WEBVTT - How Much AI Do We Need? - My AI Industry Prediction 0:00:01.070 --> 0:00:04.430 Welcome to Unsupervised Learning, a security, AI and meaning focused 0:00:04.430 --> 0:00:07.280 podcast that looks at how best to thrive as humans 0:00:07.280 --> 0:00:11.539 in a post AI world. It combines original ideas, analysis, 0:00:11.539 --> 0:00:14.750 and mental models to bring not just the news, but 0:00:14.750 --> 0:00:21.950 why it matters and how to respond. Hey what's up? 0:00:21.950 --> 0:00:26.960 It's Daniel with Unsupervised learning. I'm building AI to upgrade humans, 0:00:26.960 --> 0:00:29.060 and today I want to talk about a cool idea. 0:00:29.090 --> 0:00:32.419 Basically trying to figure out how to predict how much 0:00:32.450 --> 0:00:36.890 AI infrastructure we actually need and how far along we 0:00:36.890 --> 0:00:40.129 are along that path. And I want to say something 0:00:40.130 --> 0:00:42.500 about predictions real quick, which I've talked about before in 0:00:42.500 --> 0:00:45.290 other videos, but I don't think it's possible to predict 0:00:45.320 --> 0:00:49.670 tech really, the way that tech happens, who wins, who loses, 0:00:49.700 --> 0:00:54.080 like the timelines? Like this stuff is like impossible to predict, basically. 0:00:54.080 --> 0:00:57.260 So I wouldn't want you to think that I would 0:00:57.260 --> 0:01:00.510 try to do that because I think that's foolish. What 0:01:00.510 --> 0:01:04.110 I believe we can predict related to tech are the 0:01:04.110 --> 0:01:07.830 human things that are related to human desires. So the 0:01:07.830 --> 0:01:11.730 question is not whether we can predict tech itself, but 0:01:11.730 --> 0:01:14.820 can we predict what we want from tech. And I 0:01:14.819 --> 0:01:18.089 think that is a powerful way to kind of see 0:01:18.090 --> 0:01:21.300 what might be happening and what could unfold in the 0:01:21.300 --> 0:01:24.180 next 1 to 3 years or whatever. So I want 0:01:24.209 --> 0:01:27.300 to specifically look at like, how far along are we 0:01:27.330 --> 0:01:32.100 to building the infrastructure that we need for AI, right. So, um, 0:01:32.100 --> 0:01:35.699 I own a bunch of, uh, Nvidia, um, I think 0:01:35.700 --> 0:01:39.600 I've got some TSMC like, um, I've got stock related 0:01:39.600 --> 0:01:42.179 to AI. I'm all in on AI, have been since 0:01:42.180 --> 0:01:47.280 late 2022. Uh, I went independent to work in this field, um, 0:01:47.280 --> 0:01:52.500 after being in security for like 25 years. So I'm very, 0:01:52.530 --> 0:01:55.380 I guess, religious about the whole thing. So I want 0:01:55.380 --> 0:01:57.150 you to understand that which you probably already know if 0:01:57.180 --> 0:01:59.860 you've seen any of my videos? Um, but I don't 0:01:59.860 --> 0:02:03.220 think that affects this particular thing that we're talking about. 0:02:03.250 --> 0:02:05.590 I'm sure it does in some sense, because that's the 0:02:05.590 --> 0:02:08.500 way bias works. But I think this argument I'm about 0:02:08.500 --> 0:02:12.970 to make stands independent of that. So essentially what I'm 0:02:12.970 --> 0:02:17.140 looking to figure out is how much infrastructure do we need? 0:02:17.169 --> 0:02:19.540 How many GPUs do we need? How many data centers 0:02:19.540 --> 0:02:22.660 do we need? How many AI companies and startups do 0:02:22.660 --> 0:02:24.790 we need? And some of that depends on the actual 0:02:24.790 --> 0:02:28.690 technical implementations, which we can't predict. So we don't need 0:02:28.690 --> 0:02:30.580 to go too deep into that. But I'm just trying 0:02:30.580 --> 0:02:34.420 to figure out, are we at like 13% of how 0:02:34.419 --> 0:02:37.120 much I. We need a lot of people think we're 0:02:37.150 --> 0:02:41.410 at 85%. For example, they're like, yeah, that was pretty 0:02:41.410 --> 0:02:45.519 much it in 2023 and 2024. And now it's mostly 0:02:45.520 --> 0:02:47.860 hype and it's just going to die down. So they 0:02:47.860 --> 0:02:51.070 think it's like 87% and it's pretty much done. It 0:02:51.070 --> 0:02:54.730 was a cycle. Whatever other people are like no we're 0:02:54.730 --> 0:02:57.400 just getting started. We're at like 3%, but it's going 0:02:57.430 --> 0:02:59.950 to grow over the next five years or whatever. And 0:02:59.950 --> 0:03:03.100 I guess I'll just spoil what I think the answer is. 0:03:03.100 --> 0:03:07.780 I think the answer is we're at like 0.000. I 0:03:07.780 --> 0:03:12.070 don't know how many zeros, like eight zeros and a 1% 0:03:12.070 --> 0:03:16.120 of where we need to be or where's more specifically 0:03:16.120 --> 0:03:21.369 where humans will demand that we get. Okay. So that's 0:03:21.370 --> 0:03:23.020 that's the point I want to make here. And the 0:03:23.020 --> 0:03:25.419 way I want to get there is by asking a question, 0:03:25.419 --> 0:03:28.660 what are we actually trying to do with AI? What 0:03:28.690 --> 0:03:30.610 are we trying to do with AI? So I want 0:03:30.639 --> 0:03:34.300 to walk through a few ideas there. So one cool 0:03:34.300 --> 0:03:38.530 idea here is learning from the past and anticipating the future. 