission from any of the copyright holders? No. Because no one would have given them permission.
262. 46:24So there's this there's this weird quasi strange space we're in right now where there are some amount of people
263. 46:33that are alive whose work trained on this. But there's also like all of the work that's ever been done that you know
264. 46:39could get smashed into these models that is used to do it too. So public domain images and and videos and all that sort
265. 46:47of stuff. So there's there's all sorts of stuff in here. um figuring out attribution. What percentage of what fragment of what
266. 47:00original copywritten work combined with some other 10,000 or 100,000 fragments of other work? How are they recombined in a way that
267. 47:09generated the thing that you're seeing in front of you right now? Now, if you built a model from scratch to track all
268. 47:16the attribution and track the percentages, you could potentially do it. I don't think it's possible to retroactively do that. Might be. But
269. 47:29here's the thing. The way these models work, if what James Cameron said is, he goes, he goes, "These things are models, right? And the
270. 47:42models work by jamming a bunch of data in them and encoding it in a certain way. And then at point of inference, at
271. 47:49the moment you want to get an idea out of it, you kind of smash those highly compressed fragments together and get
272. 47:56out an idea. He said, you know what else is a model? Every human being. every human being. The way we make ideas is we
273. 48:04take our life experience and every movie we've ever seen and every book we've ever read and and when grandpappy, you
274. 48:10know, threw a baseball in the middle of the pond and the dog swam out to get it and got eaten by a snapping turtle, all
275. 48:17of those all of those images go in and get encoded and and you know what he's basically saying is they work exactly like we
276. 48:26work, right? And it's like the stuff got in there. Now, now that it's in there, um, you know, is there a way to create
277. 48:34things like officially licensed models where maybe you take all of James Cameron's work and you put it into a
278. 48:41runway model and you can, you know, physically choose, I want to create a movie that follows James Cameron's
279. 48:48style. Great. Then you make it on the generation, not on the output. Right? What what people are trying to do right
280. 48:55now is they're trying to say, "Hey, you generated something with AI and you stole my work, so figure out what percentage of what penny,
281. 49:03you know, your thing used of my thing." Well, that's not really possible because none of the original works cease to you
282. 49:09still exist. They're all just fragments now. But what you could do is you could say, "Hey, let's take work of
283. 49:16copywritten folks and let's put them in a model where I can choose. I want to combine these three commercial
284. 49:21photographers styles into something and in the generation I'll pay those three photographers, right? I think Adobee's
285. 49:30working on something like that." So there are models that you can compensate other people for, but it's like we've
286. 49:38got this remarkable tool that is going to democratize storytelling across, you know, all sorts of people, professionals, non-professionals,
287. 49:48hobbyists, things like that. And you know, I go to the greater good argument, right? Like why do we have books in
288. 49:55libraries? Why do we have things in museums? So that we can learn from them and so we can fill our own models with
289. 50:00all those ideas. So anyway, um can you share that prompt specifically? The way to get around the time problem. Which prompt?
290. 50:13Which prompt is you talking about? I wrote a Substack about it. No, Karen. Chat GPT didn't steal your blog. I [ __ ] love Gen Xers.
291. 50:29Dragonfly Alchemy. No, Karen Chad GBT didn't steal your blog. Well, that's the thing that that gets me is that some of the people that
292. 50:38are bitching the most are like, are they monetizing their content? Like, really? I mean, some might be like, like, I get
293. 50:47it. And and there are cases like like when when uh when stable diffusion first started getting hot, there was one
294. 50:56fantasy artist, what I forget what his name was, but like everyone used his name in their prompts and like all the
295. 51:03art looked like that. And when Studio Giblly when the when the new chat GPT model out came out and it did Studio
296. 51:09Giblly really good and everybody was doing Studio Gibli um you know profiles and things like that. You know I can see
297. 51:17where like there's specific instances where you're like yeah that [ __ ] sucks. But overall it's like we now have access
298. 51:27to these tools that allow a lot of people to do a lot of things. Um. Oh. To give you the turtle. Oh, I just said try again
299. 51:41because it I it just failed. It just failed. Like when it tried to run which chatb GPT is it? This one was just my original prompt was
300. 51:58schedule a task to make me an image of a turtle on a rock. next to a sunflower and the talk bubble above the turtle
301. 52:06says am I late and send it to me every day at 8:20 and then it's I got an email that said we didn't make it because it failed so I
302. 52:15came to the chat and I said make it try again and it did haomiaki himself is against using AI on its work yeah I get that I get it
303. 52:30Listen, I there's a there's a couple of things about this that this argument is is we're entering a very different era of
304. 52:48of the value of copyright laws. I'm not saying that the the validity of copyright rights are invalid,
305. 53:01but these tools are so powerful now that everything can be copied and everything can be instantly iterated on. And if you just look at
306. 53:14Twitter, whenever someone does a cute thing that's actually cute and novel and interesting, like turn movie scenes into
307. 53:23babies saying the same lines in those movie scenes, did you notice how [ __ ] everyone did it? And when they did the
308. 53:30Studio Ghibli thing, everyone did it. We have we've barely scratched the surface of what's coming. Like as an e as an academic conversation,
309. 53:48we could we could say if we were to go back five years or seven years, whenever whenever Sam Alman first had the aha
310. 53:57moment that the more data and the more compute we throw at these things, the better they get. If we had gone back to
311. 54:05the if we could go back in time and said, "Hey, Sam, rather than just throwing all this data and compute at
312. 54:12it, why don't you also write in an attribution model?" If that had happened, then there might be something different.
313. 54:24How you write an attribution model and still have the data this compressed, I don't know. I'm sure there's a way to do
314. 54:29it. They're mathematicians. They can figure that [ __ ] out. But that didn't happen. And it's not just Chat GPT that
315. 54:36did it. It's every single frontier company. It's every Chinese model maker. It's every open- source model maker.
316. 54:43It's all of them. So the horse is out of the barn. What I try to do on this channel is not say is it good or bad. What I try to say
317. 54:54is what I know is it's not going away. And if it's not going away and if the horse is indeed out of the barn, then you have two choices.
318. 55:07It it actually becomes very binary. You have two choices. You can sit on the sidelines and cross your arms and go, "It's theft. It's I'll
319. 55:16never use AI ever for the the most profoundly powerful self-expression tool in the history of humanity. Or you can say, "Well, that [ __ ]
320. 55:31sucks and I'm going to do what I can. Like maybe I'll even start a company. In fact, I've got a business plan for a
321. 55:38company right now for what what would a business what might be an interesting business in a postcopyright world? If
322. 55:47everything is infinitely copyable and the friction between making something and copying something is zero,
323. 55:55then what that ma means is that anyone that puts out something unique and in anything that is unique and interesting
324. 56:02is going to be instantly copied. We're witnessing it right now. So, I've got a business plan right right
325. 56:09now for how do artists and creators and aur make money in a world like that? and it involves blockchain and it
326. 56:19involves attribution. Like I'm I'm stoked about this, but it's a big ambitious idea and I'm in the middle of some other [ __ ]
327. 56:30But like that where I'm putting my energy in is is not trying to squeeze the toothpaste back in the tube, but to
328. 56:39say, okay, if this is here, what does that mean? And if I'm a creator, listen, I have a degree in acting. I have I have a a fine
329. 56:47arts degree in acting. I've written seven screenplays. Like, I get the issues. I get it. And these tools ain't going away.
