| Not checkable |
Not checkable |
12 Apr 2024 |
简单来说,我不确定,也没有确切的答案,但无论如何我们都需要尝试。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我偶尔会幻想,几年后我们可能还要担心有人利用 AI 系统创立一个宗教或什么类似的东西。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
但我后来发现,我仍然需要对城市及其环境有所了解,只是这些知识被转移到了大脑的其他区域处理。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我内心有些疑问——虽然不确定。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我不相信任何一个系统能完全模仿人类的行为方式。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
或许这听起来有点乐观,但你总会找到一种工作流程,使这些工具能够辅助你。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
因此,我这里只能说一些老生常谈:学会使用这项技术。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
教育你的孩子要有应变能力,做好迎接快速变化的世界的准备。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
所以我认为这是一个我们在历史上可能已经见过几次的现象,就是存在一个基本过程,它本质上是平滑的,这里则是指数增长。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我认为,对当前经济中断的唯一合理反应就是探索新的经济结构,这种结构在初期可能不明显,但将随时间不断发展。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我的第二个推荐是科幻系列《星际迷航》。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我真的相信指数增长。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
对于这些问题,我也没有确切的答案,我无法提供具体的解决方案,但我们必须努力找到解决的方法。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
但公司可以尝试自行开发对社会有益的应用。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
这也许看起来太过乐观了。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我认为通过灾难来学习是最糟糕的方式。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
一个基本建议是,尽可能地了解技术及其工作原理。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我并不认为挖掘比特币所消耗的能量有任何附加价值,这种能量在之前是不存在的。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我希望能有更好的答案,但这已是我所能想到的最好的建议了。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我觉得这也是 AI 与社会互动复杂性的一个体现:有时候我们很难分辨 AI 到底是在帮我们减轻重复性劳动,还是在介入更有趣的活动部分。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我认为,不仅仅是监管问题,整个技术的监督和使用权最终掌握在私人手中,这在某种程度上是不民主的,虽然现在看来可能尚可,但终究不应该这样。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我发现,经常有一些技术产品或系统,你原以为它们只是执行你的核心任务,但实际上,这些任务包含的内容远比你想象的要丰富,功能也更多。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
再次强调,如果这些系统变得足够通用和强大,我们可能需要重新考虑我们的策略。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我们正在尽力投入更多资源来研究模型的可解解性。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
如果任何步骤的准确性不高,例如不到 99.9%,这些步骤的组合错误率会非常高。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
我个人认为,这项技术的优点会超过其代价。
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
12 Apr 2024 |
现在有了谷歌地图 (Google Maps),我开始担心自己是否过于依赖它?
Not checkable · 12 Apr 2024 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
There needs to be some kind of legitimate process for managing this technology, which includes the role of the people building it, includes…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
We're already at the point where to me the things the models can't do don't seem to be different in kind from the things they can do.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I'm definitely a fan of Szilard for having kind of figured it out.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I suspect that it may roughly work to think of the model as if it's trained in the normal way, just getting to above human level.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think we are at a very high standard of security compared to the size of company that we are.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I just tried the simplest experiments.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
It's all costs and benefits.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
For two models to have a debate that is coherent enough that a human can judge it so that the training process can actually work, you need…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
it's the closest we're going to get to something that isn't actively optimizing against us.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
That's a metaphor.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
In a sense, we're a little bit lucky that predict the next word gets us all these other things we need.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
But what we know is it's on a metaphorical exponential, and it's going to happen fast.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
And the trade offs are more positive than they appear because of a phenomenon that I've called Race to the Top.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
One of the reasons why I'm sort of very empiricist about AI, about safety, about organizations, is that you often get surprised.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I would be kind of worried.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Even money changing hands is not quite the same thing as economic value being created.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think in principle it's possible that mechanistic interpretability could be helpful with capabilities.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I want people to think in terms of the nameless, bureaucratic institution and its incentives more than they think in terms of me.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Certainly my track record on predicting these things is terrible but also looking around, I don't really see anyone whose track record is…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
If you're a 1000 person company and everyone knows every secret, one, I guarantee you have a leaker and two, I guarantee you have a spy.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I don't think we should just hand the model over to the UN or whoever happens to be in office at a given time.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I don't think we're aligned by default, I don't think we're doomed by default and have some problem we need to solve.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
For me, the big realization on what intelligence is came with the blob of compute thing.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Now that may not be the threshold where the models are existentially dangerous.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
We should be planning for success not for failure.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
It seems to me at least that the evidence is more consistent with, when something is really high enough value, then someone acts and it's…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think the LTBT (Long Term Benefit Trust) is the right thing on this.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Clearly there's certainly a large amount of overlap because a lot of the things these models do have business applications and many of…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
