id
int64 1
325
| guest
stringlengths 3
45
⌀ | title
stringlengths 3
75
⌀ | text
stringlengths 1
528
| start
stringlengths 9
11
| end
stringlengths 9
11
|
|---|---|---|---|---|---|
1 | Max Tegmark | Life 3.0 | which is way more than there are atoms in our universe, | 1:02:36.960 | 1:02:38.960 |
1 | Max Tegmark | Life 3.0 | right, so in order to, | 1:02:38.960 | 1:02:41.000 |
1 | Max Tegmark | Life 3.0 | and then for each one of those, | 1:02:42.040 | 1:02:43.200 |
1 | Max Tegmark | Life 3.0 | I have to assign a number, | 1:02:43.200 | 1:02:44.640 |
1 | Max Tegmark | Life 3.0 | which is the probability that it's a dog. | 1:02:44.640 | 1:02:47.080 |
1 | Max Tegmark | Life 3.0 | So an arbitrary function of images | 1:02:47.080 | 1:02:49.440 |
1 | Max Tegmark | Life 3.0 | is a list of more numbers than there are atoms in our universe. | 1:02:49.440 | 1:02:54.440 |
1 | Max Tegmark | Life 3.0 | So clearly I can't store that under the hood of my GPU | 1:02:54.440 | 1:02:57.360 |
1 | Max Tegmark | Life 3.0 | or my computer, yet somehow it works. | 1:02:57.360 | 1:03:00.640 |
1 | Max Tegmark | Life 3.0 | So what does that mean? | 1:03:00.640 | 1:03:01.480 |
1 | Max Tegmark | Life 3.0 | Well, it means that out of all of the problems | 1:03:01.480 | 1:03:04.960 |
1 | Max Tegmark | Life 3.0 | that you could try to solve with a neural network, | 1:03:04.960 | 1:03:08.200 |
1 | Max Tegmark | Life 3.0 | almost all of them are impossible to solve | 1:03:10.120 | 1:03:12.880 |
1 | Max Tegmark | Life 3.0 | with a reasonably sized one. | 1:03:12.880 | 1:03:14.560 |
1 | Max Tegmark | Life 3.0 | But then what we showed in our paper | 1:03:15.480 | 1:03:17.440 |
1 | Max Tegmark | Life 3.0 | was that the fraction, the kind of problems, | 1:03:17.440 | 1:03:22.360 |
1 | Max Tegmark | Life 3.0 | the fraction of all the problems | 1:03:22.360 | 1:03:23.800 |
1 | Max Tegmark | Life 3.0 | that you could possibly pose, | 1:03:23.800 | 1:03:26.520 |
1 | Max Tegmark | Life 3.0 | that we actually care about given the laws of physics | 1:03:26.520 | 1:03:29.480 |
1 | Max Tegmark | Life 3.0 | is also an infinite testimony, tiny little part. | 1:03:29.480 | 1:03:32.480 |
1 | Max Tegmark | Life 3.0 | And amazingly, they're basically the same part. | 1:03:32.480 | 1:03:35.440 |
1 | Max Tegmark | Life 3.0 | Yeah, it's almost like our world was created for, | 1:03:35.440 | 1:03:37.560 |
1 | Max Tegmark | Life 3.0 | I mean, they kind of come together. | 1:03:37.560 | 1:03:39.000 |
1 | Max Tegmark | Life 3.0 | Yeah, well, you could say maybe where the world was created | 1:03:39.000 | 1:03:42.800 |
1 | Max Tegmark | Life 3.0 | for us, but I have a more modest interpretation, | 1:03:42.800 | 1:03:44.960 |
1 | Max Tegmark | Life 3.0 | which is that the world was created for us, | 1:03:44.960 | 1:03:46.680 |
1 | Max Tegmark | Life 3.0 | but I have a more modest interpretation, | 1:03:46.680 | 1:03:48.040 |
1 | Max Tegmark | Life 3.0 | which is that instead evolution endowed us | 1:03:48.040 | 1:03:50.360 |
1 | Max Tegmark | Life 3.0 | with neural networks precisely for that reason. | 1:03:50.360 | 1:03:53.120 |
1 | Max Tegmark | Life 3.0 | Because this particular architecture, | 1:03:53.120 | 1:03:54.640 |
1 | Max Tegmark | Life 3.0 | as opposed to the one in your laptop, | 1:03:54.640 | 1:03:56.040 |