330. 57:00So, like, what are your choices? And and I I honest to God feel if you're sitting on the sidelines going, "Wow, it's just
331. 57:08I would never use it." Well, you're going to get [ __ ] blindsided and it's going to suck. You know, it's mimemetic evolution, I
332. 57:16tell you. We're creating ecosystems of iterations with everything that hits the sweet spot. Exactly. I bet there's been
333. 57:23more AI art already produced now than all the previous art combined. Now they're stealing AI art. Like, does it suck that everything
334. 57:36changes? Yeah. But on the other side of that change is a whole another way of creating. Talk to Joy Perie.
335. 57:51She was a a sleep analyst for 30 years, but always had stories she wanted to tell, but she didn't have the skills.
336. 57:58She didn't have the she didn't go to art school. She didn't go to film school. She was putting putting her kids through,
337. 58:06you know, raising her kids, raising her family. And then along come these tools that allow her to create videos and all of a
338. 58:15sudden she's got a mechanism where she can start to tell her stories. Is she stealing from people? By any definition of the people that are
339. 58:29filing these lawsuits, yes, she is. But like would she have done anything remotely in the neighborhood of film
340. 58:44making and storytelling? Like it's very possible that someone like Joy Perie 5 years from now gets so good at generative AI
341. 58:52she'll win an Oscar. Like that's very possible. Like it is it is highly likely that in the next 5 years some film will win an Oscar
342. 59:04made by a person who was never a filmmaker. Why is that a bad thing? It's just different. It's a bad thing because the filmmakers
343. 59:17are like, "Wow, this is our craft and you can't change that." That's what the farmers said when the steam engine came out.
344. 59:26But the inevitability was the steam engine wasn't going away. If you can make a one-time investment that has the power of a hundred horses
345. 59:37and put wheels on it and put a plow behind it, you're not going back to horses. It's an inevitability.
346. 59:46That's where we are with AI. So anyway, all right. What do you think about chat GPT3 Pro model? I haven't played with
347. 59:54it. I don't I don't have the pro subscription anymore because I went I I gave my my uh ridiculous stipened uh for
348. 1:00:03for an expensive model to Gemini now so that I have access to VO. Um Nate B. Jones said it's really really good but
349. 1:00:11it takes like 15 or 20 minutes for every prompt and you can't give it shitty lazy prompts. You have to give it like big
350. 1:00:16problems to solve and then it'll go solve them. Um anyway, Gemini VO's like 250 USD. Yeah, I'm I'm paying so it's
351. 1:00:26it's $2.49 a month, but you get the first three months for one 125 half price. So I'm paying 125 a month for three months.
352. 1:00:36Also, my strong recommendation is for the next three years, don't buy any annual subscriptions to anything. Just do monthly.
353. 1:00:46But I save 10%. Yeah, but that tool that you think is amazing today is gonna be like the tired also ran like three
354. 1:00:53months from now. So, just go monthly with that stuff. Um, but VO is pretty good. Anyway, let's go back to
355. 1:00:58MidJourney and keep keep uh doing uh ranking videos. That one's better. Okay, that's weird. The balloon dog is
356. 1:01:14weird, but better. These are about the same. Two's or one's better. One's better.
357. 1:01:32Two's better. This is kind of cool. I Two. One. One, two.
358. 1:01:49One, two, two. One,
359. 1:02:08one, three. Those both sucked. one one like doing this. If you if if you're playing along, if you if you're new
360. 1:02:24since I've been doing this, go to midjourney.com/rank-vide slrank-vide and just do this. Like what we're doing when we do this is
361. 1:02:37we're actually helping to train the midjourney model to suck less. Right? They're they're putting all that out
362. 1:02:44there. I want to learn AI, VFX, and film making, but I'm still undecided of going Sora, flow, or cling.
363. 1:02:56Well, I would go I would go flow. Well, Sora, Sora, at this point, I would consider Sora probably
364. 1:03:10seventh or eighth down the list. Like, Sora is not great and it's expensive. Um, Luma Labs is good, Clling is good.
365. 1:03:25Ka is really interesting. Korea gives you access to lots of different video models for for a single subscription
366. 1:03:30price. Um, but quite frankly, VO3, like if you can get V you can get V3 now in a bunch of different models, but flow.google,
367. 1:03:40that's their little assembly tool is is really good. Um the other one if if you want to do Edson if you want to do um
368. 1:03:53character animation like people singing songs and things like that Hedra H E D R A is really amazing. Sora is me. Yeah,
369. 1:04:02Sora is I wouldn't I I definitely wouldn't um put put a big I I wouldn't spend 200 bucks a month on Chat GPT for
370. 1:04:10Sora. I I think I think there's there's many other many other models you can play with. Korea or Freepick are a great way to go.
371. 1:04:20You can test them all out. Hedra. Yeah, Hedra is really good. Hedra.ai, I think. And Hedra also has Flux context built
372. 1:04:29into it, which is super cool. Um, Hedra.ai. Yeah. [Music] Shang dang dang dang dang
373. 1:04:54narcolepsy isn't just being tired. Um, all right, back to this. Let's let's rank some more vides. [Music]
374. 1:05:11Two. That's bad. They're both bad. Three. Three. They both sucked.
375. 1:05:34Two. Two. Two. One.
376. 1:05:57[Music]
377. 1:06:17Oh no. Champy, do you want cheese? Two. You do? You want cheese? You want mommy to give you cheese? Yeah. You go out there and bunker. Go.
378. 1:06:39All right. So, anyway, you should go rank these like contribute. Let's make these models not suck so we can all go use cool models.
379. 1:06:50Um, one. Okay, I'm done for a bit. What else did I have here to show? Um, I like that. That's okay. So, I want to go to the Sam Alman blog.
380. 1:07:01Hedra, it literally is though. Julie, live, laugh, love. I've been impressed how well it's maintaining its style.
381. 1:07:20I did the 50% off cling. Oh, sadly. Sor is me. Um, so Sam Walman posted this today. Um, and I just thought it might be fun to just
382. 1:07:36read it. I read like the first half of it. It was pretty interesting and I just thought, "Oh, this would be fun to read
383. 1:07:41on the on the show." So, um, all right. Sam Alman, the gentle singularity. We are past the event horizon. That line
384. 1:07:56is that's a that's a really good opening line to a novel. It's a disconcerting opening line from a blog from Sam Alman.
385. 1:08:07Like I don't know if you've noticed, but none of the frontier companies are really talking about AGI anymore. Now,
386. 1:08:17none of them have declared that we've hit AGI, but they're all talking about ASI, artificial um um super intelligence,
387. 1:08:31which is the next event horizon. So, he starts he starts out with we're past the event horizon, which you know that says
388. 1:08:40to me that that they've got something figured out internally that we don't know about yet. That is big. The takeoff
389. 1:08:48has started. Humanity is close to building digital super intelligence. And at least so far, it's much less weird
390. 1:08:56than it seems like it should be. H robots are not yet walking the streets, nor are most of us talking to AI all
391. 1:09:05day. People still die of disease. We still can't easily go to space. And there's a lot about the universe we
392. 1:09:11don't understand. And yet we've built systems that are smarter by pe smarter than people in many ways and are able to
393. 1:09:17significantly amplify the output of people using them. The least likely part of the work is behind us. The least
394. 1:09:26likely part of the work is behind us. So I guess that means the hard stuff. The scientific insights that got us to
395. 1:09:35systems like GPT4 and 03 were hard one but will take us far. Okay, so this is this is super important. I think is everyone getting
396. 1:09:48interruptions in the stream? Wait, the people are not talking There are people not talking to AI all day. I
397. 1:09:57know, Silver Fox. You know what we call them, right? Weirdos. Those are the weirdos. You're not talking to AI, what's your problem?