By the way I'd encourage all the other companies to do this as well.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Whenever you do something at a scale where it's never been done before, every single component, every single thing has to be done in a new…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I definitely think that things like alignment and values are not guaranteed to emerge with scale.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Everyone wants a friendly face, but actually, friendly faces can be misleading.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I don't think of this as binary. It works or it does not work.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I've actually always been a fan of studying these circuits at the lowest level of detail that we possibly can.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I feel like I've been right about some things but still with these theoretical pictures ahead, been wrong about most things.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Like the models learn a physical model of the world to some extent, but they certainly don't learn how to actually move around in the world.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
To get this right in the service of predicting the next word the models are going to have to solve all these theory of mind problems, solve…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Any plan that actually succeeds, regardless of how hard misalignment is to solve, is going to need to solve misuse as well as misalignment.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
It's a less plausible case now.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
The main way to control that is to make sure that the capabilities of the model that we test are not such that they're capable of doing…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I'm skeptical that the value is all that high or the learning can be fast enough to really be in favor of the task.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
The model's constitution should be very simple. It should only have very basic facts that everyone would agree on.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Basically, it's posing to the model the equivalent of these developmental tests that get posed to children.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
But I wouldn't say it's 100%. It could be 50-50.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
And I think RNNs and LSTMs artificially close things off because they close you off to the distant past.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
It's not one shot, it's a long process.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I don't think so at all.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
That was one area where actually I was not prescient and I was surprised as well.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Can't comment on anyone else's security, don't know what's going on in there.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
So it turns out that intelligence isn't a spectrum.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
The very premise of misalignment means that we should be worried about misuse as well, with similar levels of consequences.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
attaching your incentives very strongly to the approval or cheering of a crowd can destroy your mind, and in some cases, it can destroy…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think what's more relevant at this point is just measuring the abilities of the model and seeing how far they are from humans, and they…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
Those words, at least, were the right attitude.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I just feel like you discover ten new ideas and ten new ways that things are going to go wrong by trying these in practice.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
There's kind of no limit to it.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think we're never going to have a guarantee, but some process where we do those things together.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think right now there's way too many assumptions, there's way too much overconfidence about how all this is going to go.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
This was almost like beginner's luck.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
The question I'm much more interested in is, what could we learn that shifts probability mass between them?
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
A lot of them would have happened anyway.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think compartmentalization is the best way to do it.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I actually think often people are thinking about alignment in the wrong way.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
A second thing that seems to be the case is that we seem to be bad at controlling the models.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I aspire to this and I'm proud of this.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I'm not super into questions like what's your probability distribution of X?
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I think today's models just probably aren't smart enough that we should worry about this too much
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
One of the things I think is very important is we should never train for interpretability because that's taking away that advantage.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I used to think that we didn't have to worry about this at all until models were operating in rich environments, like not necessarily…
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I genuinely don't know.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
It's very easy to come up with these methods.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I just don't think that ends well.
Not checkable · 8 Aug 2023 |
|
| Not checkable |
Not checkable |
8 Aug 2023 |
I don't know if their experience is positive or negative.
Not checkable · 8 Aug 2023 |
|