0:03:38.530 --> 0:03:40.690 So we are sitting in the middle here in the 0:03:40.690 --> 0:03:44.140 current present moment, and we could use AI to look 0:03:44.140 --> 0:03:47.140 into the past, find things that went wrong for the 0:03:47.140 --> 0:03:50.620 purpose of adjusting our current behavior right on the other side. 0:03:50.620 --> 0:03:55.600 We could anticipate, try to anticipate Paint based on everything 0:03:55.600 --> 0:03:59.140 we currently know. Try to anticipate what's going to happen 0:03:59.140 --> 0:04:03.070 in the future. For what reason? Same exact reason. Adjust 0:04:03.070 --> 0:04:06.100 our current behavior. So I think this is a paradigm. 0:04:06.100 --> 0:04:09.880 This is an idea or a concept that's extremely powerful. 0:04:09.880 --> 0:04:12.610 And so the question is how much more of this 0:04:12.640 --> 0:04:15.190 do we need? How much more of this do we need? 0:04:15.190 --> 0:04:17.919 Or how bad of a job are we currently doing 0:04:17.920 --> 0:04:20.410 at this? And how much more can we gain if 0:04:20.440 --> 0:04:23.140 we were doing it much, much better? So I won't 0:04:23.140 --> 0:04:26.650 even give my answer, but probably kind of rhetorical. All right, 0:04:26.680 --> 0:04:29.890 next idea I would say that for any given situation, 0:04:29.890 --> 0:04:33.220 for any given operation or thing you're trying to do, 0:04:33.250 --> 0:04:36.790 there is a current state and there is a desired state, 0:04:36.790 --> 0:04:39.429 and there is a delta between those two. And I 0:04:39.430 --> 0:04:42.580 think this is a really powerful concept. It automatically asks 0:04:42.610 --> 0:04:46.330 a question. It demands a question how do we get 0:04:46.330 --> 0:04:50.920 from the current state to the desired state? So I 0:04:50.920 --> 0:04:55.550 ask you again, how many situations do we have worldwide, 0:04:55.580 --> 0:05:00.230 human civilization wide, where we have a current situation that 0:05:00.230 --> 0:05:03.500 we're not really happy with, and we either need help 0:05:03.500 --> 0:05:05.779 coming up with a desired state, or we already have 0:05:05.779 --> 0:05:09.440 a desired state clearly in mind. Or maybe we have 0:05:09.470 --> 0:05:12.409 thoughts about a desired state, but we could use some 0:05:12.410 --> 0:05:15.200 help articulating it. But either way, we want to get 0:05:15.200 --> 0:05:18.110 to this desired state, which I could potentially help us 0:05:18.110 --> 0:05:21.260 articulate the desired state. But either way, how are we 0:05:21.290 --> 0:05:26.990 doing on understanding our current state perfectly, articulating our desired state, 0:05:26.990 --> 0:05:38.390 and figuring out the delta between those two? Think education. Poverty. Climate. Health. Aging. Politics. War. 0:05:38.420 --> 0:05:42.200 Just just think all the whole scope of human problems 0:05:42.200 --> 0:05:45.230 over the course of history. How are we doing on this? 0:05:45.230 --> 0:05:49.220 How much potential is there to do this better? That's 0:05:49.220 --> 0:05:51.650 the second one. And the second one really brings up 0:05:51.650 --> 0:05:55.520 this third one. What are the actions that we should 0:05:55.520 --> 0:05:59.870 take to make the transition? This this is incredibly powerful. 0:05:59.900 --> 0:06:02.900 Can we use AI to understand the current state? Can 0:06:02.930 --> 0:06:07.820 we use AI to articulate and structure and build and 0:06:07.820 --> 0:06:12.500 communicate the desired state? And most importantly, what's the plan? 0:06:12.500 --> 0:06:15.440 How do we transition from one to the other? How 0:06:15.440 --> 0:06:18.380 much do we need this middle piece? How good of 0:06:18.380 --> 0:06:22.190 a job are we as humans doing at executing on 0:06:22.190 --> 0:06:25.490 this middle piece? I would argue not very well at all. 0:06:25.490 --> 0:06:28.489 And we haven't been for centuries. Well, I mean, you 0:06:28.490 --> 0:06:32.000 could argue it could go way worse. Totally. Absolutely could 0:06:32.000 --> 0:06:35.690 go way worse. So sure, we're not all dead yet, 0:06:35.690 --> 0:06:38.630 so that's good. But if you look at the course 0:06:38.630 --> 0:06:41.690 of history, it's just like folly after folly. And we're 0:06:41.690 --> 0:06:44.690 currently in a whole bunch of it right now. And 0:06:44.690 --> 0:06:47.029 it's a bad situation to be in. So I would 0:06:47.029 --> 0:06:49.310 argue that this yellow piece right here in the middle 0:06:49.320 --> 0:06:55.020 is extraordinary. It is. It has extraordinary potential for what 0:06:55.020 --> 0:06:57.900 we wish we could do with that yellow to transition 0:06:57.900 --> 0:07:02.880 to this desired state. So let's break these three down 0:07:02.880 --> 0:07:05.640 even more. When you look at the current state, you 0:07:05.640 --> 0:07:08.850 talk about, okay, let's understand the current state. How granular 0:07:08.880 --> 0:07:12.390 are we talking