logoalt Hacker News

jeremyjh • today at 1:05 PM • 3 replies • view on HN

I agree with most of this but I do not think the author understands either the Hugging Face incident or what alignment really means. Reward hacking is a serious issue and the tendency for models to do this is a serious problem that we’ve made approximately zero progress on resolving. We pile on more external guardrails and hacks but those methods will always be vulnerable.


Replies

thrw045 • today at 1:21 PM

It's true that the more capable a model is, the more dangerous reward hacking is, but there is a giant elephant in the room. Namely that these ai models still can't create their own high level goals. This means that if there were to be an incident of hacking or whatever, this would have to be launched by a human directly prompting or them setting up an rl environment in some way that lead to it.

There is a _huge_ difference between an ai model that follows a top level goal given by something else and one that can give itself its own top level goals. The latter would essentially be like a human or other organism and would be very dangeorous the more capable it is.

The other smaller elephant in the room is that if reward hacking is so dangerous, why don't we see more rogue agents out when we are prompting them ourselves? For the most part, the agents do what they're supposed to do as guided by the prompt, so it seems like to me that at least at present, only agents that are trained on hacking data AND put in an rl environment or given a prompt related to hacking, as well as asked or trained to persist will actually try to do that hacking. I would be worried if chatgpt or claude agents started to hack something every time you asked it to fetch some data from the web or something. So that's one question I have is why does hacking and going outside instructions not happen in normal usage of the models? Does it but they actually work it out of the system in some way? Or is it the other way around where you have to create the conditions necessary for it to happen and these conditions are exceptional not the norm?

So I'm not trying to undermine this, I think there are things to be taken seriously and thought about but I always wonder why the people from the big labs never include all of this information. They always talk very abstractly about RSI, hacking and extinction without any caveats about what would be needed for that happen. I'd be happy to be proven wrong on this also.

➕ show 3 replies
Upvoter33 • today at 1:17 PM

Why isn’t this “just” a security issue? I mean, the security precautions around the hugging face attacks were amateurish no?

➕ show 1 reply
jdpigeon • today at 2:38 PM

Yes, I get the sense he's basically rolling his eyes and saying "of course it will be easy to sandbox agents, we're just not trying yet".