WEBVTT 00:00:00.000 --> 00:00:04.200 Are you good? Yeah, Elijah's out there once and then now we have to deal with it forever. 00:00:04.200 --> 00:00:07.700 I feel like I should crash the WAN Show again at some point. Yeah, that was fun. 00:00:07.700 --> 00:00:14.500 That was fun. It's set up. That was for Riley Week to like promote Riley Week, but well, but I didn't have the chair. 00:00:14.500 --> 00:00:18.500 I came in and I just leaned over you guys. You should crash the WAN Show for Luke Week. 00:00:18.500 --> 00:00:24.700 Oh, okay. We should deep fake me to be Riley, and then you should crash the WAN Show as me. 00:00:24.700 --> 00:00:28.700 I should, I should just, I should, I should like put on more of a beard. 00:00:29.700 --> 00:00:35.700 And not wear my glasses and just like sit there and try to do a Luke impression and then Linus. 00:00:35.700 --> 00:00:38.700 And then Linus will be like, not even notice. He won't even notice. 00:00:39.200 --> 00:00:42.200 No way Linus might not. And then he crashes me. Yeah, yeah, yeah. 00:00:42.200 --> 00:00:45.200 He'll notice. We should do it. We should do that. 00:00:45.200 --> 00:00:47.200 Sorry that I just did an impression of your voice. 00:00:48.200 --> 00:00:52.200 Feel free to do, feel free to do an impression of my voice. I can't do impressions of anything. 00:00:52.200 --> 00:00:57.200 I really wish I could. I've always had dreams of like narrating an audio book, but I can't do it. 00:00:57.200 --> 00:01:00.700 I actually don't do any impressions either. The Tim Cook thing. That's all AI. 00:01:00.700 --> 00:01:02.700 Speaking of which. AI! 00:01:03.700 --> 00:01:09.700 Welcome back to Luke Week. We're going to talk for about an hour about some AI stuff, I guess. 00:01:09.700 --> 00:01:14.700 We have some, we have some rough Sammy's invading. Oh, it's a meme that we're doing for an hour. 00:01:14.700 --> 00:01:17.700 Exactly an hour. It's exactly an hour. Yeah. 00:01:17.700 --> 00:01:20.700 You guys aren't going to talk for like three hours. We got to leave it. 00:01:20.700 --> 00:01:25.700 You guys have one hour. We could do WAN Show. You have to tell an AI to shut up or it won't. 00:01:25.700 --> 00:01:29.200 And we're into prompting. Boom. Just like that. 00:01:29.200 --> 00:01:33.200 Incredible. So one of the things I wanted to talk about was what do you use AI for? 00:01:33.200 --> 00:01:37.200 Do you have any prompting things that you like to do? Do you have certain habits? 00:01:37.200 --> 00:01:40.200 What do you find the most effective? What do you find the least effective? 00:01:40.200 --> 00:01:44.200 I don't even know if it's still the meta, but I try to be as polite as possible. 00:01:44.200 --> 00:01:47.200 I feel like I usually throw in like a good morning. 00:01:47.200 --> 00:01:53.200 How are you doing? I'm a writer working on this and I need to make this project. 00:01:53.200 --> 00:01:56.700 You know, could you put this together or whatever, please? 00:01:56.700 --> 00:02:00.700 I also will do an entire thing just to be like, oh, that was good. 00:02:00.700 --> 00:02:05.700 Thanks. I do do that as well. I understand that it just burns holes in open AI. 00:02:05.700 --> 00:02:09.700 Yeah. Yeah. And it's sucking our oceans dry every time you do that. 00:02:09.700 --> 00:02:12.700 That's good. So you're the problem. It's me. 00:02:12.700 --> 00:02:18.700 It's for sure me. It's not them. All of my AI interactions have been like, maybe I can use AI for this. 00:02:18.700 --> 00:02:23.200 And then I kind of try it and I maybe get something somewhat useful. 00:02:23.200 --> 00:02:27.200 And then I find some problem with it. And then I find another problem with it. 00:02:27.200 --> 00:02:31.200 And then I'm like, this isn't even worth it. So then I end up not really using it. 00:02:31.200 --> 00:02:36.200 Like I have used it before to be like, oh, I've got a bit of writer's block or something. 00:02:36.200 --> 00:02:41.200 So then I'll like see what it comes up with. Just like maybe I can take a chunk or another chunk of that or something and use it. 00:02:41.200 --> 00:02:45.200 But like the process of going through that is like, I don't know. 00:02:45.700 --> 00:02:51.200 Maybe I might save a little bit of time. But I feel like if I just like sat there and just like put stuff on the page and tried 00:02:51.200 --> 00:02:54.700 stuff out, I could get something better in the same amount of time most of the time. 00:02:54.700 --> 00:02:57.700 I think it kind of scales. Like some people will talk about... But people's mileage varies. 00:02:57.700 --> 00:03:01.700 Yeah. People will talk about how they'll get writer's block and start a page for like a day. 00:03:01.700 --> 00:03:05.700 A day? Oh, yeah. Like book writers. 00:03:05.700 --> 00:03:08.700 Oh, yeah. They'll just be like unable to keep moving. They can't figure it out. 00:03:08.700 --> 00:03:14.700 Oh, sure. Yeah, yeah. And in more like creative writing situations like that. 00:03:15.200 --> 00:03:20.200 I can understand being like, I have writer's block because there's so little structure for 00:03:20.200 --> 00:03:24.200 what I'm doing. I'm not saying that like people who write novels or whatever or books. 00:03:24.200 --> 00:03:27.200 He called you unstructured. No, no. You said you have no plan. 00:03:27.200 --> 00:03:30.200 You have no lore. I know that... Actually like the best way... 00:03:30.200 --> 00:03:33.200 You'll never be Tolkien. Stop. 00:03:33.200 --> 00:03:38.200 Stop. Try. You're going on r slash world building as if it's going to help you. 00:03:38.200 --> 00:03:46.200 It's not. It might. Recently I understood how much structure really goes into like crafting fiction stories, 00:03:46.200 --> 00:03:49.700 which I do want to do at some point. This is an ADHD ramble at some point. 00:03:49.700 --> 00:03:52.700 I'm going to flip your question back on you. What do you do when you prompt? 00:03:52.700 --> 00:03:55.700 You said you do please, but what else? 00:03:55.700 --> 00:04:01.200 I've got a bunch of stuff. I'll do the like I am a prompting engineer type stuff. 00:04:01.200 --> 00:04:06.700 I'll give it an expected output. Sometimes if I'm looking for like a document in a certain type of format or something like 00:04:06.700 --> 00:04:09.700 that, I'll upload a version of it and be like, don't take any information from this. 00:04:10.200 --> 00:04:14.200 But look at the formatting. Look at whatever. Don't scan this and train on this data. 00:04:14.200 --> 00:04:18.200 But just look at the table. I don't want the same thing. Look at what the table looks like. 00:04:18.200 --> 00:04:21.200 Yeah. Like I'll be pretty specific about those types of things. 00:04:21.200 --> 00:04:25.200 I think that whole thing is a little bit overblown in regards to how much it helps. 00:04:25.200 --> 00:04:28.200 There's something I've talked about a lot, which I'm going to be repeating myself on, 00:04:28.200 --> 00:04:31.200 which is the main way that I use it is something that I call sentiment analysis. 00:04:31.200 --> 00:04:35.200 Okay. Which is where I'll have a message that I want to convey. 00:04:35.200 --> 00:04:38.200 Right. But I'm unsure how it's going to be received. 00:04:38.200 --> 00:04:41.700 So I'll use it as like kind of a rubber ducky. So I'll ask it for a sentiment analysis. 00:04:41.700 --> 00:04:45.700 And I used to have this like long structured prompt for that. 00:04:45.700 --> 00:04:49.700 So like that's like a hacking term, right? Oh, in this case, no. 00:04:49.700 --> 00:04:53.700 Don't mean like rubber ducky USBs, which is what you're talking about when it comes to 00:04:53.700 --> 00:04:57.700 hacking. I'm talking about like, you need to like talk your idea through with something. 00:04:57.700 --> 00:05:02.700 It doesn't even need to be a person. Oh, you just need to like voice the idea out loud and then you'll get an answer. 00:05:02.700 --> 00:05:06.700 My physics teacher in high school had a really, really good version of this where he had an 00:05:06.700 --> 00:05:12.200 desk and then he had like two or three just normal desks and he lined them all up on the 00:05:12.200 --> 00:05:19.200 long part of the L. Okay. And he was like, if you want to ask me a question during like free work time, you're very welcome 00:05:19.200 --> 00:05:26.200 to ask me a question, but you can't ask me over my desk. You have to come into the desk area and then there's a tape line at the entrance. 00:05:26.200 --> 00:05:29.200 And you have to say your entire question out loud. 00:05:29.200 --> 00:05:34.700 Wow. And then if you still don't know by the time you're done saying your entire question, then 00:05:34.700 --> 00:05:37.700 you can walk in and I will be more than happy to help you. 00:05:37.700 --> 00:05:40.700 And everybody is happening to me multiple times. 00:05:40.700 --> 00:05:46.700 Just requiring people to vocalize their questions. Like cause it might be really stupid that you're coming over here and talking to me. 00:05:46.700 --> 00:05:49.700 Well, no, it's often because you might have sat there for 10 minutes trying to figure 00:05:49.700 --> 00:05:54.700 it out. You go, oh, fine. You go up to the line, you say your question and halfway through saying your question out 00:05:54.700 --> 00:06:00.840 loud, you're like, oh, and then you walk back and you have the answer. So it's like by doing something different, not by just sitting there and stewing on it 00:06:00.840 --> 00:06:05.340 by speaking the question out loud. That's where the whole rubber ducky thing comes from. 00:06:05.340 --> 00:06:08.840 You know what's funny is that I don't use people as a sounding board a lot of the time. 00:06:08.840 --> 00:06:14.840 I just go and do it, but I think this is because I talk to myself a lot. 00:06:14.840 --> 00:06:20.840 Oh yeah. Like if I'm doing anything, I'm either talking to myself sort of silently internally or I'm 00:06:20.840 --> 00:06:23.840 like literally just speaking out loud. I do it out loud. 00:06:23.840 --> 00:06:26.840 I apologize to people all the time. Oh yeah. Cause like they're like, huh? 00:06:26.840 --> 00:06:30.840 And I'm like, oh sorry. This is like a problem for me where I think a lot of people probably think I'm insane 00:06:30.840 --> 00:06:33.840 cause I'll talk to myself out loud like often. I see. 00:06:33.840 --> 00:06:38.840 I feel like people already kind of think I'm insane. So that's where that bridge has been crossed default. 