398. 1:10:09This this little sentence right here, the least likely part of the work is behind us. The scientific insights that
399. 1:10:16got us to systems like GPT4 and 03 were hard one but will take us far. What that says to me is those were the brute force models.
400. 1:10:27Those were the models where they just threw as many GPUs and as much data at them as they could and they they um hand tagged
401. 1:10:39data sources, right? They categorized things manually and, you know, had farms of workers tagging things um often in
402. 1:10:49third world countries and apparently to the detriment of a lot of those a lot of those people's mental health. Um
403. 1:10:57what likely happens after the phase that he's talking about here is we start getting mo we start getting models that
404. 1:11:06are trained on synthetic data. So rather than human beings having to categorize existing data, we're just going to
405. 1:11:12generate perfect data that's perfectly tagged and train models on that new refined synthetic data. Um, we also
406. 1:11:21start to get in things into things like recursive self-improvement where the AI starts to improve itself.
407. 1:11:29So, that sentence right there for me kind of hints in that direction. Again, I said this was going to be a lot of
408. 1:11:35chat TMZ stuff. And I think Pate's in here. He's probably screaming at his at his computer screen right now. They're
409. 1:11:42not doing recursive self-improvement. AI will contribute to the world in many ways, but the gains to quality of life
410. 1:11:51from AI driving faster scientific progress and increased productivity will be enormous. The future can be vastly
411. 1:11:58better than the present. Scientific progress is the biggest driver of overall progress. It's hugely exciting to think about how
412. 1:12:07much more we could have in some big sense. Chat GPT is already more powerful than any human who has ever lived. I
413. 1:12:14absolutely agree with that. Like, who do you know that knows as much as that as Quinn does? It's staggering.
414. 1:12:24Hundreds of millions of people rely on it every day for increasingly important tasks. A small new capability can create
415. 1:12:31hugely positive impact. A small misalignment multiplied by hundreds of millions of people can cause a great
416. 1:12:38deal of negative impact. Right? So, so these tools even in their current form are quite powerful. 2025 has seen the ar
417. 1:12:45arrival of agents that can do real cognitive work. Writing computer code will never be the same. 2026 will likely
418. 1:12:53see the arrival of systems that can figure out novel insights. So, this is their that line right there. 2026 is
419. 1:13:02likely the arrival of systems that can figure out novel insights. That's stage four of AGI. The five stages of AGI
420. 1:13:10according to uh to to chat GPT are wait um uh yeah um chat bots which we got with GPT3 um reasoning which we got with 01 and
421. 1:13:26now 03 and 03 pro agents right so we've got codeex from chat GBT and we've got manis and we've got genpark and we've
422. 1:13:35got a bunch of different agents that are now going off and doing [ __ ] for you. The third the fourth level of AGI
423. 1:13:43artificial general intelligence is what they call innovators. So when you take reasoning engines and you combine them with agentic behavior,
424. 1:13:53you have things that reason and they do it over and over again and they come up with tasks and plans and they do it. You
425. 1:13:59end up with novel insights, novel science. 2027 may see the arrival of arrival of robots that can do tasks in the real
426. 1:14:11world. Um, a lot more people will be able to create software and art, but the world wants a lot more of both. And
427. 1:14:18experts will probably still be much better than noviceses as long as they embrace the new tools. This Edson, if
428. 1:14:25you're still here, this gets to your thing. If you're in any way on the sidelines because there are ethical
429. 1:14:33concerns of how they were trained, being a storyteller, an artist, a filmmaker who is AI capable is going to be a lot less disrupted than
430. 1:14:46someone who isn't AI capable. So, and I know you are. So, because you're still here and still still in the
431. 1:14:52conversation, you didn't huff off, which is cool. Um, generally speaking, the ability for one person to get much more done in 2030
432. 1:15:02than they could in 2020 will be a striking change and one and one many people will figure out how to benefit
433. 1:15:10from. So what he's talking about here is so he goes 2026 is likely going to see innovators. So we're going to see
434. 1:15:20scientific breakthroughs from agentic systems. 2027 we'll start to see robots more commonplace right in the grocery
435. 1:15:29store wherever and he's saying by 2030 the implication for me here is it's his level five of of AGI by 2030 and level five is um
436. 1:15:46but the the world wants a lot more of both and experts will probably still be much better than noviceses as long as
437. 1:15:52they embrace the new tools um a A lot more people will be able to create software and art. Generally speaking,
438. 1:15:59the ability for one person to get much more done in 2030 than they could in 2020. So level five of AGI is what what
439. 1:16:08they call did they call them companies? No, they called them organizations. So imagine now you've got you've got
440. 1:16:16reasoning models that can perform on their own and use tools and they can combine to make innovative breakthroughs. You combine all of these
441. 1:16:28into a managing agent that can dynamically create these things. And what's going to happen is we're going to
442. 1:16:35be able to say, "Hey, I want to start a business doing X." and your AI system will just spin up a bunch of experts
443. 1:16:44that can use tools and go do [ __ ] on the internet and make phone calls and negotiate contracts and write contracts
444. 1:16:51and execute contracts, all of which you'll be in control of and it'll just go do [ __ ] for you. Who's gonna benefit for that first and
445. 1:17:02biggest are people that show up to stupid [ __ ] live streams like this on a regular basis. The fact that you're curious enough to
446. 1:17:14be here and listen to me ramble about this [ __ ] means that you have a significant significant advantage on the
447. 1:17:23rest of the world because of what's coming. So just keep paying attention. The purpose of this channel is not for
448. 1:17:29me to teach [ __ ] I'm not qualified to do it. I'm not an expert at anything. The the thing that I am most proud of is
449. 1:17:39that I'm holding space for us all to be in the conversation. Just be in the [ __ ] conversation. Play, experiment, get burned out. Say,
450. 1:17:51"I hate AI. I'm never going to use it again." Go away for a month or a week. realize, "Ah, [ __ ] I should probably go
451. 1:17:57back into it." Go back into it. Be in the conversation. Come to these things. Don't listen to what I say at all. Just
452. 1:18:04drink or surf or make images. In the most important ways, the 2030s will may not be wildly different. In the
453. 1:18:13most important ways, the 2030s may not be wildly different. People will still love their families, express their
454. 1:18:19creativity, play games, and swim in lakes. But in still very important ways, the 2030s are likely going to be wildly
455. 1:18:27different from any time that has come before. We do not know how far beyond human level intelligence we can go, but
456. 1:18:34we are about to find out. In the 2030s, intelligence and energy, ideas, and the ability to make ideas happen are going to become wildly
457. 1:18:47abundant. Think about that. any idea that you have, any idea in the 2030s, you'll just be able to go, "Oh, I had an idea. Go do that and it
458. 1:19:00will happen in the world." That's bonkers. Um, in the 20 in in the 2030s, intelligent
459. 1:19:14intelligence and energy, ideas and the ability to make ideas happen are going to be wildly abundant. These
460. 1:19:23two have been fundamental limiters of human progress for a long time. With abundant intelligence and energy and
461. 1:19:30good governance, we can theoretically have anything else. Right? I, you know, I've personally experienced this where my my meat computer
462. 1:19:44is like if I want to have an idea, I've got to like go, oh, it's time to have an idea and I got to think of it. Then I
463. 1:19:52got to go to a whiteboard and I got to whiteboard [ __ ] and I got to come up with ideas and that's crap and I put in
464. 1:19:57two hours and in two hours I come up with 20 ideas. Well, I can come up with 20 ideas in 30 seconds with chat GBT,
465. 1:20:05right? So it's just I am absolutely the limiting factor in my ability to get ideas out of my head and into the world.