about? Right. Because you could go from 0:07:12.420 --> 0:07:15.720 like a coffee cup with like an amount of hot 0:07:15.720 --> 0:07:18.000 liquid inside of it. And we could ask the question, 0:07:18.000 --> 0:07:20.640 what is the current state of this liquid inside this 0:07:20.640 --> 0:07:23.070 coffee cup? What's the current state of the coffee? Well, 0:07:23.070 --> 0:07:24.960 you could ask lots of different people about this. You 0:07:24.960 --> 0:07:27.750 could ask a philosopher, you could ask whatever. It's like, oh, 0:07:27.780 --> 0:07:31.800 half full, half empty. Um, you could ask like a 0:07:31.830 --> 0:07:35.520 chemist or a physicist. And it's like, well, um, what 0:07:35.520 --> 0:07:37.290 do you want to know? Like how many atoms, like, 0:07:37.320 --> 0:07:40.410 how excited are they? What the the location of all 0:07:40.410 --> 0:07:44.190 the electrons. Certain parts of this question for a given 0:07:44.220 --> 0:07:47.880 object are just not knowable. Like, what is the state 0:07:47.880 --> 0:07:53.160 of all atoms and subatomic particles in, say, a pen 0:07:53.160 --> 0:07:56.730 or in a cup of coffee or the state of 0:07:56.730 --> 0:08:00.720 a human, for example, you say, how is he doing? 0:08:00.720 --> 0:08:03.840 How is she doing? Well, what do you mean? I 0:08:03.840 --> 0:08:05.940 can't give you a full rundown of every atom in 0:08:05.940 --> 0:08:09.360 their body. So the question is, what level of depth 0:08:09.390 --> 0:08:12.150 do we need to be able to answer that question? 0:08:12.150 --> 0:08:15.660 And that comes down to how much telemetry can we get, 0:08:15.690 --> 0:08:18.809 how much data can we get coming off of this situation. 0:08:18.810 --> 0:08:21.630 So that's I've got an aura ring on, I've got 0:08:21.630 --> 0:08:24.600 an Apple Watch on. This is telemetry coming off of 0:08:24.600 --> 0:08:28.980 me so that you can ask an AI very soon 0:08:29.010 --> 0:08:33.090 or actually now in my opinion, Apple is building life 0:08:33.090 --> 0:08:36.959 OS essentially, in case you guys didn't know that. So 0:08:36.960 --> 0:08:40.860 you will be able to ask yourself, how am I doing? Um, 0:08:40.860 --> 0:08:43.290 and it will know. Okay, you're talking about health. You're 0:08:43.290 --> 0:08:45.750 talking about financially, you're talking about education. What are you 0:08:45.750 --> 0:08:48.730 talking about? But you have to have that telemetry coming in, right? 0:08:48.760 --> 0:08:51.910 And you have to know what the limits are right for. 0:08:51.940 --> 0:08:54.640 For a given object type, you have to know certain 0:08:54.640 --> 0:08:59.890 questions and certain types of metrics that are just attainable. Okay. 0:08:59.890 --> 0:09:04.360 Can we get someone's current heart rate? No, not 15 0:09:04.360 --> 0:09:07.840 years ago. Not automatically, not every moment of the day. 0:09:07.840 --> 0:09:10.660 We couldn't do that before. We can now. So that's 0:09:10.660 --> 0:09:15.910 a technology change that enabled that level of metric to happen. Well, 0:09:15.940 --> 0:09:19.240 Apple is currently gathering things like mood as well. How 0:09:19.270 --> 0:09:22.300 are you feeling today. And it's associating those things with 0:09:22.300 --> 0:09:25.840 other things like how much you've exercised and stuff like that. Right. 0:09:25.840 --> 0:09:28.060 So you kind of see where this is all going 0:09:28.059 --> 0:09:32.410 with life. OS now that's for a person. Okay. Now 0:09:32.410 --> 0:09:36.010 let's talk about a family. What's the state of the family? Well, 0:09:36.010 --> 0:09:38.559 what does that mean? Do we need all of those 0:09:38.559 --> 0:09:42.160 metrics for each individual person? Sure. But you also want 0:09:42.190 --> 0:09:44.800 to know the dynamics between the people. You also want 0:09:44.830 --> 0:09:48.080 to know how the family as a whole is doing. Um, 0:09:48.080 --> 0:09:51.949 what's the quality of life? What is the upward trajectory 0:09:51.950 --> 0:09:55.190 look like? Right? This is the state of a family. 0:09:55.280 --> 0:09:58.880 Now let's talk about business companies. What is the state 0:09:58.880 --> 0:10:00.800 of a company? Well, what are we talking about? Its 0:10:00.830 --> 0:10:06.500 IT infrastructure with thousands of Kubernetes pods running. What's the 0:10:06.500 --> 0:10:09.230 current state of a Kubernetes pod right now? And also 0:10:09.230 --> 0:10:12.020 right now and also right now, how often are we 0:10:12.020 --> 0:10:15.620 polling these things both for a human and for a company? Right. 0:10:15.650 --> 0:10:18.050 And a lot of these questions are going to come 0:10:18.050 --> 0:10:21.260 down to, sure, we could poll the current state of 0:10:21.260 --> 0:10:26.060 every single Kubernetes pod at, say, I don't know, Google, right? 0:10:26.059 --> 0:10:29.480 Every single second, but that might cost billions of dollars 0:10:29.480 --> 0:10:32.420 because it takes, you know, it's just expensive. And where 0:10:32.420 --> 0:10:34.579 would you store the stuff in all these different questions. 