-
3:41 · watch on youtube.com↗
Their wordsthere is this thing where effect where people who are optimistic people who are working on the technology tend to underestimate the time it takes to get there.
- 1 min later
-
4:12 · watch on youtube.com↗
Their wordsAnd we might be in a similar place with respect to our models where it also looks like we can do everything. And at the same time, it will be, we'll need to do some more work until we really iron out all the issues and make it really good and really reliable and robust and well -behaved.
- 1 min later
-
5:36 · watch on youtube.com↗
Their wordsit would be reliability that you still have to look over the answers and double check everything. And that's just really puts a damper on the economic value that can be produced by those systems.
-
6:05 · watch on youtube.com↗
Their wordsI think this paradigm is going to go really, really far and I would not underestimate it. I think it's quite likely that this exact paradigm is not going to be the quite AGI form factor.
- 1 min later
-
7:15 · watch on youtube.com↗
Their wordsMaybe such a person doesn't exist, but there's a pretty good chance that the neural net will be able to extrapolate how such a person could behave.
-
7:39 · watch on youtube.com↗
Their wordsPredicting the next token well means that you understand the underlying reality that led to the creation of that token.
- 1 min later
-
8:58 · watch on youtube.com↗
Their wordsI mean, already most of the data for reinforcement learning is coming from AIs.
- 1 min later
-
10:04 · watch on youtube.com↗
Their wordsyou might want to think of it as being in a world where the human teachers do 1 % of the world work and the AI do 99 % of the work. You don't want it to be 100 % AI, but you do want it to be a human -machine collaboration which teaches the next machine.
- 1 min later
-
10:41 · watch on youtube.com↗
Their wordsI actually think that they are bad at mental multi -step reasoning, but they're not allowed to think out loud. But when they are allowed to think out loud, they're quite good.
- 1 min later
-
11:30 · watch on youtube.com↗
Their wordsI would say the data situation is still quite good. There are still lots to go. But at some point, yeah, at some point data will run out.
- 1 min later
-
12:04 · watch on youtube.com↗
Their wordsI think that you can still go very far in text only, but going multimodal seems like a very fruitful direction.
- 2 min later
-
13:53 · watch on youtube.com↗
Their wordsI'd say that now it is possible to create a path forward, but one needs to really commit to the task of robotics. You really need to say, I'm going to build like many thousands, tens of thousands, hundreds of thousands of robots and somehow collect data from them
- 2 min later
-
15:41 · watch on youtube.com↗
Their wordsrather than achieving one mathematical definition, I think we'll achieve multiple definitions that look at alignment from different aspects. And I think that this is how we will get the assurance that we want.
- 2 min later
-
17:22 · watch on youtube.com↗
Their wordsEvery approach like this reduces the probability of misalignment. And you also want to be in a world where your degree of alignment keeps increasing faster than the capability of the models.
-
17:46 · watch on youtube.com↗
Their wordsI would expect that ultimately the thing that will really succeed is when we will have a small neural net that is well understood, that's given the task to study the behavior of a large neural net that is not understood to verify it.
- 1 min later
-
18:34 · watch on youtube.com↗
Their wordsI think the bottleneck there is good ideas, good insights, and that's something which the neural net could help with this.
- 3 min later
-
21:24 · watch on youtube.com↗
Their wordswe'll all be able to become more enlightened, because we'd interact with an AGI that will help us see the world more correctly, but become better on the inside as a result of interaction.
- 2 min later
-
23:29 · watch on youtube.com↗
Their wordsI'd much rather have a world where people are still free to make their own mistakes and suffer their consequences and gradually evolve morally and progress forward on their own, through their own strength.
- 3 min later
-
26:33 · watch on youtube.com↗
Their wordsSo I really feel like hardware, like in some sense, the only thing that matters about hardware is cost, cost per flop, overall systems cost.
- 1 min later
-
27:17 · watch on youtube.com↗
Their wordsCertainly coming up with new ideas is important, but I think even more important is to understand the results, to understand the existing ideas, to understand what's going on.
- 3 min later
-
30:06 · watch on youtube.com↗
Their wordsInference of better models will indeed become more expensive. But is it prohibitive? Well, it depends on how useful is it. Like if it is more useful than it is expensive, then it is not prohibitive.
- 4 min later
-
34:24 · watch on youtube.com↗
Their wordsThe thing which I'm really excited about or the thing which I'd like to see is reliability and controllability. I think that this will be a very, very important class of emergent properties.
- 1 min later
-
35:32 · watch on youtube.com↗
Their wordsThere is a whole separate challenge of linking next word prediction accuracy to reasoning capability. I do believe that, indeed, there is a link, but this link is complicated.
- 3 min later
-
38:08 · watch on youtube.com↗
Their wordsAll this progress in all these dimensions is very intertwined. It's not a coincidence that you don't get to pick and choose in which dimensions things improve, if you see what I mean.
- 2 min later
-
39:40 · watch on youtube.com↗
Their wordsI think with the current level of capabilities, I think we have a pretty good set of ideas of how to align them. But I would not underestimate the difficulty of alignment of models that are actually smarter than us, of models that are capable of misrepresenting their intentions.
-
40:08 · watch on youtube.com↗
Their wordsAnd I think alignment research is one place where I think academic researchers can make very meaningful contributions.
-
40:31 · watch on youtube.com↗
Their wordsI don't think there's anything fundamental about academia. Like it's not like academia can't. I think maybe they're just not thinking about the right problems or something because maybe it's just easier to see what needs to be done inside these companies.
- 3 min later
-
43:53 · watch on youtube.com↗
Their wordsBut if you are trying to just engineer a good system, there is no reason to not use backpropagation.
- 1 min later
-
44:43 · watch on youtube.com↗
Their wordsI think there is an art into being inspired by humans in the brain correctly, because it's very easy to latch on to a non -essential quality of humans or of the brain.
- 1 min later
-
46:12 · watch on youtube.com↗
Their wordsI think it's a necessary, but not a sufficient condition. Like, you know, many things need to come together in order to really figure something out. Like you need to, to really go for it and also need to have the right way of looking at things.