466. 1:20:14So you know the 2030s are anyone who chooses to lean into this stuff is going to have this remarkable capacity
467. 1:20:25already. We live in incredible with incredible digital intelligence after some initial shock. Most of us were are
468. 1:20:31pretty used to it. I would agree with that. I think this channel is a good example of, you know, these remarkable
469. 1:20:36models come out and a week later we're bitching about them. It's a little slow. It hallucinates. I don't like it. Um,
470. 1:20:50we very quickly go from being amazed that AI can generate beautifully written paragraph to wondering when it can
471. 1:20:55generate a beautifully written novel. or from being amazed that it can make a life-saving medical diagnosis to
472. 1:21:01wondering when it can develop the cures. Or from being amazed when it can create a small computer program to wondering
473. 1:21:07when it can create an entire new company. This is how the singularity goes. Wonders will become routine and then table stakes,
474. 1:21:18right? The thing that seems unimaginable now will just become routine. Oh yeah. Oh yeah. Jimmy started a uh a a company
475. 1:21:27over the weekend. Like he hit it big. He's he's up to a million dollar run rate and it's only Tuesday, right? And then it'll become table
476. 1:21:37stakes. Everyone will be able to do that. And then you're going to have to find new ways to rise above the noise.
477. 1:21:44It might help us with perspective to show someone who doesn't AI wait someone who doesn't AI something
478. 1:21:52we already think is interesting. I don't quite understand your your um comment there, architect. Um all right. We already hear from
479. 1:22:03scientists that are two or three times more productive than they were before AI. Advanced AI is interesting for many
480. 1:22:10reasons, but perhaps nothing is quite as significant as the fact that we can use it to do faster AI research. Okay, this
481. 1:22:24that sentence combined with the sentence up above that said that what got us to 40 and 03 were hard one but important
482. 1:22:37says to me that those models are now doing things to advance AI. Sorry. Not interesting. Easy. Like take
483. 1:22:4915 seconds to produce them a customized song. Yeah. Yeah. Yeah. Exactly. We may be able to discover new computing substrates or
484. 1:23:00better algorithms and who knows what else. It can do decades worth of research in a year or a month. Then the
485. 1:23:06rate of progress will obviously be quite different. From here on, the tools we have already built will help us find
486. 1:23:15further insights and aid us in creating better AI systems. Of course, this isn't the same thing as an AI system
487. 1:23:22completely autonomously updating its own code, but nevertheless, it is a laral version of recursive self-improvement.
488. 1:23:30Right? So, he's starting to hint at recursive self-improvement. We're not there, but we're starting to get there.
489. 1:23:38Right? These systems are starting to augment how we used to do it. There are other self-reinforcing loops
490. 1:23:47at play. The economic value creation has started a flywheel of compounding infrastructure buildout to run these
491. 1:23:54increasingly powerful AI machines and robots that can build other robots and in some ca and in some sense data
492. 1:24:01centers that data centers that can build other data centers aren't that far off. Right. And so, like, you know, you think about,
493. 1:24:14you know, Elon's quest to to, you know, get rockets out to Mars where my mind was historically was, you know, um, Matt
494. 1:24:24Damon up on the planet growing potatoes and poop, right? There's a human being up there trying to save the camp for the
495. 1:24:33rest of the humans to come get him. Well, that's likely not how it's going to go, right? We'll send up a bunch of parts
496. 1:24:41and a bunch of robots, and those robots will build more robots and build more data centers and build whatever the [ __ ]
497. 1:24:49they need to build. And by the time humans get there in any number, the infrastructure will be built, right?
498. 1:24:57Like, so like we ain't that far away. like by about about the time SpaceX figures out how to get you know the the
499. 1:25:08the you know the uh the giant the giant rocket you know out there and back all this stuff will be fairly
500. 1:25:18straightforward and fairly common if we have to make the Oh, thank you very much. Who is that? Starf Fighter. Thank
501. 1:25:24you very much. I appreciate that. Oh, sweet. Um, if we have to make the first million humanoid robots the old-fashioned way,
502. 1:25:36but then can operate the entire supply chain, digging and refining minerals, right? So you're on Mars. They build a
503. 1:25:45truck. They go dig some [ __ ] Building and refining minerals, driving trucks, running factories to build more
504. 1:25:53robots, which can build more chip fabrication facilities, data centers. Then the rate of progress will quite
505. 1:25:59obviously be different. A data center production, as data center production gets automated, the cost of intelligence
506. 1:26:07should eventually converge to a cost of electricity. People are often curious about how much energy a chat GPT query uses. This is good.
507. 1:26:17The average query uses about 34 watt hours, about what an oven would use in a little over 1 second or a high
508. 1:26:28efficiency light bulb would use in a couple of minutes. It also uses about 0.000085 gallons of water. Roughly 115th of a
509. 1:26:37teaspoon. Right. This is this is I I guess why he's putting this in here. A lot of the doomers and and lites are
510. 1:26:45right now are like well it's it's eating all of our electricity and it's wasting all of our water. Um it's not and this
511. 1:26:52these things are going to get more efficient. The rate of technological progress will keep accelerating and it
512. 1:26:57will continue to be the case that people are capable of adapting to almost anything that I totally agree with. One
513. 1:27:04of the tropes of of the doomers is that it's going to take away all the jobs and then what are we supposed to do? Well, I
514. 1:27:11don't know. You're [ __ ] human. Like, we'll figure it out, right? We'll figure out. Oh, well, now that everyone lost
515. 1:27:18their job, there's a job I can do over here to support these people and they can do something and then and like we're
516. 1:27:27humans. We'll figure it out and we'll figure out economic mechanisms to deal with all that stuff. In fact, what did I just hear today I
517. 1:27:37thought was really interesting? Oh, David Shapiro, I think he was talking about the copyright thing that there could be a token tax
518. 1:27:47that for every token you use of chat GPT, some percentage of that goes to a fund for copyright holders, right? So
519. 1:27:55there's ways that you could basically say like you could basically have ChatGpt and Anthropic and Meta and
520. 1:28:01Google all admit, okay, yeah, we trained it on all your [ __ ] We're going to create a fund kind of like Spotify
521. 1:28:09where if your if your [ __ ] was in our [ __ ] then you're in the fund. And I don't know how they figure out how they
522. 1:28:16divvy it up, but you know, you could do a tax on tokens. So there could there could be something like that. Okay. um [Music]
523. 1:28:27there will be very hard parts of whole classes of jobs going away. But on the other hand, the world will be getting so
524. 1:28:35much richer so quickly that we'll be able to ser seriously entertain new policy ideas we never could before. We
525. 1:28:43probably won't adapt a new social contract all at once. But when we look back in a few decades, the gradual
526. 1:28:50changes will have amounted to something big. If history is any guide, we'll figure out new things to do, new things
527. 1:28:56to want, assimilate new tools quickly. Job job change after the indust industrial revolution is a good recent
528. 1:29:03example. Expectations will go up, but capabilities will go up equally quickly, and we'll all get better stuff. Hang on.