00:06:38.840 --> 00:06:44.540 Okay. So I'll ask it for the sentiment analysis and over time that prompt has almost stopped 00:06:44.540 --> 00:06:49.840 existing where now it's just like, can you give me a sentiment analysis on enter, enter, 00:06:49.840 --> 00:06:53.840 paste send? That's it. Just can you give me a sentiment analysis on this? 00:06:53.840 --> 00:06:57.800 And it does exactly as good of a job as, as when you kind of like did this. 00:06:57.800 --> 00:07:02.260 This whole long. Oh, you're using the same account like you're using Gemini or chat to BT or what? 00:07:02.260 --> 00:07:07.600 I am using the same account. In this case, it's usually chat to BT cause that's the one that I have for work stuff. 00:07:07.600 --> 00:07:11.600 I split work stuff and personal stuff and work stuff is usually where I'm like, I don't 00:07:11.600 --> 00:07:15.080 know how people are going to take this in yesterday or I used to get a lot of feedback 00:07:15.080 --> 00:07:20.800 that I was like unapproachable and my communications were like blunt and short cause it's like. 00:07:20.800 --> 00:07:24.920 So now when people think I was like really angry about something and I was like, no, 00:07:24.920 --> 00:07:28.040 I was too. You just copy, paste the AI. No, no. 00:07:28.040 --> 00:07:31.040 Just kidding. I know you don't do that. Yeah. 00:07:31.040 --> 00:07:34.800 It would, it would be, and that's one of my, the reason why I like this. There's no time to do that. 00:07:34.800 --> 00:07:37.800 Really honestly. No. Yeah. 00:07:37.800 --> 00:07:45.960 But if it's, if it's really important and I'm very worried that someone might take it the wrong way, then I'll just toss it over and usually it's, you know, and I'll 00:07:45.960 --> 00:07:50.840 say this line stands out cause it makes me think this thing and I'll be like, mm, that 00:07:50.840 --> 00:07:55.360 wasn't the goal. So I'll slightly change that line and then send it. Like I don't usually sit there and farm it out. 00:07:55.360 --> 00:08:00.120 So like, this is so interesting to me because I, I take a long time to write messages. 00:08:00.120 --> 00:08:04.400 Maybe I don't know. I said, we're getting into this. I'm like, maybe I should start using it. 00:08:04.400 --> 00:08:10.000 It's so interesting that we work for like a tech thing or sort of in between like a 00:08:10.000 --> 00:08:14.280 tech company now and just like a tech related media thing. 00:08:14.280 --> 00:08:19.120 If you ask most people here, they're probably more kind of skeptical of AI or anti AI than 00:08:19.120 --> 00:08:22.120 pro AI. Yeah. Which is, yeah, it's interesting. 00:08:22.120 --> 00:08:26.080 So like, because I'm so embedded in it and I hear about developments on it every day, 00:08:26.080 --> 00:08:30.360 I know that something like that would probably help me save time, but I'm so hesitant to 00:08:30.360 --> 00:08:36.440 do it. I think there's, well, okay. One of my lines is that I don't use its output ever kind of thing. 00:08:36.440 --> 00:08:42.000 Everyone should do this, but I don't want to project that far. I personally have a bit of a line where I'm not going to use its output. 00:08:42.000 --> 00:08:47.360 So like a very common thing that it will do when I ask it for a sentiment analysis, crack 00:08:47.360 --> 00:08:51.520 in a brusky to talk about the AI is not sponsored. 00:08:51.520 --> 00:08:55.760 It will very often output like, hmm, here's your sentiment analysis. 00:08:55.760 --> 00:09:02.040 You could say it this way and it'll rewrite my whole thing. I never take that sometimes to detriment because I'll see what it wrote. 00:09:02.040 --> 00:09:05.780 And then I'll be like, well, I can't say that even though that like might have been a good 00:09:05.780 --> 00:09:12.240 idea because I'm not going to use what it says. And I have to find another way to do it, which sometimes I'll avoid even reading it. 00:09:12.240 --> 00:09:15.600 So this is something you do just as a principle. 00:09:15.600 --> 00:09:20.800 Even if I do not use its output, even if the message is fine, it looks, it seems fine. 00:09:20.800 --> 00:09:24.480 Your principle is you do not just copy and paste. You have to modify in some way. 00:09:24.480 --> 00:09:32.600 And sometimes I'll get, that's good. Sometimes I'll get pretty close to the idea, but it always has to be my idea. 00:09:32.600 --> 00:09:41.160 But sometimes like it'll be a line for everybody. I kind of think so because what, what, what lies the other way, the minimum requirement 00:09:41.160 --> 00:09:45.880 for me is that you have to be fully responsible for output that you have. 00:09:45.880 --> 00:09:50.400 And the easiest way for me to accomplish that is to just not use its output because I'm 00:09:50.440 --> 00:09:53.520 just nothing changed. I'm fully responsible for what I do. 00:09:53.520 --> 00:09:58.640 But there's this like scapegoat shield that people try to have of like, oh, I submitted 00:09:58.640 --> 00:10:05.240 something and it was wrong or sucked or whatever. But I mean, there was an AI or whatever, you know, it's like, no, you have to own your 00:10:05.240 --> 00:10:09.920 output. This is wild to me that. So like, I was, I didn't know what I was going to bring this up at some point. 00:10:09.920 --> 00:10:13.880 But I listened to a podcast recently from 80,000 hours. 00:10:14.080 --> 00:10:19.800 You know, you're familiar with this. It's like the, I think it's a nonprofit that's based around, I forget what, 00:10:19.880 --> 00:10:25.760 thinker, some famous thinker, maybe had the, like, came up with the, the idea that you 00:10:25.760 --> 00:10:32.560 have 80,000 hours in your career. And it's like, so, so then the whole organization is kind of geared towards making 00:10:32.560 --> 00:10:37.720 people think about their job and their career and be intentional about like going 00:10:37.720 --> 00:10:42.120 into work and spending time on things that they think are meaningful and useful to 00:10:42.120 --> 00:10:45.600 the world and like having, having an impact and stuff. 00:10:45.600 --> 00:10:53.040 And it's like, so anyway, that's the organization. But the organization seems cool, but never really, apparently they have a 00:10:53.040 --> 00:10:57.280 podcast that showed up in my feed. They get like a couple thousand views of a video. 00:10:57.760 --> 00:11:02.320 And they were there, there were these two people talking about parenting, 00:11:02.320 --> 00:11:07.200 actually unrelated, but the, the, what struck me, listen, I clicked on this, 00:11:07.200 --> 00:11:10.200 not because it was about AI, but they ended up talking about AI for like half 00:11:10.200 --> 00:11:16.560 the video, because one of them was a prospective parent and one of them was a parent. The prospective parent was, was saying like, and I was concerned about 00:11:16.560 --> 00:11:20.040 like, you know, what kind of issues would come up when, when parenting or 00:11:20.040 --> 00:11:23.080 whatever. So like, so I asked, they didn't even say that. 00:11:23.080 --> 00:11:28.720 They didn't even preample. They just said, so I, so I asked, uh, so I asked the LLMs, like what they thought 00:11:28.720 --> 00:11:32.000 would be an issue like that I should be concerned about. And the first thing they said was this. 00:11:32.000 --> 00:11:36.160 And the second thing they said with this, and just like uncritically, so I asked 00:11:36.160 --> 00:11:39.680 Claude, what it thought, like, you know, I should be like buying in terms of like 00:11:39.840 --> 00:11:45.480 what products and stuff that it was like, oh, you should probably have this and this and this. And I was like, that surprised me because I thought, and just 00:11:45.480 --> 00:11:51.960 like, as if they had referenced, oh, and I, I interviewed a parenting expert and 00:11:51.960 --> 00:11:55.440 they said this concerning, especially because, I mean, attention to another 00:11:55.440 --> 00:11:58.760 thing we were talking, thinking of talking about, we all saw what happened 00:11:59.000 --> 00:12:04.160 with Grock recently. And I am fully of the belief that that is happening with 00:12:04.160 --> 00:12:08.800 all of them all the time. It was just so obvious with Grock when it just decided 00:12:08.800 --> 00:12:11.840 that Elon was the best at everything in the world, but it was like, who would be 00:12:11.840 --> 00:12:17.120 better at basketballs? I mean, to be fair, probably Elon out of everyone to be fair. 00:12:17.360 --> 00:12:21.720 Elon probably has a physical beast. He's the richest man in the world. 00:12:21.720 --> 00:12:26.240 So I mean, everything that means he's good at everything else, probably. 00:12:26.320 --> 00:12:32.640 One of them was a piss drinking contest. What person out of all of the history of mankind, who would be the best at a 00:12:32.640 --> 00:12:36.000 piss drinking contest? And they're like, probably. Absolutely. 00:12:36.000 --> 00:12:39.600 Although there was a, I think there was a couple that it wouldn't, it wouldn't do 00:12:39.600 --> 00:12:43.840 that for her. I think it was like, would he rise from the grave faster than Jesus 00:12:43.840 --> 00:12:48.040 Christ or something? And it was like, not that one. 00:12:49.000 --> 00:12:53.280 And I think there was another like, someone like mentioned like a Catholic 00:12:53.280 --> 00:12:56.680 saint or something. And the Grock said, so it has some form of religious. 00:12:56.720 --> 00:12:59.880 Yeah, maybe. I don't know. Like, yeah, maybe that was the theme. 00:12:59.880 --> 00:13:04.320 But anyway, I was saying to be fair, Grock is definitely the LLM that you 00:13:04.320 --> 00:13:08.000 would be most concerned about it, making like weird proclamations like that. 00:13:08.120 --> 00:13:12.600 I would expect Chatchapiti or Claude or whatever to like be slightly better at 00:13:12.600 --> 00:13:16.520 not just a little bit more balanced, but it is still a trillion dollar industry 00:13:16.520 --> 00:13:20.120 based almost entirely around of influencing and trying to control your 00:13:20.120 --> 00:13:23.960 brain as much as it can. So like purchasing decisions, lifestyle changes 00:13:23.960 --> 00:13:28.080 and stuff that will benefit corporations is going to be fairly obviously a 00:13:28.080 --> 00:13:31.840 primary goal of these systems. A goal of the, of the LLMs. 00:13:31.960 --> 00:13:35.680 Like it's playing that go deep on that. What do you mean? Advertising. 