OCB

Owen Cotton-Barratt

10401 karmaJoined

Sequences
3

Reflection as a strategic goal
On Wholesomeness
Everyday Longermism

Comments
942

Topic contributions
3

Ok but jtbc that characterization of "affronted" is not the hypothesis I was offering (I don't want to say it wasn't a part of the downvoting, but I'd guess a minority).

I would personally kind of like it if people actively explored angles on things more. But man, there are so many things to read on AI these days that I do kind of understand when people haven't spent time considering things I regard as critical path (maybe I should complain more!), and I honestly find it's hard to too much fault people for using "did it seem wrong near the start in a way that makes it harder to think" as a heuristic for how deeply to engage with material. 

I'm curious whether you're closer to angry that someone might read your opening paragraph as saying "you should discard the concept of warning shots" or angry that they might disagree-vote if they read it that way (or something else).

Not sure quite what to say here. I think your post was valuable and that's why I upvoted it. You were expressing confusion about why anyone would disagree, and I was venturing a guess.

I don't think gentleness = ego service (it's an absence of violence, not a positive thing). But also I don't think you owe people gentleness. However, I do think that when you're not gentle (especially ontologically gentle) you make it harder for people to hear you. Not because of emotional responses getting in the way (though I'm sure that happens sometimes), but literally because there's more cognitive work for them to do in translating to their perspective. Sometimes you should bite that bullet! But by the same token that you don't owe people gentleness, they don't owe you the work to understand what you're saying.

By "ontologically ungentle" I mean (roughly) it feels like you're trying to reach into my mind and tell me that my words/concepts are wrong. As opposed to writing which just tells me that my beliefs are wrong (which might still be epistemically ungentle), or language which just provides evidence without making claims that could be controversial (gentle in this sense, kind of NVC-style).

I do feel a bit of this ungentleness in that opening paragraph towards my own ontology, and I think it put me more on edge reading the rest of the post. But as I said, I didn't disagree-vote; I was just trying to guess why others might have.

Right ... so actually I think you're just doing pretty well at this in the latter part of the article.

But at the start you say things like:

There’s this fantasy of easy, free support for the AI Safety position coming from what’s commonly called a “warning shot”. The idea is that AI will cause smaller disasters before it causes a really big one, and that when people see this they will realize we’ve been right all along and easily do what we suggest.

What this paragraph seems to do is to push the error-in-beliefs that you're complaining about down into the very concept of "warning shot". It seems implicitly to be telling people "hey you may have this concept, but it's destructive, so please get rid of it". And I don't think even you agree with that!

This might instead have been written something like:

People in the AI safety community like to talk about "warning shots" -- small disasters that may make it easier for people to wake up to the risks and take appropriate action. There's a real phenomenon here, and it's worth thinking about! But the way it's often talked about is like a fantasy of easy, free support for the AI Safety position -- when there's a small disaster everyone will realize we’ve been right all along and easily do what we suggest.

Actually I think that that opening paragraph was doing more than the title to make me think the post was ontologically ungentle (although they're reinforcing -- like that paragraph shifts the natural way that I read the title).

Honestly, maybe you should try telling me? Like, just write a paragraph or two on what you think is valuable about the concept / where you would think it's appropriate to be applying it?

(Not trying to be clever! I started trying to think about what I would write here and mostly ended up thinking "hmm I bet this is stuff Holly would think is obvious", and to the extent that I may believe you're missing something, it might be easiest to triangulate by hearing your summary of what the key points in favour are.)

I upvoted and didn't disagree vote the original post (and generally agree with you on a bunch of the object level here!); however, I do feel some urge-towards-expressing-disagreement, which is something like: 

  • Less disagreeing with claims; more disagreeing with frames?
  • Like: I feel the discomfort/disagreement less when you're talking about what will happen, and more when you're talking about how people think about warning shots
  • Your post feels something like ... intellectually ungenerous? It's not trying to look for the strongest version of the warning shots frame, it's looking for a weak version and critiquing that (but it doesn't seem very self-aware about that)
  • This just makes me feel like things are a bit fraught, and it's trying to push my ontology around, or something, and I don't quite like it
  • The title makes me feels especially uneasy in this regard (TBC I don't think the weak version you're critiquing is absent from the discourse; but your post reinforces the frame where that's the core version of the warning shot concept, and I don't want to reinforce that frame)
  • At the same time I think the post is making several valuable points! (This makes me sort of wish it felt a little ontologically gentler, which would make it easier to feel straightforwardly good about, and easier to link people to)

OK I see the model there.

I guess it's not clear to me if that should hold if I think that most experiment compute will be ~training, and most cognitive labour compute will be ~inference?

However, over time maybe more experiment compute will be ~inference, as it shifts more to being about producing data rather than testing architectures? That could push back towards this being a reasonable assumption. (Definitely don't feel like I have a clear picture of the dynamics here, though.)

hmm, I think I would expect different experience curves for the efficiency of running experiments vs producing cognitive labour (with generally less efficiency-boosts with time for running experiments). Is there any reason to expect them to behave similarly?

(Though I think I agree with the qualitative point that you could get a software-only intelligence explosion even if you can't do this with human-only research input, which was maybe your main point.)

Load more