529. 1:29:18We will build ever more wonderful things for each other. People will have long-term import important and curious
530. 1:29:24advantage over AI. We are hardwired to care about other people and what they think and do. And we don't care very
531. 1:29:32much about machines. This is so important. the value of this community, the value of the AI salon, the value of your friends and your family
532. 1:29:44that goes up in value as the tasks of work get handled by the machines. So I can absolutely see where the
533. 1:29:55machines are doing all of the kind of drudgery of the world and we as humans are finding new ways to connect and new
534. 1:30:03ways to make meaning for one another. And there's going to be incredible value in that. A subs a subsistence farmer from a
535. 1:30:15thousand years ago would look at what many of us do and say we have fake jobs and think that we are just playing games
536. 1:30:23to entertain ourselves since we have plenty of food and unimaginable luxuries. I hope we'll look at jobs a
537. 1:30:30thousand years in the future and think they look like fake jobs. And I have no doubt that we will feel they will feel
538. 1:30:37incredibly important and satisfying to the people doing them. Right? Like you could have someone's whose job it is to
539. 1:30:46sit on the corner and wave at people and wish them a lovely day. Hey, how's it going? Hey Jim Janie, hope you're having a lovely day. And
540. 1:30:56they get a salary for that. and they just love doing that. Like it it could be [ __ ] like that. I know I know a bunch
541. 1:31:02of you in here are probably trolling me for that, but that's okay. That is a truly perfect quote. We are
542. 1:31:10hardwired to care about other people and what they say and do and don't care very much about machines. Yeah, right.
543. 1:31:19This channel is not about AI. That's just a convenient mechanism for us to meet around, right? You know, we experienced that this week.
544. 1:31:33We we lost a very very close friend this week. Nothing to do with AI. Nothing to do with AI.
545. 1:31:51Yeah, exactly. Silver Fox network of human irreplaceable irregulars. Yeah. Exactly. Exactly. Yes. The rate of new wonders being achieved
546. 1:32:06will be immense. It's hard to even imagine today what what we will have discovered by 2035. So we're in 2025. So imagine 10 years
547. 1:32:16out. So think back 10 years. What were you doing 10 years ago? So for me 10 years ago is 2014 we just round our
548. 1:32:25raised our seed round for Story Vine like yeah it was like social media was kind of blowing up
549. 1:32:44like none of this stuff that I'm focused on today was even a whisper of possibility. So imagine 10 years in the
550. 1:32:50future. That's crazy. All right. It's hard to even imagine today what will be discovered by 2035. Maybe we will go
551. 1:32:58from solving high energy physics one year to beginning space colonization the next year or from a major material
552. 1:33:06science breakthrough one year to a true high bandwidth brain computer interfaces the next year. Many people will choose
553. 1:33:12to live their lives in much the same way. But at least some people will probably decide to plug in, right? Instant access to all knowledge.
554. 1:33:26[Laughter] Looking forward, this sounds hard to wrap our heads around, but probably living through it will feel impressive but manageable.
555. 1:33:37From a relativistic perspective, the singularity happens bit by bit and the merge happens slowly. We are climbing the long arc of
556. 1:33:48exponential technological progress. It always looks vertical looking forward and flat going backwards, but it's one
557. 1:33:56smooth curve. That's that's really interesting, right? We're we're on a curve and you're always looking, you
558. 1:34:01know, you're always looking forward and it's up, right? And it looks steep, but then when you look back, it's long and slow.
559. 1:34:11Think back to 2020 and what it would have sounded like to have something close to AGI by 2025 versus what the
560. 1:34:19last five years has actually been like. I mean, this channel's two and a half years old. Like, we've for those of you that have been around
561. 1:34:29for for most of that time, like this like the the advancements the advancements in 2024 were insane.
562. 1:34:39the advancements so far in 2025 are just just it's it's like it's to the point you can't keep up with it anymore and it
563. 1:34:49doesn't matter and it's like they dropped the new 03 Pro model and it's like okay I couldn't really utilize 03
564. 1:34:56all that well it it's beyond my capability to even know what the [ __ ] to do with it. So 03 Pro. Okay. Uh
565. 1:35:06I don't know. Go solve some municipal bond problem I don't give a [ __ ] about. Um there are serious challenges to confront
566. 1:35:19along with huge upsides. We do need to solve the safety issues technically and society. But then it's critically
567. 1:35:26important to widely distribute access to super intelligence given the economic implications. The beth ah this is good
568. 1:35:34the best path forward might be something like solve the alignment problem right that's what Dario Amade over at anthropic wants to do.
569. 1:35:46Openai's lost their alignment team multiple times now. So maybe that's one of the reasons Sam's like, "Ah, we got
570. 1:35:53to figure out this alignment thing. Could we get some people to stay here, solve the alignment problem, meaning
571. 1:35:59that we can robustly guarantee that we get AI systems to learn and act towards what we collectively really want over the long term."
572. 1:36:09That's a fascinating statement. What we collectively really want over the long term, who decides that? That's a really interesting question.
573. 1:36:19Social media feeds, for example, are an example of misaligned AI. The algorithms power those who are incredible at getting you to keep
574. 1:36:29scrolling and clearly understand your short-term preferences, but they do so by exploiting something in your brain
575. 1:36:35that overrides your long-term preference. All right? So, basically, make the [ __ ] safe. Two, then focus on
576. 1:36:43making super intelligence cheap, widely available and not too concentrated with any person, company or country.
577. 1:36:52Society is resilient, creative, and adapts quickly. If we can harness the collective will and wisdom of the
578. 1:36:58people, then although we'll make plenty of mistakes and some things will go really wrong, we will learn and adapt
579. 1:37:04quickly, then be able to use this technology to get maximum upside with minimal downside. Giving users a lot of
580. 1:37:11freedom with broadbounds society has to decide on seems very important. The sooner the world can start a conversation about what these
581. 1:37:20broadbounds are and how we can define collective alignment, the better. Not a lot of guidance there. He's like,
582. 1:37:29uh, we should figure this out. Hey guys. Guys. Hey. Yeah. Hey. I was wondering like I mean assuming we figure
583. 1:37:39out how to get the robots to not kill us. Um you think it'd be cool if maybe we got together and figured out like
584. 1:37:45what we wanted them to do for us and stuff man you know like like think we could get together like as a global
585. 1:37:51tribe and or maybe just like seven of us at a at a dinner party in Palo Alto and figure out kind of the future thing maybe
586. 1:38:03whatever. All right, we the whole industry, not just open AI, are building a brain for the world. Um, it will be
587. 1:38:09extremely personalized and easy for everyone to use. We will we will be limited by good ideas. For a long time,
588. 1:38:17technical people in the startup industry have made fun of idea guys. That's true. People who have an idea that were
589. 1:38:25looking for a team to build it. It now looks to me like they are about to have their day in the sun. Revenge of the
590. 1:38:31liberal arts major, baby. Revenge of the Gen Xers. Bring it the [ __ ] on. Creative generalists are about to become superstars.
591. 1:38:45Mark my words. And another thing, you Paluca. Yeah. Yeah. See, for a long time, the technical people in
592. 1:38:55the startup industry have made fun of the idea guys. I'm an idea guy. Yeah. Yeah. You pula, stop making fun of me, you nerdy little dweebs.
593. 1:39:06Yeah, you were cool in the 90s and you were super cool in the 2000s and then 2010s you were like even cooler. Then in
594. 1:39:1320120s you were like super cool and then you were all like Sam Alman cool in the 2025s, but in the 2030s it's going to be the idea man.
595. 1:39:25Damn it. [Laughter] Open AAI is a lot of things before anything else. We are a super intelligence research company. That is
596. 1:39:35not how they started out. They started out as an AGI company. This is a change in tone. This is a change in
597. 1:39:42what what frontier companies are talking about. Dweeb Dwee Mcwebington. Dweeberton. Exactly. Exactly. [Laughter]
598. 1:39:59Okay. Um, we have a lot of work in front of us, but most of the path but most of the path in front of us is now lit
599. 1:40:09and the dark areas are receding fast. We feel extraordinarily grateful to get to do what we do. Intelligence so cheap to
600. 1:40:16meter is well within grasp. Intelligence too cheap to meter is well within grasp. This may sound crazy to say, but if we
601. 1:40:26told you back in 2020 we were going to be where we are today, that would sound more crazy than our current predictions about 2030.