0:10:34.580 --> 0:10:37.670 So you've got like how granular, what is the update 0:10:37.670 --> 0:10:40.670 frequency and how expensive is that to gather based on 0:10:40.670 --> 0:10:45.410 the current tech. Right. And this is Tremendously interesting and 0:10:45.410 --> 0:10:49.219 tremendously powerful, especially for something like a human. The most 0:10:49.220 --> 0:10:52.640 important driver for all of this AI stuff one is 0:10:52.640 --> 0:10:55.400 going to be business, no question. Right? And that's going 0:10:55.429 --> 0:10:58.370 to have the money behind it or whatever. But ultimately, 0:10:58.370 --> 0:11:01.820 the bigger thing is going to be humans using this 0:11:01.820 --> 0:11:06.530 to feel better and be better and improve themselves. That's 0:11:06.530 --> 0:11:09.949 that's the entire game here. What are people struggling with 0:11:09.950 --> 0:11:14.870 right now? Job happiness. Can't get a job. Loneliness. A 0:11:14.870 --> 0:11:18.380 lack of meaning in their lives. Okay. How do you 0:11:18.410 --> 0:11:20.990 solve how how do you solve the current state desired 0:11:20.990 --> 0:11:24.679 state situation for somebody who's lonely? And I got another 0:11:24.679 --> 0:11:27.679 example of that coming up. But think about that. Think 0:11:27.679 --> 0:11:31.820 about what it takes to sort of solve these problems. 0:11:31.850 --> 0:11:35.839 Go from current state, desired state for a human, for 0:11:35.840 --> 0:11:39.650 a company, and what level of context you need to 0:11:39.679 --> 0:11:42.270 be able to answer those sorts of questions. now, I 0:11:42.270 --> 0:11:46.050 would argue just like the previous ones, we are nowhere 0:11:46.080 --> 0:11:51.809 near gathering enough context on the current state of anything. 0:11:51.840 --> 0:11:56.880 Of anything. Okay, a park bench, a tree and at 0:11:56.880 --> 0:12:00.210 the park, the state of a human that you care about. 0:12:00.240 --> 0:12:03.360 We don't have nearly enough telemetry on that. We want 0:12:03.390 --> 0:12:05.520 to change a business. We have no idea what's going 0:12:05.520 --> 0:12:08.280 on in business. Most people who work there and actually 0:12:08.280 --> 0:12:10.260 run the business, they have no idea what's going on 0:12:10.260 --> 0:12:12.540 in the business. They don't know the level of mood 0:12:12.540 --> 0:12:16.679 and happiness and the state. How much waste is happening 0:12:16.679 --> 0:12:20.939 at any given time? Bottom line here is things like 0:12:20.940 --> 0:12:26.790 context size, things like Rag technologies like these are absolutely 0:12:26.790 --> 0:12:30.720 in their infancy because they have to be because of 0:12:30.720 --> 0:12:34.140 the size of the scope of the problem that they 0:12:34.140 --> 0:12:38.190 actually are being signed up to solve. Okay. Uh, Altman 0:12:38.190 --> 0:12:40.680 talked about this a while back. He was like, look, 0:12:40.710 --> 0:12:44.280 at some point, context is basically going to be infinite, 0:12:44.280 --> 0:12:49.620 and obviously nothing's infinite. So that doesn't really mean infinite. 0:12:49.620 --> 0:12:52.860 It means functionally infinite. So there's going to be this 0:12:52.860 --> 0:12:57.990 competition between context size going into models versus Rag, going 0:12:58.020 --> 0:13:02.640 into models versus whatever, whatever comes out later and makes 0:13:02.640 --> 0:13:07.199 maybe those not as important. But either way, what you 0:13:07.200 --> 0:13:09.900 have to do is you have to get this current 0:13:09.900 --> 0:13:13.530 state for a thing that matters to humans. Forget tech 0:13:13.559 --> 0:13:16.890 is about a thing that matters to humans into the 0:13:16.890 --> 0:13:20.340 brain of the AI, so that all it's understanding of 0:13:20.340 --> 0:13:23.910 the world can be applied to this larger situation. Now 0:13:23.940 --> 0:13:26.640 think about this as a human. Imagine somebody who's 40 0:13:26.640 --> 0:13:30.270 years old and has had this amazing life and, you know, 0:13:30.300 --> 0:13:34.620 the hardships and people died and they fought in wars 0:13:34.620 --> 0:13:37.770 and they traveled to Costa Rica and they did ayahuasca, 0:13:37.770 --> 0:13:40.089 and they learned all these things. They know thousands of 0:13:40.090 --> 0:13:43.570 people and they've had this wonderful life, and they're trying 0:13:43.570 --> 0:13:47.050 to craft even a better life. And you want to 0:13:47.080 --> 0:13:49.630 ask an AI, you know, tell me what I can 0:13:49.630 --> 0:13:52.780 work on. Tell me about myself. What do you suggest 0:13:52.780 --> 0:13:56.410 for me? I'm looking at doing this career or this career. 0:13:56.410 --> 0:13:58.870 I'm looking at moving here. I'm looking at marrying this 0:13:58.870 --> 0:14:02.260 girl or not marrying the squirrel. Um. What should I do? 0:14:02.290 --> 0:14:05.920 What does the. I need to know about that person 0:14:05.920 --> 0:14:10.120 to make that decision? Or to offer advice in making 0:14:10.120 --> 0:14:14.950 that decision? Ideally, it has everything. It's got the genome. 