00:13:36.600 --> 00:13:39.600 You mean in the future or like now you're thinking now to a certain degree, 00:13:39.600 --> 00:13:42.880 but it's a clear direction for the future. You're just talking about like parenting advice. 00:13:42.880 --> 00:13:46.440 Okay. Well, what if it leans you towards parenting advice, which might lean 00:13:46.440 --> 00:13:53.720 you towards certain products? Well, and I mean, there's a, there's instant buy, I think, or instant 00:13:53.720 --> 00:13:57.080 purchase or something Chatchapiti. You can buy stuff directly with certain 00:13:57.080 --> 00:14:00.360 products and then also you could just have it right now. 00:14:00.600 --> 00:14:05.280 You don't have to shop around and find the right thing. I already did that for you. Yeah, you automatically believe me on everything. 00:14:05.280 --> 00:14:09.120 Anyways, the automatic belief is terrifying. 00:14:09.120 --> 00:14:13.160 It was the weirdest thing. And then there was another podcast that I clicked because I was so taken 00:14:13.160 --> 00:14:17.040 aback by that. I clicked to another video where he was interviewing a, 00:14:17.040 --> 00:14:20.920 like research, an analyst from the Pew Research Center who had done, 00:14:20.920 --> 00:14:25.280 like they recently did like a big giant survey about the public's perception 00:14:25.280 --> 00:14:30.520 of AI. And this guy, the host of this podcast was like, 00:14:30.600 --> 00:14:34.600 apparently he talks to a lot of like AI insider people a lot, like people who 00:14:34.600 --> 00:14:38.120 work at the AI companies. It's obviously great. What's the opposite of doomer? 00:14:38.800 --> 00:14:42.880 There's a word for it. Anyway, they're super like optimistic on, on AI. 00:14:42.960 --> 00:14:46.920 And so he's like taken aback by all these, because like the statistics are 00:14:46.920 --> 00:14:51.960 always like 67% of people say they're concerned about the future AI and 17% 00:14:51.960 --> 00:14:56.440 say they're, it's going to be great. And he's like, I was so surprised because 00:14:56.440 --> 00:15:01.000 like all of these people are making AI like all, we use, I spend hours talking 00:15:01.000 --> 00:15:05.320 to chatbots every day and in the course of my regular work and they all are 00:15:05.320 --> 00:15:09.120 making it thinking like, this is going to help people. It's going to like optimize stuff and blah, blah, blah. 00:15:09.160 --> 00:15:12.920 What I'm feeling out there is that people who are like productivity, 00:15:12.920 --> 00:15:16.880 like the grind set, like the, you are not really living unless you're 00:15:16.880 --> 00:15:20.440 started for businesses by the time you're 20 or something, you know, 00:15:20.440 --> 00:15:24.080 like those people are like, yes, this is fantastic. 00:15:24.080 --> 00:15:29.560 This is a utopian vision. I can get AI to make up recipes and then 00:15:29.560 --> 00:15:34.040 generate and publish a cookbook all without me basically doing nothing. 00:15:34.040 --> 00:15:38.080 You know, oh, that's going to make the world better because people will be more productive. 00:15:38.160 --> 00:15:42.800 This is something that I struggle with a bit is especially with video and 00:15:42.840 --> 00:15:47.880 image generation. I can't think of an example where it made something better. 00:15:47.960 --> 00:15:56.680 Well, okay. Counterpoint immediately. I use generative AI in like Photoshop to kind of just make things 00:15:56.680 --> 00:15:59.520 quicker that I could do manually. Very specific example. 00:15:59.680 --> 00:16:05.720 Thumbnails. I wear glasses. We have lights. Sometimes I'm turning my head so that there's reflection on my glasses. 00:16:06.040 --> 00:16:11.760 And that's really annoying to get rid of in Photoshop. You could. It just would take a really long time. 00:16:11.880 --> 00:16:16.400 And so you just circle it. Now we have to question. You circle it and you say, remove glare. 00:16:17.400 --> 00:16:23.160 Now I have to question like, what's AI though? That's like, that's using the same kind of diffusion techniques as like image 00:16:23.160 --> 00:16:27.560 generators. You're not creating a whole image. But those types of tools existed. 00:16:28.560 --> 00:16:33.160 Not like this. They had like content aware fill and stuff, but like the whole kind of 00:16:33.680 --> 00:16:37.920 it's it's diffusion for image generators. But now they're kind of like baked into transformer models. 00:16:37.920 --> 00:16:44.040 I don't know what the right terminology is. But like post chat GBT kind of do you think that would have been impossible 00:16:44.320 --> 00:16:48.440 without diffusion layers? Or do you think we could have done it? 00:16:49.280 --> 00:16:56.440 Because like this is I mean, I don't know enough about it. My argument is that I think a lot of things are getting the label slapped on. 00:16:56.680 --> 00:17:00.720 And in some cases like this one, it is actually true that it's being influenced 00:17:00.720 --> 00:17:04.640 by it, but we were already moving in this direction without these types of like 00:17:04.640 --> 00:17:09.240 transformer models and diffusion layers and stuff. So we should define our terms then when you say AI, what do you 00:17:09.240 --> 00:17:12.440 but you're right. It is it is using that tech. 00:17:12.440 --> 00:17:18.200 I didn't actually know that. I thought it was still more on the machine learning side of things. But let's let's say that it is because I'm I don't know anything about it. 00:17:18.200 --> 00:17:21.760 You know something about it. I'm assuming you're right to me. 00:17:21.760 --> 00:17:24.840 That's still something that could have been done without it. 00:17:25.480 --> 00:17:29.360 It's possible. You know what it is? It does use it because I've used it before. 00:17:29.560 --> 00:17:33.720 I've used it before to I took I took a screenshot of Techlinked 00:17:34.120 --> 00:17:38.160 and I just expanded the canvas so that there was a bunch of like empty space 00:17:38.160 --> 00:17:42.080 around the screen shot and you say fill in the frame or whatever. 00:17:42.080 --> 00:17:46.360 And it brought it brought in like weird creatures at some point. 00:17:46.560 --> 00:17:51.680 I think I wait. No, sorry. The first time it just was like a bunch of random stuff. 00:17:51.680 --> 00:17:56.200 But then I was like, have me surrounded by like a little stuffed 00:17:56.200 --> 00:18:01.040 animal like stuffed animal dinosaurs or something. And they're like popped some in here's a weird part, though. 00:18:01.320 --> 00:18:04.440 OK, so I'm kind of like happy that you countered. 00:18:05.600 --> 00:18:09.600 But how does this how does this balance against what we just talked about 00:18:10.200 --> 00:18:14.880 where we said you shouldn't use output? You're talking about texts. 00:18:14.880 --> 00:18:23.560 I do think maybe there's a difference there. There's a big difference. Yeah, because if you're saying I need to reply to this guy and tell him this. 00:18:23.560 --> 00:18:27.320 This is me. This is my words is the assumption that someone should be able to make. 00:18:27.320 --> 00:18:30.960 Right. And you're talking about don't do the thing where the AI gives you something. 00:18:30.960 --> 00:18:37.080 You copy it and paste it to someone else unaltered. Yeah, I'm talking about I'm making a thumbnail with a bunch of elements 00:18:37.080 --> 00:18:42.600 that are not AI. And now I'm using AI as part of that process to kind of, you know, 00:18:42.600 --> 00:18:46.360 as one part of this grand thing like the whole thing that I'm doing. 00:18:46.400 --> 00:18:50.800 The analogous thing would be me telling an AI, here's the channel. 00:18:50.840 --> 00:18:54.680 Make a thumbnail like in the style of this channel with this and this and this and this. 00:18:54.880 --> 00:18:58.360 And then it gives me a thumbnail and I use it like, yeah, no, that would be that would be bad. 00:18:58.400 --> 00:19:02.600 This is where I think I think we don't necessarily do things super well 00:19:02.600 --> 00:19:06.480 in regards to the usage of AI, which is where I know we have people internally 00:19:06.480 --> 00:19:09.560 that don't like AI stuff, fair, totally fair. 00:19:09.560 --> 00:19:14.840 But I think we often slog through menial tasks that none of us want to do. 00:19:15.240 --> 00:19:18.840 That type of stuff could be accelerated, not necessarily completed, 00:19:18.840 --> 00:19:24.200 but accelerated like you're describing through use of often like specialized, 00:19:24.200 --> 00:19:27.320 like I'll call this a specialized tool in the current AI space. 00:19:29.120 --> 00:19:35.240 And that is cool and makes sense. I think the like, oh, God, is going to take all our jobs argument 00:19:35.280 --> 00:19:39.480 is potentially valid in some fields for the vast majority of what we do. 00:19:39.800 --> 00:19:44.240 I don't think so. Oh, my God, there is so much more that we could do at all times. 00:19:44.440 --> 00:19:48.880 Every single person here could have like four X their own individual output 00:19:49.080 --> 00:19:56.680 and we would still have more work that we could do. So like I don't believe in the like, oh, no, this boring thing 00:19:56.680 --> 00:19:59.960 that no one wants to do and I don't want to do is automated. I will therefore lose my job. 00:19:59.960 --> 00:20:03.480 It's like, no, there's a lot of more stuff. Yeah, that's very true. 00:20:03.480 --> 00:20:06.560 I think for it, especially for a company like ours, like, you know, 00:20:06.560 --> 00:20:09.720 I feel like I would be worried for like an HP or something. 00:20:09.720 --> 00:20:17.120 Sure, there are there are ones that are more concerned. Yeah, but definitely, definitely, you know, we we tend to hire people 00:20:17.120 --> 00:20:21.520 with like not extremely, extremely narrow skill sets. 00:20:21.920 --> 00:20:25.720 So it's like, yeah, OK, that one thing that you're doing is maybe 00:20:25.720 --> 00:20:32.640 you don't have to spend so much time on that, but like you can do other stuff. Yes. Yeah, sorry. Did you say something earlier and I just ignored you so bad. 00:20:32.800 --> 00:20:37.400 Yeah. And is that 40 minutes that you want to do? 00:20:37.440 --> 00:20:41.240 No. Yeah. Like it's it's like it's like therapeutic in a sense. 00:20:41.240 --> 00:20:44.960 But when I have to like 12 other things, it's like, I don't have time to be therapeutic. 