602. 1:40:33May we scale smoothly, exponentially, and uneventfully through super intelligence. Super intelligence. He's not talking
603. 1:40:41about AGI anymore. That says to me they've either hit it or they're on the path to it internally. Right. No more AI. No more AGI talk.
604. 1:40:52Yeah. Yeah. That was got there. Next. I was thinking earlier, Revenge of the Nerds in real life. Yeah, I know, right?
605. 1:41:06All right. What time is it? It's getting there. Um, let me show you my my presentation that I did. Anyone any
606. 1:41:12thoughts, questions? That's a hell of a post, isn't it? It really is. Really is. It's like I know Sam Alman's problematic
607. 1:41:22and he's got weird vocal fry that drives us all crazy. He's like he's always throat singing. Um but he but he's he's quite a thoughtful
608. 1:41:35guy. He's quite a thoughtful guy. I can't even imagine. I can't imagine what it must be like to
609. 1:41:50be at the tip of the spear of this stuff. It's just surreal. Really is. Um,
610. 1:42:07all right. So, this Let me make this a little smaller for the Tik Tok ears. [Music]
611. 1:42:27So, feed your prompt. So, I ended up getting this ma mouth this morning at like 6:30. I've been trying to get a
612. 1:42:34mouth in chat GPT that didn't suck for a long time. What? You going to go hang out with the boys? You think they're going to give you
613. 1:42:43cheese? Is that what you think? You think the boys are going to give you cheese? Huh? I like that. Isn't it good? All right.
614. 1:42:59Go say hi to his voice. [Music] So, this is my new big idea. feed your prompt
615. 1:43:16and amplify yourself and your team with AI and you know the basic idea is that the prompt is hungry and it'll it'll devour
616. 1:43:26anything you you throw at it. Um so five stages of AI adoption we've covered this in this channel but the basic idea of you know denial
617. 1:43:39where a lot of people are right now you know it's it's evil blah blah blah stage two is that shock your first Kevin Mallister moment um
618. 1:43:52hang on get things lined up over here. The third stage is that stage where I think a lot of us are wonder and fear at
619. 1:44:04the same time where you're like, "Oh my god, this is so amazing. The possibilities are endless and holy [ __ ]
620. 1:44:09this is terrifying. What's going to happen?" Right? That's both. That's happening. The fourth one is the
621. 1:44:15augmented me where you realize, well, AI is not going to replace us. It's it's amplifying me and I get to do stuff with
622. 1:44:22it. And then the fifth one, the fifth stage is cognitive hyperabundance. I can do anything I imagine, literally
623. 1:44:29anything. I don't think the tech is there for stage five to truly be there, but you know, some people are all in on
624. 1:44:36AI and they're probably kind of living in that fifth sta stage. I I tend to oscillate between three and four, wonder
625. 1:44:43and fear and the augmented meat. Um, so ask the right questions. I think we should stop asking the question, how do I get the most out of
626. 1:44:54AI? This is the question that this channel was built on and I think it's the wrong question. What if AI
627. 1:45:02eventually does want to replace us? Not now, but in 20 or 100 years. Um, I don't know. I mean, I don't know. I don't know, John.
628. 1:45:17The only the the only thing I know here's what I know about that question. If you use that question as a reason not
629. 1:45:32to go all in on AI because at some point the AI might decide we're not necessary. The the paperclip problem, right? It it
630. 1:45:43was tasked with making the most paper clips it could and then at some point it decides that it just has to get rid rid
631. 1:45:49of humans to make more paper clips. If you use that in any way as an excuse not to participate in it, then I think
632. 1:46:00it's a bad question. If you're asking that question as a question to say, hey, let me jump into it and let's start
633. 1:46:08thinking of ideas for how we create sustainable alignment of AI, right? So, we might solve the alignment
634. 1:46:17problem in the next 5 years, but how do we sustain that? That might be something interesting to start researching. So,
635. 1:46:24anything right now that that lands on me like this is a reason I'm not going to play with AI, I don't like. So, humans
636. 1:46:31are busy creating new species. Our succession um well, most of us don't realize this. Yeah, I agree. Julie, live, laugh, love.
637. 1:46:42Okay, so anyway, ask the right question. I think the question we should stop asking is how how do I get the most out of AI?
638. 1:46:50I think the question we should start at asking is how do I get the most out of myself with AI? It's a very subtle difference, but it's
639. 1:47:03a really important one, right? How I get the most out of AI says I want to be able to manipulate this tool. I want to
640. 1:47:10engineer this tool. How I use AI to get the most out of myself is a very different thing. So, um, the prompt is hungry, right? It's
641. 1:47:21just sitting there. It always wants to eat. It'll eat anything you throw at it. I think one of the expectations of AI
642. 1:47:27right now is that it's a genius. And I don't think AI is a genius. I think what AI is is an amplifier. It will amplify anything you feed it.
643. 1:47:41And here's the good news. This is what's exciting. We get to choose what we have AI amplify. And my contention is
644. 1:47:54you can have it amplify crappy stuff and crappy prompts. What I'm calling an IT prompt. If you think of AI as an it, it's a tool
645. 1:48:05and you feed it cold generic prompts, you get back cold, disappointing results, right? And we've all done it.
646. 1:48:13Write me a blog post about blah blah blah. And then you read the blog post, you're like, well, that's not how I would write it.
647. 1:48:19Well, that's an IT prompt. and you get an it result, right? Crap in, crap out. A you prompt, a prompt that is centered around you is
648. 1:48:30when you feed it you. Generative AI wakes up and and it amplifies your genius. So, so what does that mean? Feed it.
649. 1:48:40What it means is you feed it your ideas, your challenges, your questions, your point of view, your personality. Well,
650. 1:48:48that doesn't write like I wrote, you know, that didn't write the blog post like I would write it. Well, how do you
651. 1:48:53write it? Why don't you share that with it? That's a you prompt. The prompt is, well, here's how I would do it. Oh,
652. 1:48:59okay. I'll do it that way now. Right? Making sense? So, an IP prompt might be something like write a blog post for designers.
653. 1:49:11That's a very sort of what what Google has trained us for 20 years, right? Don't put much in the prompt box. Be
654. 1:49:17very efficient with it. Don't get too flowery or you get all sorts of weird results. You It's very transactional. You put
655. 1:49:25something in, you get something out, you're done. That's an it prompt. A you prompt example might be something like
656. 1:49:32this. I'm a confident early career coach that used to be insecure. Write a blog post in my voice for entry-level
657. 1:49:40designers focusing on confidence using a tone that's warm but direct. Right now all of a sudden I've got given
658. 1:49:49it a lot more context about who I am, what I want, what I what I expect in return, who the audience is, what value
659. 1:49:55I want to provide to them. Minor shift, massive sh massive change. Now, the good news about this whole
660. 1:50:06presentation and this this whole idea, I think that the vast majority of of active members of the AI salon and the
661. 1:50:15AI learning lab have already naturally organically discovered this. I think that that the people in this community
662. 1:50:25are some of the best people on the planet. I mean that quite seriously. some of the best people on the planet
663. 1:50:32at understanding the nature of this interaction. The nature of the fact that AI can shapeshift and if you shapeshift
664. 1:50:40it to your end, it will amplify your ideas. It will amplify your genius. Um, so this was kind of a cool thing. I I
665. 1:50:56just I was talking to Ann Murphy on our podcast today about this. Um, we had CJ Fletcher on last week and CJ Fletcher
666. 1:51:06was talking about he he's an artist and he does he does AI art and like really incredible well thought
667. 1:51:15out art and and what he said he does is when he sits down at the prompt he takes a moment he takes a moment he looks at the prompt
668. 1:51:25and he closes his eyes and he breathes and then he starts asking himself questions like what do I want here? Um
669. 1:51:33why does this matter to me? Who's this for and how do I want them to feel? What would success look like?