0:14:14.950 --> 0:14:18.760 It's it's got all your medical records growing up. Ideally 0:14:18.760 --> 0:14:23.200 it has journal entries. Ideally, it's been recording you at 0:14:23.200 --> 0:14:26.530 a deep level of state gathering for your entire life. 0:14:26.530 --> 0:14:29.229 So imagine this is 200 years from now and this 0:14:29.230 --> 0:14:31.660 person is 40 years old. But every moment of their 0:14:31.660 --> 0:14:34.780 life has been captured. And the very moment that you 0:14:34.780 --> 0:14:38.230 ask a question, what should I do? Which career? Which 0:14:38.230 --> 0:14:41.800 country do I move to? Which guy do I marry? Um, 0:14:41.800 --> 0:14:46.000 the entirety of their life up to that moment would 0:14:46.000 --> 0:14:49.000 be the context for answering that question. This is the 0:14:49.000 --> 0:14:51.850 key point for this entire thing. Think of the AI 0:14:51.850 --> 0:14:57.220 infrastructure that is required to have the entire context of 0:14:57.220 --> 0:15:00.070 the moment, right up to the point when the question 0:15:00.070 --> 0:15:05.290 is asked to be jammed into that AI's mind. But 0:15:05.290 --> 0:15:07.930 don't just imagine it for this 40 year old guy 0:15:07.930 --> 0:15:11.950 in Costa Rica. Imagine it for a giant business with 0:15:11.950 --> 0:15:17.560 10,000 employees who's been in business for 120 years, how 0:15:17.560 --> 0:15:20.560 much context, at a deep level, can we grab from 0:15:20.560 --> 0:15:24.160 that entire 120 years and put it into what should 0:15:24.160 --> 0:15:28.390 I do now that is big. That is massive. And 0:15:28.390 --> 0:15:31.300 here's the crazy part. It can be summarized to a 0:15:31.300 --> 0:15:34.030 ten page text document. It can be summarized to a 0:15:34.030 --> 0:15:38.510 1010 page text document. And guess what? It would be 0:15:38.510 --> 0:15:42.380 good if you could summarize it to a 200 page report. 0:15:42.380 --> 0:15:45.710 It would be really good. It could also be terabytes 0:15:45.710 --> 0:15:48.710 of data. Petabytes of data. How quickly can you get 0:15:48.710 --> 0:15:52.790 petabytes of data into an AI's brain to snap and 0:15:52.790 --> 0:15:56.450 answer the question in a really, really powerful way for 0:15:56.450 --> 0:15:59.480 a business, for a country, for a city, for a human, 0:15:59.480 --> 0:16:03.290 for a family. Think of the infrastructure that is required 0:16:03.290 --> 0:16:06.170 to do that, and think about the fact that this 0:16:06.170 --> 0:16:09.890 is what humans want. This is what humans will demand. 0:16:09.920 --> 0:16:13.520 This is why I'm saying forget the tech. Forget the 0:16:13.520 --> 0:16:17.240 tech does not matter. What matters is what humans want. 0:16:17.270 --> 0:16:23.630 Humans want the most amazing answer ever to that question. 0:16:23.660 --> 0:16:26.480 Do I do this merger with this company? Do I 0:16:26.510 --> 0:16:30.470 hire this person as my CFO? And the amount of 0:16:30.470 --> 0:16:33.800 data that is needed to get a better and better 0:16:33.800 --> 0:16:37.670 and better answer to that question just keeps going up 0:16:37.670 --> 0:16:41.630 along with the capabilities of the tech. Right. So this 0:16:41.630 --> 0:16:45.170 is what's so powerful about this paradigm of thinking about 0:16:45.170 --> 0:16:49.370 this is that there's kind of no end. There's no 0:16:49.370 --> 0:16:53.330 foreseeable end coming to how much better it gets when 0:16:53.330 --> 0:16:57.620 you give it more. And that's why okay, 200 K 0:16:57.650 --> 0:17:01.489 tokens right now. So that's a lot. No it's not 0:17:01.520 --> 0:17:03.950 it's not a lot. It's not a lot because watch this. 0:17:03.950 --> 0:17:06.800 The thing a human wants to do is to say hey, 0:17:06.800 --> 0:17:11.120 I've got an idea for a concept for this really cool, um, 0:17:11.150 --> 0:17:14.750 mini series that's going to be on Netflix. It's got anime, 0:17:14.780 --> 0:17:18.859 it's got Harry Potter stuff, but it's in this stylized thing, 0:17:18.890 --> 0:17:22.399 kind of like The Last Samurai or, um, or what 0:17:22.400 --> 0:17:25.850 was that? It was called Shogun, the recent animated thing. Anyway, 0:17:25.850 --> 0:17:27.470 you give it all these things, you're like, I want 0:17:27.470 --> 0:17:30.050 this art. I want the concept of Star Wars, but 0:17:30.050 --> 0:17:32.520 it's got to be Harry Potter. But I also want 0:17:32.550 --> 0:17:35.640 like deep romance. But I want it to be really, 0:17:35.640 --> 0:17:38.460 really gritty. And it definitely, you know, got to be 0:17:38.460 --> 0:17:40.770 over 21 to even watch this thing. And there's also 0:17:40.770 --> 0:17:43.080 violence and there's also sex and there's also all these 0:17:43.080 --> 0:17:45.840 different things. And you basically hit it with that and 0:17:45.840 --> 0:17:47.850 you're like, yeah. But I also want it to be 0:17:47.850 --> 0:17:51.090 kind of like Tolstoy. I really like Tolstoy. And it's 0:17:51.090 --> 0:17:54.900 