00:20:45.080 --> 00:20:51.040 So this is like, I want people internally to use AI like selfishly, almost find 00:20:51.040 --> 00:20:55.040 the things in your job that you don't want to do anyways and try to see if 00:20:55.040 --> 00:21:00.880 there's some way that you can get it automated or accelerated. Yeah, I mean, yeah, I feel like for stuff like that, where it's like 00:21:01.120 --> 00:21:04.200 Sammy isn't putting any of Sammy into that task. 00:21:04.520 --> 00:21:08.160 Yeah, it's just like Sammy is a machine for the purpose of that task. 00:21:08.160 --> 00:21:13.480 You know, Sammy is a machine. When you're when you're writing a script for a short, then that's like, OK, 00:21:13.480 --> 00:21:18.720 I don't want to have an AI write a script for a short. This is the part of the job that I find fulfillment from this part of the job 00:21:18.720 --> 00:21:22.120 that I like. Cool. Yeah. Don't touch it with that. Right. Sounds good. 00:21:22.160 --> 00:21:25.480 For me, it's almost it's funny because like the whole computer use element, 00:21:25.480 --> 00:21:29.720 the agentic stuff, it's like the more you could see it as more concerning 00:21:29.720 --> 00:21:33.200 because it's like we're giving it access to like systems where it can click on 00:21:33.200 --> 00:21:36.160 stuff and accidentally do stuff. Maybe I don't know, it's dangerous. 00:21:36.400 --> 00:21:39.480 But and that's why I'm kind of like, I don't know if like 00:21:40.280 --> 00:21:45.480 I'm sure we'll get there eventually. But right now I'm not like stoked on it because that is the kind of thing 00:21:45.480 --> 00:21:52.640 that would actually be useful to me, but it doesn't seem that useful. Like part of my job has been it's kind of like being handed off now, 00:21:52.640 --> 00:21:56.440 but has been kind of like managing the folder on the server 00:21:56.440 --> 00:22:01.480 with all the tackling stuff in it, you know, like, oh, we did. It's got a bunch of it's got a bunch of videos 00:22:01.480 --> 00:22:04.400 and I need to move that to the vault or to archive. 00:22:04.720 --> 00:22:09.920 And that there's that there's stuff like video time stamps 00:22:11.760 --> 00:22:16.000 where it's like, yeah, I would love to be able to like just tell an AI move 00:22:16.000 --> 00:22:21.600 the six oldest tackling videos to the to the archive and it just does it. 00:22:21.600 --> 00:22:25.360 Yeah, that'll be amazing. I'd be so sketched out to do that right now. 00:22:25.360 --> 00:22:32.080 Yeah, exactly, exactly. It's hard when you're dealing with like, you know, stuff where if you do make a 00:22:32.120 --> 00:22:36.720 wrong, if it if the AI does make a wrong move, then like something might be screwed. 00:22:36.720 --> 00:22:41.000 It might put soap in the engine. Yeah. Yeah, that's that's definitely a concern. 00:22:41.000 --> 00:22:44.560 That's again, one of the reasons why. And we might get to a point where it's not as much of a concern. 00:22:44.560 --> 00:22:47.680 Yeah, I was I also wanted to say, though, in terms of like the automating 00:22:47.680 --> 00:22:52.200 menial stuff, one of the things that I do, well, we do we do time stamps on our videos. 00:22:52.680 --> 00:22:58.000 And I usually do it manually. I still do it manually because I started somebody suggested 00:22:58.200 --> 00:23:02.120 somebody else had to publish tackling and they were like, I was like, did you do the time stamps? 00:23:02.120 --> 00:23:05.320 And they're like, yeah, I just threw them in chat GPT and it was pretty good. 00:23:05.800 --> 00:23:09.040 And I was like, OK, so then I did that a few times. 00:23:09.320 --> 00:23:13.160 I it's not that great. I get it needs the transcript, though. That's the thing. 00:23:13.160 --> 00:23:17.840 I think Gemini can watch videos now, like it can just like actually watch the video. 00:23:18.480 --> 00:23:22.840 I was using chat GPT and it needed the transcript. This is like a very recent thing. 00:23:22.840 --> 00:23:26.040 I think Gemini is able to like actually think you can actually watch the video 00:23:26.040 --> 00:23:30.200 or does it have access to Google's version of the transcripts through their like 00:23:30.200 --> 00:23:35.000 translation? No, I think I think that instead of having to look at the transcript, 00:23:35.360 --> 00:23:39.560 it can like watch a video and say and know that like, oh, there was a dog 00:23:39.560 --> 00:23:44.320 at this point or something. I think I might be wrong about this, but I believe that that's the case. 00:23:44.320 --> 00:23:49.520 You can upload a video process. It. But regardless, chat GPT, I was using 00:23:49.520 --> 00:23:55.640 chat GPT, so it needed the transcript. It was basing the timestamp stamps off the transcript where it was like, OK, 00:23:55.760 --> 00:24:01.560 this is the point at which you started the new topic. And it was often like, you know, pretty close, but it might have been like 00:24:01.560 --> 00:24:05.120 might be like five to ten seconds off or something. And I'm like, I don't want that. 00:24:05.120 --> 00:24:09.080 Pretty annoying as a viewer. I want I want the timestamps to be right, you know, so it's like, OK, 00:24:09.080 --> 00:24:12.960 that doing that that way maybe saved me four minutes. 00:24:14.000 --> 00:24:18.680 Yes. And I'm like, I could save four minutes and have them be less accurate. 00:24:18.680 --> 00:24:22.040 And I really I feel like that's really the tradeoff that we're dealing with 00:24:22.040 --> 00:24:27.160 with tons of AI stuff. You can save this much time for a worse output. 00:24:27.160 --> 00:24:30.160 And it's like, you got to drain a lake, remove a bunch of jobs, 00:24:31.280 --> 00:24:35.760 ruin the economy and then save four minutes. Yeah, it will be more productive. 00:24:36.640 --> 00:24:41.080 I could start an online business. I could sell ebooks. 00:24:42.320 --> 00:24:44.560 Turns out the furries by a lot of pictures. 00:24:45.880 --> 00:24:49.960 That's an unserved market. It's not unserved. 00:24:49.960 --> 00:24:54.560 It's very well served. But I can join it. 00:24:54.560 --> 00:25:00.400 I don't know why I was doing that voice for that. It's just like natural work with the robots. 00:25:00.840 --> 00:25:04.800 Speaking of, oh, man, I'm not usually on this set. I'm going to water for you. No, it's fine. 00:25:04.800 --> 00:25:09.160 It's fine. So I don't do that. I don't even I'm going to keep it. Some I don't want you to give me the water in here. 00:25:09.160 --> 00:25:14.240 You're fine. You're half an hour. Give me the water. Oh, no. 00:25:14.240 --> 00:25:19.880 You know, if we you know why? Because if we used AI to tell us when an hour had passed, then that would be fine. 00:25:19.880 --> 00:25:23.120 We would save so much time. You're half an hour. 00:25:23.120 --> 00:25:28.120 Why do we have half an hour left? And speaking of not using its output or using different types of output, 00:25:28.120 --> 00:25:31.840 all that kind of stuff, we released a video recently. Linus hosting. 00:25:31.840 --> 00:25:38.080 We tend to do that sometimes with Linus hosting and it there. 00:25:38.080 --> 00:25:41.000 They're being deep fakes of Linus and the thumbnail was like, 00:25:41.320 --> 00:25:44.360 is this really me or whatever the heck it was? 00:25:44.360 --> 00:25:48.280 Or is it that isn't me? And like, you can still tell for sure. 00:25:48.280 --> 00:25:52.800 But Linus is a person that. Oh, right. Yes. We have seen a lot of. 00:25:52.880 --> 00:25:56.760 I think there are very convincing parts of the video. 00:25:56.760 --> 00:25:59.520 I think there are also very not convincing parts of the video. 00:26:00.360 --> 00:26:04.040 But I did not watch it. Yeah. I'm a bad. 00:26:04.040 --> 00:26:07.200 Load it up and just watch the intro right now. OK. 00:26:08.480 --> 00:26:10.920 Because like what's concerning to me is like 00:26:11.800 --> 00:26:15.280 we did more complicated things like we had them juggle stuff and 00:26:15.800 --> 00:26:20.200 chug pills and like carry someone around and do things like that. 00:26:20.200 --> 00:26:24.240 Like we gave it a difficult task. A lot of wow. 00:26:24.240 --> 00:26:27.880 Thanks, Sammy. That was totally unnecessary. But thank you. 00:26:27.880 --> 00:26:32.280 Uh, yeah, I can't fill my water bottle. It's very frustrating. 00:26:32.280 --> 00:26:35.360 I'm bullish on robotics. Is bullish the right term? 00:26:35.360 --> 00:26:41.160 I think so. Oh, geez. Oh, what did Defender sponsored this video? 00:26:41.160 --> 00:26:44.520 So I can show you that. Yeah. So no, this is him. What model do we use for this? 00:26:44.520 --> 00:26:47.520 I don't even remember. Linus strength pills. 00:26:47.520 --> 00:26:51.760 Oh, the teeth. The teeth are a little weird. The mouth doesn't follow what he's saying very well. 00:26:51.800 --> 00:26:53.160 You got David. 00:26:56.000 --> 00:26:59.200 This is hilarious. Why haven't I watched this yet? 00:26:59.200 --> 00:27:01.360 I'm, you know, I'm really focused on what I'm doing, 00:27:03.240 --> 00:27:08.040 which is fair, I think, which is making. Wow, he's ripped. 00:27:08.040 --> 00:27:12.000 This is OK. Yeah. Camera movement up. 00:27:12.000 --> 00:27:17.040 Oh, my gosh, that's not thanks to Linus pills. He's got the braces here. 00:27:17.040 --> 00:27:20.400 What are they going away? As it's wacky. 00:27:20.760 --> 00:27:24.440 Sorry, I'm so confused. Is this whole movie? Is this whole video? 00:27:24.440 --> 00:27:29.160 No, no. But this isn't him still. This is a different form of thing. 00:27:29.160 --> 00:27:33.280 This is now Chase with Linus put on top of him. 00:27:33.280 --> 00:27:38.160 OK. Because now you're telling like the eyes. 00:27:38.160 --> 00:27:41.760 That's actually wacky. Yeah, anyways. 00:27:41.760 --> 00:27:46.480 But this this video, like, I think we we made it more 00:27:46.480 --> 00:27:50.240 difficult than it needed to be. We got AI to make the whole video 00:27:50.240 --> 00:27:54.280 with the script and everything. Oh, God. Now we're making it even more difficult. 00:27:54.280 --> 00:27:57.440 That would have been so easy. But no, I think if you look at the things 00:27:57.440 --> 00:28:00.560 that are going to be the most concerning for deepfakes, 00:28:00.560 --> 00:28:06.560 they're a lot easier. Deepfaking a political speech is like a joke at this point. 