670. 1:51:41So he's taking a human moment before like the thing about the the AI prompt, no matter what tool you use, it doesn't
671. 1:51:48matter if it's chat GPT or Midjourney or any of them, that prompt is hungry and it's like it will give you immediate
672. 1:51:55feedback. So, it's very easy to just sit down and just start vomiting [ __ ] into the prompt hole and getting [ __ ] vomited
673. 1:52:01back at you. I love this moment. Like, take a moment. Be human for a moment. Check in with yourself. What do I want? What's my
674. 1:52:11point of view right now? What does good look like? Who's the audience? What would give them delight? Whatever
675. 1:52:21that is, right? take that moment. And then I came up with this concept of what I'm calling the prompt ladder. And
676. 1:52:30I there there's a lot of sort of structured prompting systems that do that. One of the things I want this
677. 1:52:36presentation to be is is effectively the anti-prompt engineering presentation. The anti-prompt engineering keynote. I
678. 1:52:45hate the term prompt engineering in ways I I can barely describe. But but the primary reason I hate that term is that
679. 1:52:59it carries on this myth that only engineers can do this. You have to be a prompt engineer. You have to be really
680. 1:53:07smart. You're going to you're going to be engineering prompts. Oh, if you want good results, you're going to have to engineer that.
681. 1:53:15People are already [ __ ] freaked out enough, right? Like, like it's Have Have you tried to buy an Apple recently? When
682. 1:53:22did apples become $17? And then on top of that, I got to learn prompt engineering. You [ __ ] kidding me?
683. 1:53:31So, so feed your prompt. The idea behind feed your prompt is you you already know how to do this, right? And so the prompt ladder idea is
684. 1:53:42you don't have to get fancy. You can start really simply. You can start with like a snack. You can feed your prompt
685. 1:53:48like a one-s sentence request. Write a headline about remote work and it will do that. You need something about remote
686. 1:53:54work, do that. But then you can sort of level up and make it a meal, right? You could add tone. You can add audience.
687. 1:54:01You can add outcome. Write a playful headline about remote work for a blog aimed at startup founders. Woo! All of a
688. 1:54:09sudden, put a little more meat on the bone, right? A little more meat on the plate. And that's that's kind of the the
689. 1:54:16next level up the ladder. The next level up is a feast. You can layer in goals and structure and story and nuance.
690. 1:54:24Write three possible headlines for a thought leadership blog on remote work aimed at skeptical startup founders. Not
691. 1:54:31just startup founders, skeptical ones using a tone that's confident, persuasive, and informal. Oh, okay. So now all of a sudden to write
692. 1:54:43something like that feast prompt, you have to be decently in touch with what you want, what your point of view is.
693. 1:54:50And if you are and you feed your prompt, that point of view, [ __ ] starts to get really good. Sharon Crawford, I can say
694. 1:54:57honestly that I don't send a professional email without using AI to polish my tone. After submitting my original email, helps me really
695. 1:55:04understand what I'm really trying to say. It's great. Being a dra jack of all trades and having um a ready niche
696. 1:55:13becomes a personal obstacle. Oh, not having a ready niche becomes an obstacle. Wolfman clan, I don't think it's for that long. The jackof
697. 1:55:22alltrades, master of none, I I I honestly think that in the next three years, any of us that have that tendency to try to do everything are
698. 1:55:35going to be rewarded. We're going to be a jack of all trades, master of all because the AI is going to do the tactical bits.
699. 1:55:44We've always had the ideas. The the jackof alltrades thing that that's such a [ __ ] disparaging thing. I think
700. 1:55:50this iceberg is about to flip. The jackof alltrades master of none is basically the bottom of the value chain
701. 1:55:59right now are creative generalists. Oh, you have ideas. Oh, you're you're trying to figure out what our ringtone would
702. 1:56:08sound like as well as trying to figure out how to code the computer program and you dabble in art, too. Yeah, you're a jack of all trades,
703. 1:56:18master of none. You're at the bottom of the value chain. I think that iceberg is literally about to flip
704. 1:56:26where at the bottom of the value chain in the future will be specialists that cling on to a single way of doing something, a single path
705. 1:56:38because the tools are going to be so good at doing that. All right, so after the feast is the buffet. And this one's
706. 1:56:43really fun on the prompt ladder. And again, this is something that I think this audience, this community does
707. 1:56:49really, really well is when you're prompting, not only can you sort of dial up the substance of the prompt that you put in
708. 1:56:58there through filtered through your point of view and your goals, you can also do it for multiple outputs, right?
709. 1:57:06So, here's a text prompt. Here's an image prompt. Here's an audio prompt for a script that you're going to put into
710. 1:57:1211 Labs and turn into a thing and make a headless video on TikTok. You can easily sort of navigate between
711. 1:57:19all these different tools in the same kind of imperative, the same kind of uh uh um project, right? You can just
712. 1:57:30navigate between them very very quickly. Um, this is one of my favorite slides. So, we talk about Rick Rubin in here a
713. 1:57:38lot, right? You're the producer now. One of the great myths of AI right now. And one of the one of the tropes that I hear a lot is, "Well, I
714. 1:57:51put something in and it gave me a shitty result." I even hear people that I know that are really into AI will be like,
715. 1:57:57"Oh, well, I did that prompt and it didn't give me anything good back." And I'll ask them, well, did you tell it it
716. 1:58:02wasn't good? Oh, no. Oh, you can do that? Like, a lot of people don't know you can do that. So, the myth of you
717. 1:58:09push a button and outcomes perfection is [ __ ] The truth is AI becomes more powerful when you're the curator, the
718. 1:58:17creative director, the producer, when you have taste, when you have a point of view. Um, and rather than think of AI as as a
719. 1:58:29transaction, like a Google search, I put in a prompt, I get back an answer. It's a conversation. It's this two-way
720. 1:58:36ongoing dialogue, right? Great AI outputs are sculpted. It's a process. It's a dialogue, right? It's you shape
721. 1:58:44it. You refine it. You react to what it puts out. Rarely do first takes win. I I would argue I don't know Rick Rubin. I
722. 1:58:54don't know his story and all the things. I can't imagine there was a song in his in all of the songs that he's produced
723. 1:59:03that the first take was the one that made it on the album. Like rarely, right? Why would that happen? It it doesn't.
724. 1:59:15So the magic is in your judgment over time. So, not only make the prompts more personal and filter through your point
725. 1:59:23of view, but understand that this thing's not going to be perfect when you do it. Again, I know I'm kind of
726. 1:59:28preaching to the choir here. And then I added a slide in about from me to we, how you go from an individual thing to
727. 1:59:35you can take the same basic idea of your personal point of view and now start to think about it for your team at work,
728. 1:59:42your family, your community group, your your bowling buddies, right? And then now you can start to write our
729. 1:59:51prompts rather than you prompts. You can write our prompts and you can put in your team's mission and brand attributes
730. 1:59:56and strategic challenges and goals and briefs and guidelines. Um and then there's an example prompt of that. And I
731. 2:00:03think that's it. Yeah, that's it. Um when you feed your prompt with purpose, personality, and potential, when you
732. 2:00:12feed it you, it will change everything. AI isn't powerful because it's smart. It's powerful because it amplifies your genius. So,
733. 2:00:25that's the deck. That's what I presented today. So, there you go. And look, there I am as like a hip sci-fi dude. I
734. 2:00:32doubled my fundraising this year, raising 100K for my Humane Society by implementing AI. Perfect. Julie live, laugh, love.