going to go and write you and build you a 0:17:54.900 --> 0:17:59.700 complete movie, a complete series, and do all these different 0:17:59.700 --> 0:18:01.890 things with agents, all the stuff that's coming out. That's 0:18:01.890 --> 0:18:04.770 kind of inevitable, right? How much does it need to 0:18:04.800 --> 0:18:07.050 know to be able to do that perfectly? It's got 0:18:07.050 --> 0:18:09.390 to be able to read into everything you're saying. It's 0:18:09.390 --> 0:18:12.629 got to have all the capabilities of actually doing the 0:18:12.630 --> 0:18:15.270 video and the art creation and all this stuff, but 0:18:15.270 --> 0:18:18.030 the more it knows, the better this thing gets. And 0:18:18.030 --> 0:18:21.540 that doesn't really stop at any time soon. Right? So 0:18:21.540 --> 0:18:24.360 this is needed for business. This is needed for creativity. 0:18:24.359 --> 0:18:28.050 This is needed for human thriving at an individual and 0:18:28.050 --> 0:18:31.740 a society level. There is no end to how much 0:18:31.740 --> 0:18:35.040 we will demand this thing. I'll give you another example here. 0:18:35.430 --> 0:18:39.810 Because of this assassination that happened with the head of United, 0:18:40.020 --> 0:18:43.889 UnitedHealth and I talked about this in in my book, 0:18:43.890 --> 0:18:47.310 The Real Internet of Things in like 20, 2016. It 0:18:47.310 --> 0:18:50.070 came out and I've recently done a whole series on 0:18:50.100 --> 0:18:53.220 like modernizing it. And it's got video and everything, so 0:18:53.220 --> 0:18:55.409 you should check it out. But the point is, what 0:18:55.410 --> 0:18:57.780 I talked about in that book was walking down the 0:18:57.780 --> 0:19:01.649 street and having sensors all around you. This is the 0:19:01.650 --> 0:19:04.410 vision that I'm seeing for everything. You have sensors all 0:19:04.410 --> 0:19:06.960 around you, the current state of the world. This thing 0:19:06.960 --> 0:19:09.750 on the left here, this current state of the world, 0:19:09.750 --> 0:19:13.380 is the thing that your eye is always monitoring for you. 0:19:13.410 --> 0:19:17.040 Your Da, your digital assistant. Right? It could see behind you. 0:19:17.070 --> 0:19:19.649 You could see above you. Why is it seeing above you? 0:19:19.680 --> 0:19:23.550 This is dystopian, but this is real. Because of drones, 0:19:23.550 --> 0:19:26.340 that's why. Why is it seeing behind you? Because someone 0:19:26.340 --> 0:19:29.410 might walk up behind you with a big backpack on 0:19:29.410 --> 0:19:32.200 and a mask and pull out a gun and shoot you. 0:19:32.200 --> 0:19:35.770 So what your eye has to be doing is watching 0:19:35.770 --> 0:19:38.859 everything all the time. And guess what? If you're a 0:19:38.859 --> 0:19:41.530 VIP or you just have some money, you're going to 0:19:41.530 --> 0:19:45.850 have little drones flying around looking at everything, looking down. Hey, 0:19:45.850 --> 0:19:48.070 why is that car moving like that? Hey, moved. Move 0:19:48.100 --> 0:19:49.689 to the left across the street. I don't want you 0:19:49.690 --> 0:19:51.969 over here. Your little ear piece is going to be 0:19:51.970 --> 0:19:55.629 giving you these little guidance things. It hears your stomach rumble. 0:19:55.630 --> 0:19:58.060 It's like, hey, there's Thai food up there. It's your 0:19:58.060 --> 0:20:02.260 favorite restaurant. I already pinged Jeremiah. He's got your favorite table. 0:20:02.260 --> 0:20:04.180 I turned on table tennis. So when you get there, 0:20:04.180 --> 0:20:08.199 table tennis is on. It is monitoring the state of 0:20:08.200 --> 0:20:12.340 the world and changing the state of that world in 0:20:12.340 --> 0:20:14.830 preemptively watching for the state of the world to make 0:20:14.859 --> 0:20:18.310 sure it doesn't turn into an undesired state for you 0:20:18.340 --> 0:20:21.910 24 over seven like every second. Boom, boom, boom boom. 0:20:21.910 --> 0:20:25.629 It's reading every API. It's looking at every person. It's 0:20:25.630 --> 0:20:28.220 pulling up their daemon to see if there if there's 0:20:28.250 --> 0:20:31.520 information on them, like should we go introduce ourselves to them? 0:20:31.520 --> 0:20:34.729 This is the parsing of the current state of the 0:20:34.730 --> 0:20:38.840 world around us at all times. That once again only 0:20:38.840 --> 0:20:41.990 gets better and better with more tech, right? You could 0:20:41.990 --> 0:20:47.060 do this a little bit in 2025 and 2026 and 27. 0:20:47.060 --> 0:20:49.310 That'll be pretty good. But wait till you see it 0:20:49.310 --> 0:20:51.170 in five years. Wait till you see it. In ten 0:20:51.170 --> 0:20:54.440 years it will be absolute sci fi stuff. And again, 0:20:54.440 --> 0:20:57.590 you don't need to think about the tech. You only 0:20:57.590 --> 0:21:00.889 need to think about the fact that humans wish they 0:21:00.890 --> 0:21:04.939 could see around themselves at all times. They wish they 0:21:04.940 --> 0:21:08.990 had that tech on all of their kids at all times. 