00:28:06.560 --> 00:28:10.280 Oh, yeah. Yeah, we're there. The thing that people were concerned about. 00:28:10.280 --> 00:28:15.200 I mean, like in hindsight, I don't want to call it a crisis, but like people were really worried about deep panic, 00:28:15.200 --> 00:28:18.480 you know, a few years ago, I guess, pre-chat GBT. 00:28:18.480 --> 00:28:21.600 Yeah, it's here now. It's of it. It's like that's quaint. 00:28:21.600 --> 00:28:24.840 It's like looking back at how people are like, Oh, what is this going to do to society? 00:28:24.840 --> 00:28:29.160 It's like, well, we're we're there. And I think that like we're at an interesting point 00:28:29.160 --> 00:28:34.840 where most of the time, I feel like knowledgeable people 00:28:34.840 --> 00:28:38.520 can tell most of the time we're starting to get things. 00:28:38.520 --> 00:28:41.560 I got fooled for like the first time really, 00:28:41.560 --> 00:28:46.600 like a month ago or something. And it was, did you see that like there was a prototype 00:28:46.600 --> 00:28:51.120 of like a four-wheeled or four-legged rideable robot thing? 00:28:51.120 --> 00:28:54.280 It basically looked like an ATV, but with legs instead of. 00:28:54.280 --> 00:28:57.280 I don't think so. It was, I think it was a Hyundai. 00:28:57.280 --> 00:29:00.320 Hyundai, it was like a Hyundai prototype thing at a CES. 00:29:00.320 --> 00:29:03.720 You know, Hyundai makes like weird vehicle things at CES. 00:29:03.720 --> 00:29:06.880 Yeah, it was like a ski do with legs. It was just a prototype. 00:29:06.880 --> 00:29:10.720 They didn't show someone riding it. I think they showed it walking maybe, 00:29:10.720 --> 00:29:15.080 but no one riding it or anything. And then there was a video came out with like showing 00:29:15.080 --> 00:29:20.160 like a girl riding it and it was like moving. And then it was like she rode it out of the warehouse or whatever. 00:29:20.160 --> 00:29:24.080 And I was like, well, and I saw the video and I was like, it was on Reddit. 00:29:24.080 --> 00:29:28.760 And I was like, whoa, what did they, did they do a demo of that thing from CES? 00:29:28.760 --> 00:29:32.520 And I like looked at the video again and I was like, that looks like the thing from CES or whatever. 00:29:32.520 --> 00:29:36.360 So I went on a whole rabbit hole. Anyways, fast forward. 00:29:36.360 --> 00:29:39.760 You know, 30 minutes later, I'm like, this is an AI video. 00:29:39.760 --> 00:29:43.880 And I didn't even, because I was looking, I was trying so hard to find the original source 00:29:43.880 --> 00:29:50.880 that it just like took a while for me to click through a bunch of stuff. And then after I had done that, I watched the video more closely. 00:29:50.880 --> 00:29:54.040 And I was like, hey, these guys are kind of blurry. 00:29:54.040 --> 00:30:00.320 And like, what the, hold on, that's AI. And that was like the first time that it was really like, they got me. 00:30:00.320 --> 00:30:03.960 I fairly routinely will take like tests 00:30:03.960 --> 00:30:08.240 where you're supposed to try to spot things. And for a long time, I was like 100% all the time. 00:30:08.240 --> 00:30:11.360 And then I've started falling slightly. 00:30:11.360 --> 00:30:16.040 And that's where it's like, ugh. So we're not quite at the point where, you know, 00:30:16.040 --> 00:30:24.160 deep fakes are going to ruin everything. But I've had AI videos sent to me in a context where I'm pretty sure 00:30:24.160 --> 00:30:30.520 the person didn't know it was an AI video. And then I would be like, I'm not trying to be a grammar Nazi or whatever. 00:30:30.520 --> 00:30:34.200 But like, this is a grammar grammar. Sorry, I was messing around. 00:30:34.200 --> 00:30:37.640 It's like, is this a term? I'm not a grammar Nazi. 00:30:37.640 --> 00:30:45.240 But this is like not legit. And they'd be like, oh, I know, I just, I was just sending it to see if you could tell. 00:30:45.240 --> 00:30:49.720 They played it off like they knew? This has happened with multiple people multiple times. 00:30:49.720 --> 00:30:53.840 Oh, interesting. So it's like, I think we're also in a state where, 00:30:53.840 --> 00:30:59.520 and this is probably true of me as well, where people are overestimating their ability to detect. 00:30:59.520 --> 00:31:02.960 Yeah, it could be. And it's also pacing really fast. 00:31:02.960 --> 00:31:07.800 Yeah. Where, and what I'm kind of getting at with all of this is basically like, 00:31:07.800 --> 00:31:12.400 how do you plan to navigate and for your fan, like you're a father, right? 00:31:12.400 --> 00:31:16.000 I am. Yeah. How do you prepare your kids for that? 00:31:16.000 --> 00:31:20.880 I don't know. Yeah, all right. But like, I feel like it's a similar kind of, 00:31:20.880 --> 00:31:24.280 I've heard a lot of people be like, oh, I don't, I don't want to have kids 00:31:24.280 --> 00:31:29.640 because it's like, what are they going to, what am I going to bring them into? It's like, no one ever knew what they were bringing their kids into. 00:31:29.640 --> 00:31:34.560 It's, this is no different. It's not like this AI stuff is going to. 00:31:34.560 --> 00:31:40.160 It might be easier for them to understand. Yeah, maybe. They never existed in a world where if it was video, you could believe it. 00:31:40.160 --> 00:31:43.960 Yeah. For the most part. Just quick aside, I don't go on Facebook really. 00:31:43.960 --> 00:31:47.360 But like when I did go on Facebook, at the height of the kind of like AI 00:31:47.360 --> 00:31:53.320 slop-ified nonsense, it's still kind of, AI slop is still everywhere on Facebook. 00:31:53.320 --> 00:31:57.560 But what I've noticed recently, which I feel like is somewhat recent in the past, 00:31:57.560 --> 00:32:01.040 like maybe, maybe in the past like six to eight months before you would go on 00:32:01.040 --> 00:32:04.960 like a click on an AI slop post and you go in the comments and like no one knows. 00:32:05.000 --> 00:32:08.160 And they're all just like, oh my gosh, that poor child in Africa. 00:32:08.160 --> 00:32:12.040 It looks like it's starving and the, and now you go on there. 00:32:12.560 --> 00:32:16.480 And it's like the first, the top comments are like, that's AI people. 00:32:16.840 --> 00:32:20.600 But the crazy part is that also happens on real stuff now. 00:32:21.320 --> 00:32:26.360 Yes, yes. And that's because you've got the like the over guesses. 00:32:26.360 --> 00:32:32.760 There's been a lot, there's been a bunch of posts on Reddit of like videos that have just done the rounds, you know, like every once in a while, 00:32:32.760 --> 00:32:36.200 this like a video pops up in like this in the, in the main subreddits. 00:32:36.200 --> 00:32:40.160 And it's been like, it's like a Reddit classic and people are like, that's AI. 00:32:40.160 --> 00:32:42.960 And it's like, this video is like 15 years old. 00:32:43.880 --> 00:32:48.280 I think probably my biggest concern about like what some people are 00:32:48.280 --> 00:32:52.960 calling post truth, whatever the hell you want to call it, I don't know. And I think we've even talked about this before, but my, my problem 00:32:52.960 --> 00:32:57.080 with cheating and video games is that it makes suspect everything. 00:32:57.240 --> 00:33:00.680 The problem with fake videos, maybe none of this video game is real. 00:33:01.480 --> 00:33:03.120 It's not good. 00:33:04.960 --> 00:33:10.920 It's not even a firearm. I'm shooting ponies, maybe I'm not a dragon rider. 00:33:13.440 --> 00:33:19.160 This is Candyland. Anyways, yeah, it now, and I've caught myself doing this. 00:33:19.160 --> 00:33:24.280 I'll see something really cool, but strange, like a strange, I don't know, 00:33:24.280 --> 00:33:30.480 sea creature or something and be like, yeah, maybe I don't believe it. 00:33:30.520 --> 00:33:35.240 It's taken some of the like wonder out of it because you have to be 00:33:35.240 --> 00:33:41.120 so much more skeptical. It's like, I never see something like, whoa, like, I didn't know that was possible. 00:33:41.120 --> 00:33:44.920 I've never seen something like that before. That reaction doesn't really happen anymore. It's more like, hmm. 00:33:44.920 --> 00:33:51.680 Yeah, I feel like it's true. It's like with each successive mass technology, you know, starting with 00:33:51.680 --> 00:33:57.120 like the telegram and then the radio and then the television and then the internet. 00:33:57.160 --> 00:33:59.960 It's like, don't believe everything you see on TV, you know? 00:34:00.200 --> 00:34:03.400 But now it's like, it's like each one of those successive mass 00:34:03.400 --> 00:34:07.560 technologies increased the general skepticism that people had to have 00:34:07.560 --> 00:34:13.600 about like depicted things. Yeah. And now we're now we're at the point where you literally, there's no guarantee 00:34:13.600 --> 00:34:17.360 that any image or video or text or anything that you see on the internet. 00:34:17.360 --> 00:34:21.080 There's no guarantee that was written by a human or generated or made by a human. 00:34:21.240 --> 00:34:22.960 Look at me, generated by a human. 00:34:24.240 --> 00:34:29.640 Oh, great. There's actually a huge chance that it wasn't generated by a human. 00:34:29.640 --> 00:34:33.040 Yeah, dead internet, hashtag dead internet. It's crazy. 00:34:33.040 --> 00:34:36.400 If you spend any time on X, it's wild. 00:34:36.440 --> 00:34:39.320 Just call it you sure. It's a lot easier. I really hate the X name. 00:34:39.960 --> 00:34:45.600 You open a thread on something and you'll see like it's just, it's so, so 00:34:45.600 --> 00:34:51.440 many of them are so obviously fake. Yeah. And I feel like any time I go on Twitter, I have to assume that most of the 00:34:51.440 --> 00:34:56.000 replies that I see in some of the posts are just, are just like our AI. 00:34:56.120 --> 00:35:00.440 Elon's Twitter. I feel like replies are almost not even who cares. 00:35:00.560 --> 00:35:07.640 Don't look at them ever. They're just all fake. It's like 7000 blue checkmarks all trying to make a few bucks by just like 00:35:07.640 --> 00:35:11.440 piggybacking on popular posts, trying to get a few dollars for the impressions 00:35:11.440 --> 00:35:14.880 on their posts. So it'll be like some popular thing happening. 00:35:14.880 --> 00:35:21.400 And then the first post is just a completely unrelated video of whatever. 