1 | Max Tegmark | Life 3.0 | is very, very well adapted to solving the kind of problems | 1:03:56.040 | 1:04:01.040 |
1 | Max Tegmark | Life 3.0 | that nature kept presenting our ancestors with. | 1:04:02.480 | 1:04:05.560 |
1 | Max Tegmark | Life 3.0 | So it makes sense that why do we have a brain | 1:04:05.560 | 1:04:08.120 |
1 | Max Tegmark | Life 3.0 | in the first place? | 1:04:08.120 | 1:04:09.280 |
1 | Max Tegmark | Life 3.0 | It's to be able to make predictions about the future | 1:04:09.280 | 1:04:11.880 |
1 | Max Tegmark | Life 3.0 | and so on. | 1:04:11.880 | 1:04:12.880 |
1 | Max Tegmark | Life 3.0 | So if we had a sucky system, which could never solve it, | 1:04:12.880 | 1:04:16.440 |
1 | Max Tegmark | Life 3.0 | we wouldn't have a world. | 1:04:16.440 | 1:04:18.280 |
1 | Max Tegmark | Life 3.0 | So this is, I think, a very beautiful fact. | 1:04:18.280 | 1:04:23.280 |
1 | Max Tegmark | Life 3.0 | Yeah. | 1:04:23.680 | 1:04:24.520 |
1 | Max Tegmark | Life 3.0 | We also realize that there's been earlier work | 1:04:24.520 | 1:04:29.000 |
1 | Max Tegmark | Life 3.0 | on why deeper networks are good, | 1:04:29.000 | 1:04:32.040 |
1 | Max Tegmark | Life 3.0 | but we were able to show an additional cool fact there, | 1:04:32.040 | 1:04:34.680 |
1 | Max Tegmark | Life 3.0 | which is that even incredibly simple problems, | 1:04:34.680 | 1:04:38.360 |
1 | Max Tegmark | Life 3.0 | like suppose I give you a thousand numbers | 1:04:38.360 | 1:04:41.080 |
1 | Max Tegmark | Life 3.0 | and ask you to multiply them together, | 1:04:41.080 | 1:04:42.720 |
1 | Max Tegmark | Life 3.0 | and you can write a few lines of code, boom, done, trivial. | 1:04:42.720 | 1:04:46.680 |
1 | Max Tegmark | Life 3.0 | If you just try to do that with a neural network | 1:04:46.680 | 1:04:49.520 |
1 | Max Tegmark | Life 3.0 | that has only one single hidden layer in it, | 1:04:49.520 | 1:04:52.440 |
1 | Max Tegmark | Life 3.0 | you can do it, | 1:04:52.440 | 1:04:53.400 |
1 | Max Tegmark | Life 3.0 | but you're going to need two to the power of a thousand | 1:04:54.320 | 1:04:57.360 |
1 | Max Tegmark | Life 3.0 | neurons to multiply a thousand numbers, | 1:04:57.360 | 1:05:00.920 |
1 | Max Tegmark | Life 3.0 | which is, again, more neurons than there are atoms | 1:05:00.920 | 1:05:02.520 |
1 | Max Tegmark | Life 3.0 | in our universe. | 1:05:02.520 | 1:05:03.360 |
1 | Max Tegmark | Life 3.0 | That's fascinating. | 1:05:04.600 | 1:05:05.480 |
1 | Max Tegmark | Life 3.0 | But if you allow yourself to make it a deep network | 1:05:05.480 | 1:05:09.960 |
1 | Max Tegmark | Life 3.0 | with many layers, you only need 4,000 neurons. | 1:05:09.960 | 1:05:13.240 |
1 | Max Tegmark | Life 3.0 | It's perfectly feasible. | 1:05:13.240 | 1:05:14.520 |
1 | Max Tegmark | Life 3.0 | That's really interesting. | 1:05:16.400 | 1:05:17.960 |
1 | Max Tegmark | Life 3.0 | Yeah. | 1:05:17.960 | 1:05:18.800 |
1 | Max Tegmark | Life 3.0 | So on another architecture type, | 1:05:18.800 | 1:05:21.040 |
1 | Max Tegmark | Life 3.0 | I mean, you mentioned Schrodinger's equation, | 1:05:21.040 | 1:05:22.720 |
1 | Max Tegmark | Life 3.0 | and what are your thoughts about quantum computing | 1:05:22.720 | 1:05:26.360 |
1 | Max Tegmark | Life 3.0 | and the role of this kind of computational unit | 1:05:27.240 | 1:05:32.240 |