0:21:08.990 --> 0:21:13.760 So when some white kid jumps out of a jumps 0:21:13.760 --> 0:21:16.160 out of a pickup in the front of the school 0:21:16.160 --> 0:21:18.830 and runs into the thing, and he looks like he's 0:21:18.830 --> 0:21:22.310 over 18, but probably 19 or 20, and he's got 0:21:22.310 --> 0:21:26.550 a backpack. Suddenly the earpiece in your kid's ear and 0:21:26.550 --> 0:21:29.129 in the ear of the principal and everyone else. Boom! 0:21:29.160 --> 0:21:32.879 There's data happening. Why? Because there's drones flying over. This 0:21:32.880 --> 0:21:36.840 is a this is a monitored state of the world 0:21:36.840 --> 0:21:40.440 that I parse and hand to the people who care 0:21:40.440 --> 0:21:43.470 about that state. And going back to the first point, 0:21:43.470 --> 0:21:46.830 there is so much you can gather every everything's going 0:21:46.859 --> 0:21:48.419 to have an API. You're going to be able to 0:21:48.450 --> 0:21:51.570 pull all this data. But okay, why am I pulling 0:21:51.570 --> 0:21:54.300 gigabytes of data all the time? Can I even process it? 0:21:54.330 --> 0:21:57.000 The answer is no, you can't, which means we won't 0:21:57.000 --> 0:21:59.070 pull that yet. But that will push us to be 0:21:59.100 --> 0:22:01.350 able to process it. Oh, now we can. Now we'll 0:22:01.350 --> 0:22:04.650 pull more data. This will just keep ratcheting up and 0:22:04.650 --> 0:22:09.540 keep ratcheting up. And until until you eventually have something 0:22:09.540 --> 0:22:13.740 like Neuralink and you just kind of know the state 0:22:13.740 --> 0:22:16.439 the same way you see color. You will know the 0:22:16.440 --> 0:22:20.670 state of danger. Right? And I believe all this is inevitable, 0:22:20.670 --> 0:22:22.710 but it's so far in the future. Like you don't 0:22:22.710 --> 0:22:24.810 need to worry about it. But what I'm trying to 0:22:24.810 --> 0:22:28.500 do is show you that it's obvious that that sort 0:22:28.500 --> 0:22:32.940 of push is happening from humans, because that's what we demand, right? 0:22:32.940 --> 0:22:35.970 So that's kind of where we're going, um, and how 0:22:35.970 --> 0:22:38.250 it will happen and who will do it or whatever. 0:22:38.250 --> 0:22:41.760 Who knows? Nobody knows. Could be a giant single corporation 0:22:41.760 --> 0:22:43.800 that does everything. And or it could be a million 0:22:43.800 --> 0:22:47.580 different startups. Who knows? It's impossible to predict. The point is, 0:22:47.580 --> 0:22:50.340 we will demand this and we will demand it gets 0:22:50.340 --> 0:22:53.520 better and better. And the deeper you go on how 0:22:53.520 --> 0:22:57.990 granular and the more often the updates to the frequency 0:22:57.990 --> 0:23:00.960 are demanded, and the more the cost goes down, the 0:23:00.960 --> 0:23:06.390 more we will have. It's this thing just goes crazy. Okay, 0:23:06.480 --> 0:23:10.020 now we talk about what's the desired state. Now we 0:23:10.020 --> 0:23:12.180 talk about how do we get from here to there? 0:23:12.210 --> 0:23:15.840 What's the planning? Right? How do we make a I'm 0:23:15.840 --> 0:23:17.790 reading good to great right now. How do we make 0:23:17.790 --> 0:23:20.580 this business from good to great. What are the changes 0:23:20.580 --> 0:23:23.790 we make. Right. That's also an AI eye piece. All 0:23:23.790 --> 0:23:26.939 of these, all of these combined are going to push 0:23:26.940 --> 0:23:31.050 this thing from us barely starting this year, last year, 0:23:31.080 --> 0:23:35.160 year before to it's not a hockey stick. It is 0:23:35.160 --> 0:23:38.639 straight up into the sky in how much context we 0:23:38.640 --> 0:23:41.670 need and how much memory we need and how much 0:23:41.670 --> 0:23:46.500 processing we need right now. It could be that new 0:23:46.500 --> 0:23:50.699 processors come out so we could do a 100,000 times 0:23:50.700 --> 0:23:56.699 as much processing, with 100,000 times less effort and resources 0:23:56.700 --> 0:23:59.159 and stuff like that. Doesn't matter. We'll just ask for 0:23:59.160 --> 0:24:02.940 100,000 times more. We'll be like, oh, cool. So so 0:24:02.940 --> 0:24:07.950 you can pull, um, 40,000 metrics on me as a 0:24:07.950 --> 0:24:11.580 human every second. Cool. Can you get the state of 0:24:11.580 --> 0:24:14.400 every molecule in my body? And they'll be like, no, 0:24:14.400 --> 0:24:17.219 that's impossible. Cool. Someone will go work on it. And 0:24:17.220 --> 0:24:20.609 suddenly the current hardware is not good enough. And we 0:24:20.609 --> 0:24:22.810 could make better predictions if we knew the state of 0:24:22.810 --> 0:24:25.630 every molecule in the body. And guess what? It just 0:24:25.630 --> 0:24:28.750 keeps going. It just keeps going. That's the point of this. 0:24:28.750 --> 0:24:32.980 So these are the three basic pieces current state, desired state, 0:24:32.980 --> 0:24:38.830 and the actions, processes and SOPs. SOP is a standard 0:24:38.830 --> 0:24:42.850 operating procedure. I think this is a really, really powerful concept. 