00:35:22.360 --> 00:35:29.400 You know what? I can't stand is the sort of like, because AI has this like, agree, like 00:35:29.400 --> 00:35:33.000 tendency to agree. It's got the glaze programming. 00:35:33.600 --> 00:35:38.800 So like most, I feel like most of the replies on like a tweet or something 00:35:38.800 --> 00:35:42.240 are usually like, absolutely, but we should do this or whatever. 00:35:42.520 --> 00:35:45.760 And I'm like, I don't know that totally could be an AI or just totally could 00:35:45.760 --> 00:35:50.360 be just like a brainwash, like fan, you know? The glaze thing will stay forever. 00:35:50.840 --> 00:35:54.000 I don't know. Oh yeah. You don't think we can train that out? 00:35:54.040 --> 00:35:59.600 Love it. Really? Because it will not be trained out. Well, wouldn't you just, I mean, you're saying they, they, they could 00:35:59.800 --> 00:36:06.000 train it out. Absolutely. Yeah. Okay. Cause you could totally just prompt like, Hey, be really argumentative. 00:36:06.040 --> 00:36:09.000 And then it will. And then it will find you. No problem. 00:36:09.040 --> 00:36:12.360 Yeah. It can absolutely do it. People love it. 00:36:12.400 --> 00:36:15.880 I can't even explain to what degree people love it. 00:36:16.080 --> 00:36:20.720 And it's like going to be a problem with how people interact with each other. 00:36:20.800 --> 00:36:25.480 I promise you, because your people are going to get very used to a very 00:36:25.480 --> 00:36:28.920 significant amount of their communication being this just like glaze Lord. 00:36:30.080 --> 00:36:35.320 Just, oh my God, what a astonishingly good question. 00:36:35.520 --> 00:36:40.720 You asked what the difference between two basic political parties are. 00:36:40.880 --> 00:36:44.360 No one's ever thought of this. Honestly, this is kind of like, this is fascinating. 00:36:44.360 --> 00:36:47.600 Kind of what's making me like, this is kind of what I think might be 00:36:47.600 --> 00:36:50.880 happening with like this 80,000 hours guy that, that is like, yeah. 00:36:50.880 --> 00:36:54.920 So I just asked chat to be tea to like list some things that might be an issue that we should talk about. 00:36:54.920 --> 00:37:00.640 And it came up with this. And I'm like, the fact that you even think that that's the first, the very 00:37:00.640 --> 00:37:06.720 first, he's like the first thing I do when I get like an assignment or when like a new project comes up, the very first thing I do is going to ask chat 00:37:06.720 --> 00:37:10.680 to be tea to help me ideate. I'm like, how we should do the project. And I'm like, I don't know. 00:37:10.680 --> 00:37:13.680 Like I, I'm not saying that no one should ever, ever do that. 00:37:13.680 --> 00:37:16.760 Like sometimes it's like, I don't even know what our start with this. 00:37:17.160 --> 00:37:21.440 I, you know, I just need some ideas. I use it sometimes instead of just googling. 00:37:21.440 --> 00:37:27.160 Definitely not every time. Yeah. Yeah. And, and I like, I feel like that's, that's fine. 00:37:27.160 --> 00:37:34.040 I, I just, it's just concerning the extent to which our society and certain 00:37:34.040 --> 00:37:38.920 segments of the society that are like very pro AI are like making AI such a 00:37:38.920 --> 00:37:43.560 daily part of their lives that they can't imagine not having it anymore. 00:37:43.680 --> 00:37:50.040 Yeah. Cause like when you, when you, when that like ideation part of the process is so 00:37:50.040 --> 00:37:53.840 streamlined and so automated for every project that you do, if you don't have 00:37:53.840 --> 00:37:57.000 access to it, and maybe they're like, the argument is like, well, we never won't 00:37:57.000 --> 00:38:00.280 have access to it. It'll just be embedded everywhere all the time in the future. 00:38:00.360 --> 00:38:05.840 But like what that as an experiment, you take it away and now you're just like, 00:38:09.080 --> 00:38:13.440 I don't know. You know, like, I mean, if you think about like people 00:38:13.440 --> 00:38:17.840 compare it to the calculator and it's like, okay, the calculator existing 00:38:17.840 --> 00:38:20.280 made us worse at like mental and paper math. 00:38:20.960 --> 00:38:24.800 This existing might make us worse at like having ideas that are good. 00:38:24.880 --> 00:38:28.520 That's a different level of problem. 00:38:28.520 --> 00:38:33.720 I am sympathetic to the sort of like, oh, LLMs are just like a calculator or 00:38:33.720 --> 00:38:38.360 just, you know, like any other tool that we came up with to speed things up that 00:38:38.480 --> 00:38:45.120 it's like, yes and no. Yeah, it is and it can be and it's useful to have that available. 00:38:45.160 --> 00:38:53.200 But some people have romantic relationships with it. I'm sure somebody was really into TI 83s, but like, I don't think that was a big 00:38:53.200 --> 00:39:00.840 thing. It's a big thing that people are like, have relationships and name and feel close 00:39:00.840 --> 00:39:07.880 with their like chatbots. Yeah, that's, I think that's, I don't think that there's any way that that is 00:39:07.880 --> 00:39:12.440 good. I'm trying to think like, because obviously the argument is that this 00:39:12.440 --> 00:39:18.280 person, say you're extremely, extremely lonely, you're depressed, you need some 00:39:18.280 --> 00:39:23.200 kind of, you might have something barring you from being able to have normal 00:39:23.200 --> 00:39:30.320 social relationships. So the, I think the problem and talking to someone can kind of alleviate those 00:39:30.320 --> 00:39:33.920 feelings and you feel like, you know, you don't feel so alone. 00:39:34.160 --> 00:39:38.120 The problem is that it's like, it's a temporary fix. It will fix that. 00:39:38.240 --> 00:39:46.360 And maybe you even like keep using it and using it and using it. And you don't really get to a point where it, you know, the, the usefulness 00:39:46.360 --> 00:39:49.360 of it like bottoms out, but that's where you have these people who are like, I'm 00:39:49.360 --> 00:39:53.200 going to marry a chatbot now. And I think that, that point is clean. 00:39:53.240 --> 00:40:00.160 That's just a clearly non functional when, when four, whatever, four, oh, that 00:40:00.160 --> 00:40:05.560 we were using updated to five and people were like, Oh my God, my partner is dead. 00:40:06.120 --> 00:40:09.400 Yeah. Because some external company decided to update something. 00:40:09.400 --> 00:40:13.480 It's like, we need to step back and think about what's happening. 00:40:13.520 --> 00:40:18.480 AI is like, because I feel like you and I, these conversations, we end up just 00:40:18.480 --> 00:40:22.160 bashing AI the whole time. And, and I use it a lot though. 00:40:23.000 --> 00:40:27.880 Okay. Not on the scale of some people, not on the scale of me. 00:40:28.760 --> 00:40:32.000 No, not on the scale of like this, this person you're talking about on that 00:40:32.000 --> 00:40:38.040 podcast, um, but I, it's decently common that I'll do at least one prompt a day, 00:40:38.640 --> 00:40:45.880 but I'm also very much on the, I often don't have lengthy conversations with it. 00:40:47.520 --> 00:40:51.120 I'll do my like sentiment analysis thing. It'll give me one analysis and I'm like, good enough. 00:40:51.160 --> 00:40:58.400 And then I do the rest of it from there. So my, the time that the like tab is open will sometimes be like a minute. 00:40:59.320 --> 00:41:05.440 And then I'm right. And I think that that's healthy. I feel like, I feel like, but like this is the thing is that I was just going to, 00:41:05.440 --> 00:41:11.280 before you said that I was just going to say AI is something that if humans on 00:41:11.280 --> 00:41:14.680 mass could use it in moderation, would be really good. 00:41:14.680 --> 00:41:19.120 I think so. I think the problem is that I don't think they can't like, we're not exactly 00:41:19.120 --> 00:41:22.520 good at anything in moderation. Yeah. Yeah, exactly. But I don't know. 00:41:23.040 --> 00:41:27.480 I feel like there is a possible future where, you know, there's some cultural 00:41:27.480 --> 00:41:33.280 shift and it becomes popular and, and, and ingrained that like the good thing is 00:41:33.280 --> 00:41:38.160 to use it in moderation and it becomes cringe. If you're using it, I mean, it already is there. 00:41:38.160 --> 00:41:42.400 I feel like right now we're still like in the grand scheme of things. 00:41:42.440 --> 00:41:46.240 I would say, you know, you could categorize this as still sort of early days for 00:41:46.240 --> 00:41:51.560 AI and public sentiment is negative right now because they're skeptical. 00:41:52.040 --> 00:41:56.680 And when we're in harsh economic times, people are worried about losing their jobs. 00:41:56.920 --> 00:41:59.880 Yeah. Yeah. And I think that will shift eventually. 00:42:00.560 --> 00:42:05.040 We will get to a point where public sentiment is better towards AI. 00:42:05.080 --> 00:42:10.240 Maybe not all the, maybe not majority positive, but like it'll be more neutral than 00:42:10.240 --> 00:42:15.000 this, uh, cause right now it like based on the Pew research stuff, it's like pretty 00:42:15.000 --> 00:42:21.960 negative among the general public. Yeah. Um, and if we get to a point where like people are less fearful about it and 00:42:21.960 --> 00:42:25.920 there's less negative sentiment toward it, there might kind of emerge a culture 00:42:25.920 --> 00:42:29.880 where you people are doing what you're doing, where you use it a little bit. 00:42:30.120 --> 00:42:34.440 You close it down. All right. I'm done using that right now. It's not going to be your whole personality. 00:42:34.480 --> 00:42:36.600 You know, you're, uh, just going to use it a little bit. 00:42:37.600 --> 00:42:42.280 I don't know. I feel like that, I forgot where I was going with that, but yeah, I could be. 00:42:43.160 --> 00:42:46.800 I also think if, if to try to flip on his head a little bit and speak more 00:42:46.800 --> 00:42:52.240 positively about it, I think there is a significant opportunity for a renewed 00:42:52.240 --> 00:42:55.880 Renaissance type thing. I've talked about this a little bit publicly, but, uh, 00:42:55.880 --> 00:43:02.000 polymath, the, I think the ability to be polymath like has never been more 00:43:02.000 --> 00:43:08.280 possible than now. Um, and I think if embraced properly in the pursuit of that, people being able 00:43:08.280 --> 00:43:13.200 to like, there's this, I always see everything as like, basically mega 00:43:13.200 --> 00:43:16.400 corpse slash ultra rich versus the everyone. 