1 | Max Tegmark | Life 3.0 | in creating an intelligence system? | 1:05:32.400 | 1:05:34.880 |
1 | Max Tegmark | Life 3.0 | In some Hollywood movies that I will not mention by name | 1:05:34.880 | 1:05:39.520 |
1 | Max Tegmark | Life 3.0 | because I don't want to spoil them. | 1:05:39.520 | 1:05:41.040 |
1 | Max Tegmark | Life 3.0 | The way they get AGI is building a quantum computer. | 1:05:41.040 | 1:05:44.240 |
1 | Max Tegmark | Life 3.0 | Because the word quantum sounds cool and so on. | 1:05:45.480 | 1:05:47.600 |
1 | Max Tegmark | Life 3.0 | That's right. | 1:05:47.600 | 1:05:48.440 |
1 | Max Tegmark | Life 3.0 | First of all, I think we don't need quantum computers | 1:05:50.040 | 1:05:52.880 |
1 | Max Tegmark | Life 3.0 | to build AGI. | 1:05:52.880 | 1:05:54.920 |
1 | Max Tegmark | Life 3.0 | I suspect your brain is not a quantum computer | 1:05:54.920 | 1:05:59.240 |
1 | Max Tegmark | Life 3.0 | in any profound sense. | 1:05:59.240 | 1:06:00.640 |
1 | Max Tegmark | Life 3.0 | So you don't even wrote a paper about that | 1:06:01.600 | 1:06:03.200 |
1 | Max Tegmark | Life 3.0 | a lot many years ago. | 1:06:03.200 | 1:06:04.560 |
1 | Max Tegmark | Life 3.0 | I calculated the so called decoherence time, | 1:06:04.560 | 1:06:08.120 |
1 | Max Tegmark | Life 3.0 | how long it takes until the quantum computerness | 1:06:08.120 | 1:06:10.320 |
1 | Max Tegmark | Life 3.0 | of what your neurons are doing gets erased | 1:06:10.320 | 1:06:13.400 |
1 | Max Tegmark | Life 3.0 | by just random noise from the environment. | 1:06:15.320 | 1:06:17.960 |
1 | Max Tegmark | Life 3.0 | And it's about 10 to the minus 21 seconds. | 1:06:17.960 | 1:06:21.320 |
1 | Max Tegmark | Life 3.0 | So as cool as it would be to have a quantum computer | 1:06:21.320 | 1:06:24.600 |
1 | Max Tegmark | Life 3.0 | in my head, I don't think that fast. | 1:06:24.600 | 1:06:27.320 |
1 | Max Tegmark | Life 3.0 | On the other hand, | 1:06:27.320 | 1:06:28.360 |
1 | Max Tegmark | Life 3.0 | there are very cool things you could do | 1:06:28.360 | 1:06:33.040 |
1 | Max Tegmark | Life 3.0 | with quantum computers. | 1:06:33.040 | 1:06:34.200 |
1 | Max Tegmark | Life 3.0 | Or I think we'll be able to do soon | 1:06:35.240 | 1:06:37.480 |
1 | Max Tegmark | Life 3.0 | when we get bigger ones. | 1:06:37.480 | 1:06:39.360 |
1 | Max Tegmark | Life 3.0 | That might actually help machine learning | 1:06:39.360 | 1:06:40.960 |
1 | Max Tegmark | Life 3.0 | do even better than the brain. | 1:06:40.960 | 1:06:43.160 |
1 | Max Tegmark | Life 3.0 | So for example, | 1:06:43.160 | 1:06:45.640 |
1 | Max Tegmark | Life 3.0 | one, this is just a moonshot, | 1:06:47.040 | 1:06:50.760 |
1 | Max Tegmark | Life 3.0 | but learning is very much same thing as search. | 1:06:50.760 | 1:06:55.760 |
1 | Max Tegmark | Life 3.0 | If you're trying to train a neural network | 1:07:01.800 | 1:07:03.160 |
1 | Max Tegmark | Life 3.0 | to get really learned to do something really well, | 1:07:03.160 | 1:07:06.240 |
1 | Max Tegmark | Life 3.0 | you have some loss function, | 1:07:06.240 | 1:07:07.280 |
1 | Max Tegmark | Life 3.0 | you have a bunch of knobs you can turn, | 1:07:07.280 | 1:07:10.360 |
1 | Max Tegmark | Life 3.0 | represented by a bunch of numbers, | 1:07:10.360 | 1:07:12.080 |
1 | Max Tegmark | Life 3.0 | and you're trying to tweak them | 1:07:12.080 | 1:07:12.920 |