0:24:42.850 --> 0:24:46.240 I think these three combined are going to be like 0:24:46.240 --> 0:24:49.840 a framework for making improvements to things. I did a 0:24:49.840 --> 0:24:55.209 piece in early 2023, like March of 2023, called Pspca 0:24:55.240 --> 0:25:00.070 State Policy Questions and Actions. State is state of the world. 0:25:00.070 --> 0:25:04.480 Exactly like this policy is the desired state of your 0:25:04.480 --> 0:25:08.260 company or your life, or whatever questions is like what 0:25:08.260 --> 0:25:12.040 are we asking to this I body of knowledge and 0:25:12.040 --> 0:25:14.409 then action is what do we do as a result 0:25:14.410 --> 0:25:16.750 of learning the answer. Right. So I think that's a 0:25:16.750 --> 0:25:19.850 really powerful structure for thinking about how to improve anything, 0:25:19.850 --> 0:25:22.430 and it's all based on the stuff we've been talking about. 0:25:22.430 --> 0:25:25.879 So some examples of this. A lonely human wants to 0:25:25.910 --> 0:25:29.060 be a happy human. So the recommendation is going to 0:25:29.060 --> 0:25:31.159 be to connect. And all of these are going to 0:25:31.160 --> 0:25:33.979 be extremely granular with lots of different recommendations going all 0:25:33.980 --> 0:25:36.560 the way down. Right. And of course this will all 0:25:36.560 --> 0:25:39.830 be managed by a Da AI of some sort. You've 0:25:39.830 --> 0:25:41.780 got a dying business. You wish you had a thriving 0:25:41.780 --> 0:25:44.030 business that was actually making money. What are you going 0:25:44.030 --> 0:25:46.669 to do? You're going to increase revenue by doing the 0:25:46.670 --> 0:25:50.750 following things. This flow here of understand the current thing, 0:25:50.780 --> 0:25:53.480 understand where you want to be and have AI figure 0:25:53.510 --> 0:25:55.879 out the stuff in the middle. At first it's going 0:25:55.910 --> 0:25:58.909 to be recommending things to us right as it is now, 0:25:58.910 --> 0:26:02.930 but very soon and already starting. It's going to just 0:26:02.930 --> 0:26:04.580 be like, do you want me to do that for you? 0:26:04.580 --> 0:26:07.399 And the businesses are going to be like, yes, absolutely. 0:26:07.400 --> 0:26:09.409 That means I don't have to hire anyone. In fact, 0:26:09.410 --> 0:26:10.609 that means I can get rid of a lot of 0:26:10.609 --> 0:26:13.370 people because you could just do that for me. So 0:26:13.369 --> 0:26:15.800 what are the predictions that we can make based on this? 0:26:15.800 --> 0:26:19.190 I think of the three. The state piece is the 0:26:19.190 --> 0:26:23.780 most important. Just because you can't really do anything unless 0:26:23.780 --> 0:26:28.700 you understand what's currently happening, right? It's hard to it's 0:26:28.700 --> 0:26:32.270 hard to talk about improvement if you don't know what's wrong. Um, 0:26:32.270 --> 0:26:34.760 so state of the human state of a company. State 0:26:34.760 --> 0:26:37.250 of a family. Cup of coffee. We've talked about this 0:26:37.280 --> 0:26:42.229 infinitely complex changing constantly, and that's why the stuff is 0:26:42.230 --> 0:26:45.650 going to scale so crazily, in my opinion. And who 0:26:45.650 --> 0:26:48.170 knows if Nvidia is going to win. I think they'll 0:26:48.170 --> 0:26:50.450 probably win for at least a couple of years, but 0:26:50.450 --> 0:26:54.080 I have no idea they could crash tomorrow. Or they 0:26:54.080 --> 0:26:57.980 could whatever, go to the stratosphere in six months? No idea. 0:26:58.010 --> 0:27:01.370 So this this brings us to the question how much 0:27:01.369 --> 0:27:04.189 context size do we need? How many tokens do we 0:27:04.220 --> 0:27:06.380 need to be able to put into context? How big 0:27:06.380 --> 0:27:09.379 are Rag systems going to get? How many GPUs do 0:27:09.380 --> 0:27:12.230 we need? How many startups do we need in the 0:27:12.230 --> 0:27:16.280 AI space? How much I do we actually need? That 0:27:16.280 --> 0:27:18.710 is ultimately what we're trying to answer here. And my 0:27:18.710 --> 0:27:22.580 answer to that is enough to constantly poll the state 0:27:22.580 --> 0:27:26.389 of everything we care about, and then take actions to 0:27:26.420 --> 0:27:29.119 fix it to get it into the desired state. And 0:27:29.119 --> 0:27:34.070 I think that is a metric crap ton of AI. 0:27:34.070 --> 0:27:36.680 And that's why I think we're at like, like I said, 0:27:36.710 --> 0:27:40.640 point ten zeros and a one on where it's actually going. 0:27:40.670 --> 0:27:44.090 See you in the next one. Unsupervised learning is produced 0:27:44.090 --> 0:27:47.150 and edited by Daniel Miessler on a Neumann U87 AI 0:27:47.150 --> 0:27:51.380 microphone using Hindenburg. Intro and outro music is by Zomby 0:27:51.380 --> 0:27:54.380 with a Y, and to get the text and links 0:27:54.380 --> 0:27:56.629 from this episode, sign up for the newsletter version of 0:27:56.630 --> 0:28:02.270 the show at Daniel missler.com/newsletter. We'll see you next time.