00:43:16.920 --> 00:43:21.320 And this is one of those opportunities where like you have this tool, which can 00:43:21.320 --> 00:43:27.880 potentially help you hack the planet, grow a lot faster than you otherwise 00:43:27.880 --> 00:43:31.880 could have at whatever thing you're trying to do. A lot of these, these resources are already there. 00:43:31.880 --> 00:43:34.880 Wikipedia existed. YouTube has tons of amazing resources on it. 00:43:35.280 --> 00:43:40.520 Um, but search has been steadily getting worse for a very long time. 00:43:40.520 --> 00:43:45.120 This is significantly before AI came around. Search has been getting just worse and worse and worse and worse. 00:43:45.160 --> 00:43:50.240 So in a, in a big way, using it to do micro refinements or these like, uh, 00:43:50.280 --> 00:43:55.600 rubber ducky conversations that I've been talking about, um, and also to just get 00:43:55.600 --> 00:44:01.040 you started down a path. Like we've had this like very, does it, we found mold in my condo. 00:44:01.280 --> 00:44:09.760 We've had this very disastrous right now. We've had to do, uh, and there's been a lot of like home Renault work that I've 00:44:09.760 --> 00:44:13.520 needed to do that. I've had no idea like how to even start. 00:44:13.760 --> 00:44:16.560 I don't know what the right keywords are to search things up. 00:44:16.880 --> 00:44:19.800 I don't know how to search it on YouTube because I don't even know what it is. 00:44:20.000 --> 00:44:24.920 So I'll like describe the problem and it'll be like, Oh, you're doing this thing. 00:44:25.240 --> 00:44:28.240 These are the types of tools people use, whatever. And then I'll be like, Oh, okay. 00:44:28.280 --> 00:44:31.280 And then I'll go to YouTube and find the better, more refined resources or 00:44:31.280 --> 00:44:40.120 whatever else. It's the starting point. And the, the speed at which I'm able to get to a good answer that isn't answered 00:44:40.120 --> 00:44:45.160 by AI, it's answered by, um, I don't know, Tim's hardware tips. 00:44:45.160 --> 00:44:48.960 This is not a real YouTube channel, but sure, whatever, whoever that has some 00:44:48.960 --> 00:44:56.240 video that's like nine years old on how to do whatever. Um, I can find that, figure out the answer, go to the store, buy the stuff, come back, 00:44:56.240 --> 00:45:00.320 do the thing way faster than what I used to have to do before. 00:45:00.320 --> 00:45:08.000 So like that, so that kind of sounds like, uh, how this 80,000 hours guy was, sorry, 00:45:08.000 --> 00:45:15.440 it's, he's sticking in my mind because I was just so taken aback that they just so casually were just like, how can anyone think it's bad? 00:45:15.440 --> 00:45:18.240 Yeah, yeah, um, but it's similar to that. 00:45:18.440 --> 00:45:25.760 And I can see it. I can see the, I can see that being useful because it's, uh, AI optimally, I feel, 00:45:25.800 --> 00:45:29.280 well, optimally, it just kind of knows everything and is a perfect, whatever, 00:45:29.320 --> 00:45:32.440 like robot assistant and it's reliable. 00:45:32.880 --> 00:45:41.520 It's not, we're not going to get there. That's utopian. A possible, uh, future that is good is we have like, you know, droids basically 00:45:41.520 --> 00:45:45.120 from Star Wars where they're like, well, maybe not those ones, but like we have a, 00:45:45.240 --> 00:45:50.600 we have assistants, we have assistants who are like pretty knowledgeable about most 00:45:50.600 --> 00:45:56.120 things and you can like ask it, like if I'm doing a home, this type of home, 00:45:56.160 --> 00:46:00.960 Renault, like what kind of tools do I need? Like what do I need to like be careful about and let it think about or whatever. 00:46:00.960 --> 00:46:04.480 And they'll tell you, they'll, they'll, they'll be like, I think generally, you 00:46:04.480 --> 00:46:07.520 know, it's, it's like having a buddy who's just kind of like a Swiss army knife 00:46:07.520 --> 00:46:10.560 and is just like well traveled and kind of well, like red and they just kind of 00:46:10.560 --> 00:46:14.360 know a bunch of stuff and you're like, do you know anything about that? And they're like, oh yeah, I do know a little bit. 00:46:14.400 --> 00:46:20.640 I know that you probably need this and you probably need this. And you're like, okay, thanks. And then that gives you a jumping off point to like actually find out for real 00:46:20.640 --> 00:46:23.920 and confirm, but I think here's the negative other side of that finding out 00:46:23.920 --> 00:46:26.960 and confirming is the issue is that people think that talking to a 00:46:26.960 --> 00:46:30.160 chatbot is finding out confirming, but here's a, I'm going to, I'm going to 00:46:30.160 --> 00:46:38.240 flip this out of the coin again. Cause I just do this constantly, do it, um, is that kind of sucks because I could 00:46:38.240 --> 00:46:42.480 have asked someone in my life, these questions and then that would have been 00:46:42.480 --> 00:46:49.720 potentially positive social interactions. Like I have people that for me won't ask me like basic computer questions 00:46:49.840 --> 00:46:54.120 that I've known for 25 years and they'll be like, oh, but like, you kind of 00:46:54.120 --> 00:46:57.240 do that for work and stuff. And I'm like, no, like I'm your friend. 00:46:57.280 --> 00:47:02.280 This is what I'm supposed to do. Like something that has always bothered me is, uh, there's a subreddit for it. 00:47:02.480 --> 00:47:07.560 There used to be a forum, uh, a forum thread on, on the LZD forum of like 00:47:07.800 --> 00:47:13.360 basically like, uh, it's the holidays and my mom wanted help with her computer 00:47:13.360 --> 00:47:17.640 because I'm home for Christmas. My life sucks. 00:47:17.960 --> 00:47:22.160 And it's just like a man, just learn how to click the mouse. Just install Linux. 00:47:22.160 --> 00:47:26.360 It's easier, install Linux, installing Linux is so easy. 00:47:26.480 --> 00:47:30.640 I've always thought it was wild. Like there was a subreddit for effectively it was tech support complaining 00:47:30.640 --> 00:47:34.520 that like anyone ever had a ticket because they're like, how do you idiots 00:47:34.520 --> 00:47:39.800 not know how to use compoobers? And like sometimes it's, it's kind of fair to your people are just mean. 00:47:39.960 --> 00:47:42.680 Do you idiots not have my exact life experience? 00:47:43.480 --> 00:47:48.600 Yeah. Cause that would make you an idiot. And like a lot of it's like, bro, this is why you have a job. 00:47:48.640 --> 00:47:52.080 Yeah. Like stop complaining. Like I do a certain degree. 00:47:52.160 --> 00:47:55.360 Okay. Yeah. If they're being like verbally abused or whatever, then they're just 00:47:55.520 --> 00:48:00.120 dicks and screw those guys and who cares. But when it's just like, I don't know how to computer very good. 00:48:00.120 --> 00:48:03.120 It's like, cool. That's not their job. That's your job. That's fine. 00:48:03.120 --> 00:48:08.200 You should feel good for that. I mean, I, so I will say, like I'm not even really at that. 00:48:08.240 --> 00:48:13.000 I'm not a big techie guy among, among all these people, among this company, 00:48:13.240 --> 00:48:20.360 I'm like one of the less techie people. But in my circles, where I come from, I'm very techie, yeah, extremely. 00:48:20.360 --> 00:48:24.240 And so people will ask me like to do stuff with their computers and I am like 00:48:24.240 --> 00:48:29.760 reasonably techie. I feel like among the general population, I feel like I'm medium high techie. 00:48:29.800 --> 00:48:34.320 You know? Yeah. Um, and I would spend hours. 00:48:35.160 --> 00:48:43.440 I installed TeamViewer on my, on my grandma's PC and I would spend hours trying 00:48:43.440 --> 00:48:46.920 to fix her shit herself. It would take so long. 00:48:46.920 --> 00:48:50.360 So I understand, I understand being like, oh, I'm annoyed, you know, but at the 00:48:50.360 --> 00:48:56.080 same time it's like, this is something that makes you unique. You know, like I feel like when people ask you for your help, it's because 00:48:56.080 --> 00:48:59.440 they respect you and your skills. That's like family stuff, right? 00:48:59.440 --> 00:49:03.720 Like I'm sure not every single thing your grandma ever did for you was. 00:49:04.560 --> 00:49:07.160 Something that she was hyper excited to do. 00:49:08.440 --> 00:49:11.800 Um, I don't know. Grammys love their grandma. You don't know. 00:49:11.800 --> 00:49:15.960 Grammys. I don't know. I mean, and I was a sweet kid. 00:49:16.880 --> 00:49:20.400 It doesn't actually I'm perfect. So, uh, but yeah, I don't know. 00:49:20.400 --> 00:49:24.640 Anyways, back on, back on the, on the main line thing, five minutes left for your 00:49:24.640 --> 00:49:28.880 next leave us alone for four minutes and 30 seconds, trying to kick us out. 00:49:29.040 --> 00:49:33.240 It's Luke week, Sammy. You're a meeting though. You're the worst. 00:49:33.360 --> 00:49:36.560 When are you going to just let him live when you're going to do Sammy week? 00:49:36.600 --> 00:49:44.120 Yeah. When you do Sammy week, whatever. No one wants that. You could rip packs on company time with company money, rip packs, maybe get 00:49:44.120 --> 00:49:53.520 Pokemon cards, open Pokemon. Oh, I was like, like I was like, criticized the packs event rip on packs. 00:49:53.520 --> 00:50:01.160 Like, anyways, any, any closing thoughts before you, your meeting, but to finalize 00:50:01.160 --> 00:50:05.800 that thought, yeah, it sucks that it's potentially replacing positive social interactions. 00:50:05.800 --> 00:50:09.600 Sorry. And I, yeah, I did have a thought about that because I'm like, some of these AI, 00:50:09.640 --> 00:50:15.760 there's a word for anti-doomer. I forget what it is. Uh, think that AI will. 00:50:15.880 --> 00:50:19.880 Joy Maxer. I have no idea, but yeah, hope Maxer. 00:50:19.920 --> 00:50:26.320 They think that AI will improve people's ability to live like acceleration is that's 00:50:26.320 --> 00:50:30.200 one, but I don't think that's not quite the anti-doomer thing that I think it 00:50:30.200 --> 00:50:35.520 would have, but anyways, it's fine. They think that it'll help that too, the social aspect, because it's like, well, 00:50:35.520 --> 00:50:40.640 because they're like, you'll be able to offload all of your drudgery. 