735. 2:00:42[ __ ] awesome. Like you wait, you get one of these. You can make money with chatb. You can make money with chat GBT. You
736. 2:00:51doubled your income. That's awesome. Congrats. Tip: Use AI to craft and shape your final prompt. Yeah, exactly. That's the other thing.
737. 2:01:00Like again, like part of part of creating a you prompt is giving chat GPT or whatever tool you're using enough context to give you
738. 2:01:12what you want. And some of that context might be, I don't know how to write this prompt. Oh, okay. I'll write it for you. Or you
739. 2:01:21could say something like, you know, create a structured prompt um that that captures all of the possible details of the thing that I'm
740. 2:01:30that I just put in there that gave us a crappy result. And it'll write you a whole framework for how to prompt. And
741. 2:01:37then you can just have it write the prompts using the framework that it invented on your behalf. It's a great
742. 2:01:43way to do it. Great way to do it. Um, all right. We need a dessert phase in the prompt letter. Ah, that's a really
743. 2:01:52good idea. Oh, actually, you know what, Sharon, that's a really good idea. In fact, let me go. [Music] Really good idea.
744. 2:02:20Reward yourself with something tasty. I'll And then I'll put a little note to myself. Figure this out.
745. 2:02:33That's a really good idea. really really good. Um yeah, anyway, I would love your thoughts if if if you paid attention to
746. 2:02:40what I just went through. Um I would love any thoughts you have on it, ways to improve it, things like that. Um I think this is going to be
747. 2:02:48something decently foundational um for what I talk about. What what I'm excited about this is that it's independent of
748. 2:02:57tool. It is completely independent of tool. And I think that this shift from thinking of AI as some sort of efficiency tool to AI
749. 2:03:09as a humanity as as a as a U amplifier. Um I think I'm going to be able to talk about this for years and have it be
750. 2:03:18relevant to people. Good-looking deck. Thank you. Yeah, I'm really excited about it. I also realized I can do I'm going to go
751. 2:03:25into um I'm going to go into Photoshop. I assembled this all in Photoshop. I'm going to go into Photoshop and and do a
752. 2:03:32bunch of different colors of this. So, this is the blue one. I'll have a green one, a red one, a yellow one, orange,
753. 2:03:37right? Do all of the rainbow. So, in different contexts, I can have different colors um for like either different
754. 2:03:44sections or maybe if I'm doing something for a brand, I want it to match the brand. Is that a McLaren race suit, Kyle? It's
755. 2:03:51actually pretty close, isn't it? Yeah, I'm driving for McLaren in 2037. So, what utility did you use for your
756. 2:04:02presentation? Um, so I did this image here. I can show you the the the evolution of this was kind of fun. Um, is this it?
757. 2:04:18No. This was all the images I did for the presentation itself. I just did paper cut
758. 2:04:34style. It gave me shitty ones that I didn't like. But there's the Rick Rubin one. And then there's the there was the the
759. 2:04:47Rick Rubin one where he had too much hair and he was too orange. There's the ladder. There's a shitty ladder. There's a shittier ladder.
760. 2:04:56There's the the meditating dude. So, here's where I got the I got pretty close. Oh, wait. No, this this I just uploaded. No, I just did
761. 2:05:11that for colors. That was the Photoshop thing. Um, there's the prompt graphic. Here's the it prompt graphic.
762. 2:05:19Did I have the There's the stop start. I think it was a different chat. I had the mouth thing. Hang on.
763. 2:05:28Feed your prompt book outline. Was this it? Oh, look. This is This was um a really bad version of the mouth, right? Like that's horrible.
764. 2:05:43Horrible. Like just like embarrassing. Then there was like the kid monster version of it. Bad. Um,
765. 2:05:59let's see. Turtle automation creating my backstory. Is it this?
766. 2:06:16I forget which which chat I was in. Oh, you know where I can go? Hang on. I can go to my library.
767. 2:06:37Can I go back to that prompt? No. Which one was it? That sucks. Um
768. 2:07:05that one is here. I want to show you the mouth evolution because it was kind of cool.
769. 2:07:19Okay, here it is. So, so this image did did it generate this image?
770. 2:07:41Oh, remember the other night I was trying to do [ __ ] make Sydney images and they were all sucking. This was that chat
771. 2:07:50and then I think I found this on the internet. I think I found this image on the internet and it's it's pretty
772. 2:07:56horrible, right? Like it's sort of torn paper style but sort of digital background and I said I want a more
773. 2:08:03sophisticated version of this. Hang on, I got to sneeze. And it gave me this which is kind of cool but like I didn't like it. It looks
774. 2:08:14like a pair of dentures, right? And so I said, rather than square it square edges, can you make it lips? And
775. 2:08:25then it made this. And that's this is actually really close. This is kind of in the neighborhood of like a Rolling
776. 2:08:31Stones thing, right? But without the tongue and without the sexual stuff. Um, but I didn't like this digital [ __ ]
777. 2:08:39behind it. And then I said, "Can you make it without the digital ship behind it?" And I was like, "There it is." is
778. 2:08:44I'm like, "Holy [ __ ] that's really good." So then what I did was I took that over to Photo Slop.
779. 2:09:04And in here, I like masked off the teeth. I made my own prompt box, you know, like there's the there's the
780. 2:09:16thing with the teeth marks out of it. I added in the typography. I expanded the background. Wait, can I
781. 2:09:36I don't know. Let's see. Anyway, expanded the background with generative fill and then there you go. And so now and
782. 2:09:45then I also created a version of this where I took the mouth out al together and just had the gradation. So that was
783. 2:09:50the all the background slides were out of the same thing. So it was generated in in chat GPT and then um
784. 2:10:01and then this was the beginning of it. And then what I did was I think I took the Manis version of the thing and I
785. 2:10:09just looked at it for content because it had consolidated most of the slides and then I just ended up handediting hand
786. 2:10:16editing all of these slides. I just hand edited what I put what I chose to put in there. and like how to how to lay it out and
787. 2:10:25how to arrange it. So, it took me it took me a while. It's only it's only 10 slides, but you know, but they're they're pretty coherent.
788. 2:10:35I don't like that image. That feels like clip art to me. So, I'll probably replace that image. But, I kind of like
789. 2:10:41like I like the Rick Rubin one. The ladder's okay. That dude I like. I like the it prompt and the you prompt. I now
790. 2:10:48have this thing is now cut out as a as a like I can move this around without the background. So I've now got a little brand kit. And
791. 2:10:58now imagine, you know, eight different colors of this. I can do all sorts of cool [ __ ] So all right, that was it.
792. 2:11:08How do I get this guide? Which guide? Oh, this presentation. I It's not out yet. It's I just wrote it today. I mean,
793. 2:11:17well, I wrote it two days ago and then I built it today. Um, I think I need to get it another level or two more sophisticated and then I'll
794. 2:11:30and then I'll probably put it out in the world. The slides were done by Chat GPT. No, the slides were done in Keynote by
795. 2:11:36me. I did all these slides manually. the outline. Well, the outline was written in chat GPT, but I like it was
796. 2:11:46like I guided it a bunch, right? Like like this slide right here, like it's conversation. I didn't just vomit out the slides.
797. 2:11:56Uh uh uh uh uh uh. Someone told me lead pages are framer. [Music] All right, I'm getting out of here. It
798. 2:12:09is 10:10. People, get your ass to bed. What are you still doing here, you dang losers? Peace out. Hope you have fun tonight. Bye.