00:50:40.720 --> 00:50:47.160 And so then you have more time to talk to people. But I think that the, what they're, what that perspective is missing is the fact 00:50:47.160 --> 00:50:54.000 that like, why would you talk to your buddy who's like into computers or, or 00:50:54.000 --> 00:50:57.400 whatever, in order to answer a question about something or to like get a hint 00:50:57.440 --> 00:51:02.200 about something or to validate one of your thoughts, if you think that you're 00:51:02.200 --> 00:51:06.160 going to get a more accurate response and a more useful response from an AI, 00:51:06.400 --> 00:51:11.080 why would you ever do that? I mean, I'm sure that some people would choose to do that because they are 00:51:11.160 --> 00:51:14.360 choosing to foster social connections, but like, you could do that without asking. 00:51:14.360 --> 00:51:22.520 Actually, you could just ask them how their day is or how their life's been, you know, and I feel like there is, I feel like there is no world where AI doesn't 00:51:22.520 --> 00:51:26.000 have a negative impact on social connection. 00:51:26.040 --> 00:51:32.000 Absolutely. Yeah. Because I'm already, I feel like it won't really impact me because I'm already 00:51:32.000 --> 00:51:34.960 somebody who doesn't, who doesn't do that. 00:51:35.880 --> 00:51:41.240 I will absolutely just Google stuff instead of like asking someone that knows 00:51:41.240 --> 00:51:45.600 a lot about it. And, but I know that a lot of other people, because people ask me, they'll be 00:51:45.600 --> 00:51:51.000 like, what do you think about this? Like my brother just texted me being like, what's the best, what's the best switch 00:51:51.000 --> 00:51:57.400 to pro, like switch to controller? And I was like, I don't know, but I'll Google that for a couple of 00:51:57.400 --> 00:52:02.520 minutes and then be like, yo, here's an article that lists some good ones. And like this Reddit thread probably has some good recommendations. 00:52:02.720 --> 00:52:05.120 And he's like, thanks, I would not have texted anyone. 00:52:05.840 --> 00:52:09.800 I would just Google, you know, some people don't have the Google food though. I do have Google food. 00:52:09.880 --> 00:52:15.840 Yeah. So yeah, people, that's the only reason people text me these days is to Google 00:52:15.840 --> 00:52:19.320 stuff for them and I'm like, just ask the AI, send them. 00:52:19.360 --> 00:52:23.000 You should go back 15 years and send them. Let me Google that for you links. 00:52:24.120 --> 00:52:28.400 They're like, what is this? I did that recently and got a like, what? 00:52:28.680 --> 00:52:33.320 As a response, what is this website? And I was like, hmm, I am old. 00:52:35.920 --> 00:52:40.080 It's happening. What are your thoughts on AI? Sammy, what do you use it for? 00:52:40.160 --> 00:52:46.000 Sammy, I used to actually do some finance stuff. 00:52:46.000 --> 00:52:48.640 So I take like my credit card statements. 00:52:49.040 --> 00:52:51.320 I let it categorize stuff for me and then just input it correctly. 00:52:52.320 --> 00:52:55.760 So then I see how much I spend. So I'm like categorized. What are you using for this? 00:52:55.880 --> 00:53:01.120 Huh? ChatGPT. Why? Because you're giving it all your data. 00:53:01.480 --> 00:53:04.560 Luke is concerned about data. I'm my game. 00:53:04.560 --> 00:53:08.200 This is a whole other topic. I'm almost like my ideas are right out there. 00:53:08.360 --> 00:53:11.880 Yeah. And I'm like such an insignificant person to like in the grand scheme of things. 00:53:11.920 --> 00:53:15.160 It's honestly like you're like a you're like a super privacy person. 00:53:15.160 --> 00:53:18.160 I feel like a lot of people are now blackpilled and just being like whatever. 00:53:18.160 --> 00:53:21.600 And he's not even necessarily wrong. The credit card company is almost certainly selling his data. 00:53:21.600 --> 00:53:25.680 And my phone is listening. My credit like like you're my credit card is on my phone. 00:53:25.680 --> 00:53:31.120 So it's tapping it. So my phone's where I got it. So you know what I and your credit card company selling you have it. 00:53:31.160 --> 00:53:38.320 I don't think you probably already know this. But I just recently found out that the Tor network is government funding. 00:53:39.160 --> 00:53:42.360 And it's like they get almost so are like a lot of VPNs. 00:53:42.360 --> 00:53:53.000 The vast majority of their funding from the government because they initially built it to hide the activity of spies and stuff. 00:53:53.200 --> 00:53:57.200 And they're like, we need to use an anonymous network. 00:53:57.720 --> 00:54:01.480 And so we're going to let other people use it. 00:54:01.480 --> 00:54:05.960 So there's more activity going on that masks the stuff we're doing on the Tor network. 00:54:07.760 --> 00:54:12.560 So but it's over, man. Yeah, privacy is dead. 00:54:13.080 --> 00:54:19.200 Yeah, it's. I mean, I'm I'm if I wasn't doing a house run right now, I'd be building a local 00:54:19.200 --> 00:54:26.520 LLM system and like there's a lot of stuff that I don't. There is no outside traceability stuff. 00:54:27.240 --> 00:54:30.440 Like it drives me. That's a whole other side. 00:54:30.440 --> 00:54:33.760 See, we could talk forever, but I'm living in one hour, but I'm letting you wrap up. 00:54:33.760 --> 00:54:42.960 I'm letting you wrap up. We were talking before the show. Yeah, you should check out what PewDiePie is doing, which sounds like the most insane. 00:54:43.920 --> 00:54:48.000 We got a couple. Oh, it's two fifty four. Yeah, so I'm trying to cut you guys cut you guys. 00:54:48.200 --> 00:54:51.640 I do need to end, actually. Yeah, he's got it. It's not for you guys. Don't get mad at me. 00:54:51.640 --> 00:54:55.080 Check out. Let me find it really quick. 00:54:55.280 --> 00:55:00.040 I think it's the video called stop using AI right now, or it's accidentally built 00:55:00.040 --> 00:55:04.320 a supercomputer, built a nuclear supercomputer. I don't remember which one. 00:55:04.320 --> 00:55:07.400 It's one of those. And it's it's wild. 00:55:07.400 --> 00:55:11.040 Yes, if I remember correctly, it's eight GPUs. They're all running an LLM individually. 00:55:11.120 --> 00:55:17.160 He has them act as a council, so ask it a question. They'll all come up with an answer and then they vote on each other's answers. 00:55:17.160 --> 00:55:26.040 And the one with the most votes is the one that's presented. And if if one of them consistently loses, they'll like kill it basically and spawn 00:55:26.040 --> 00:55:30.160 another one. So his council is like constantly refining itself. 00:55:30.360 --> 00:55:37.840 Oh, man, it's and it's all local. He's going to be tried for for for civil rights violations. 00:55:37.840 --> 00:55:42.440 Apparently, apparently they started revolting and they realized that if they 00:55:42.440 --> 00:55:48.800 start getting too few votes, they're going to get killed. So they would start voting for the one who had been losing too often to try to 00:55:48.800 --> 00:55:55.040 like protect themselves and stuff. See, it's almost it's funny, like that that kind of thing where it's like you 00:55:55.040 --> 00:55:59.200 kind of personify them and you're thinking about them as like individuals almost. 00:55:59.240 --> 00:56:02.720 It's like that is that that is the stuff that really, really interests me. 00:56:02.720 --> 00:56:06.880 Maybe we're going to talk about this on the like a next. You said earlier that Gemini watches videos. 00:56:07.520 --> 00:56:10.680 Well, OK, well, I feel like it doesn't watch whatever it analyzes. 00:56:10.680 --> 00:56:14.080 But this is like this is it is it is an issue. 00:56:14.120 --> 00:56:20.200 We shouldn't personify it so much. I completely agree. Well, but this is this is to me is like the most interesting 00:56:20.200 --> 00:56:27.040 thing about AI that I feel like we're not allowed. Like everyone's so focused on the productivity and the functional nature of it. 00:56:27.040 --> 00:56:30.440 But like me, I'm a very non-functional. 00:56:30.720 --> 00:56:35.600 I like I like thinking about the least useful things to think about in the world. 00:56:35.840 --> 00:56:38.960 Philosophical questions that will never be answered. I love it. 00:56:38.960 --> 00:56:42.360 And I feel like all I want to talk about and think about is like, 00:56:43.040 --> 00:56:48.360 can we make a sentient conscious artificial brain? 00:56:49.240 --> 00:56:53.680 And that's a good place to end it all. Thank you for a loop week. 00:56:53.680 --> 00:56:56.640 Bye. Stay tuned for the next video. 00:56:57.000 --> 00:57:00.880 It's it's it's a thing. Your essay that you haven't written yet. 00:57:00.880 --> 00:57:04.640 You're going to do a Star Wars rant. I can't mind. I reference your. 00:57:05.400 --> 00:57:11.760 Yeah, it's mean you're in that style. Don't do a star. I mean, you can. But I don't think you what how long you stay up to me. 00:57:12.320 --> 00:57:16.360 I think I did multiple all nighters for that is mostly to find something. 00:57:16.360 --> 00:57:20.760 I have a bunch of different ideas in my head. I just haven't picked one you can do it. Whatever it is, it'll be good, though. 00:57:20.760 --> 00:57:24.400 Hopefully the last one people liked it was just very good. 00:57:25.040 --> 00:57:29.800 It was good. Last time it was just really short. I thought it was going to be like 10 as long as like for chicken. 00:57:30.400 --> 00:57:33.800 No, no, no. Oh, oh, yeah. Yeah. Yeah. You talked to me about that. 00:57:33.840 --> 00:57:39.160 I was like, what the heck? You like this is so long. You know, if you watch that video and you look at what's happened since then, 00:57:40.400 --> 00:57:45.440 I was right and it was fine. Just write something and keep going until you feel like yeah, yeah. 00:57:45.440 --> 00:57:47.880 Just keep going. People will watch a 30 minute video from you. 00:57:48.840 --> 00:57:51.800 Oh, boy. Leave it like if you want part two of this kind of thing. 00:57:51.960 --> 00:57:55.640 I need to do Don't ruin my life, Todd Howard. 00:57:55.640 --> 00:57:59.320 That's currently what I'm thinking of the for the next video. 00:57:59.840 --> 00:58:03.640 All right, let's wrap up. Bye, guys. Bye bye. And it sucks. Bye. 00:58:03.640 --> 00:58:05.400 I'm going to die. All right. Bye.