I suggested a similar approach in deconfusing logical counterfactuals where we erase information about the agent so that we end up with multiple possible agents, though I wouldn’t be surprised if other people have also tried asking their agent to ask about how other agents reason. Your approach is different in that the original agent isn’t included among the set of agents considered and that seems like a useful adaption that I hadn’t considered as long as we can provide appropriate justification. I also provide some more justification as to why this is a reasonable thing to do.
Anyway, it’s good to see someone else thinking along a similar (but different) track and I’d be curious to hear what you think about my approach.
“You should be extremely cautious about apologising because that’s what you think a nice person would do”—I edited in a “when it’s” to make it clearer. I only suggesting being careful about apologising based on a particular motivation.
Very glad to hear that it was a success. What component did people find most valuable? I’m particularly interested in techniques that don’t overlap CFAR.
I find it strange to say that we don’t have any plan. Surely the government could set up scholarships or research institute or some kind of committee to look into this?
Would having this information actually be beneficial. Perhaps it’d be good for us to know what is going on, but it might be negative for certain governments to know about this as it might increase the chance of an AI arms race.
I’d suggest initially making short-form a seperate section of the site as my suspicion that if it really is a compelling feature it should be able to succeed on its own without homepage integration. Otherwise, it likely doesn’t provide enough value to make up for the loss of nuance.
Submission low bandwidth: This is a pretty obvious one, but: Should we release AI x that we’re convinced is aligned?
Submission: Wei Dai wanted to ask about the best future posts. Why not ask about the best past posts as well to see if any major insights were overlooked?
Submission: What would I think about problem X if I had ten years to think about it?
What if another AI would have counterfactually written some of those posts to manipulate us?
Well, you don’t have to answer this now. You’ll probably have a better idea once you’ve promoted a few more people.
To be honest, I’m not happy with my response here. There was also a second simultaneous discussion topic about whether CEA was net positive and even though I tried simplifying this into a single discussion, it seems that I accidentally mixed in part of the other discussion (the original title of this post in draft was EA vs. rationality).
Update: I’ve now edited this response out.
I didn’t systematically review his work, just clicked on random articles to see how much value I could extract. Feel free to look me to any reasonable accessible articles.
Thanks for writing this; I’ve briefly attempted looking at his ideas, but most of it is unreadable. Most of his remaining ideas seem at least somewhat mystical, which makes me skeptical, but it’s useful to know!
I don’t have a rigorous argument against an infinite chain, but here’s my current set of intuitions: Let’s suppose that we have an infinite chain of reasons. Where does the chain come from? Does it pop out of nowhere? Or is there some intuition or finite collection of intuitions that we can posit as an explanation for the chain? While technically possible that the infinite chain could require infinite different intuitions to justify, this seem rather unlikely to me. What then if we accept that there is an intuition or there are intuitions behind the chain? Well, now we ask why these intuitions are reliable. And if we hit an infinite chain again, we can try the same trick and so on until we actually find a cycle
“The point here is that no matter how we measure complexity, it seems likely that philosophy would have a “high computational complexity class” according to that measure.”—I disagree. The task of philosophy is to figure out how to solve the meta problem, not to actually solve all individual problems or the worst individual problem
How strongly do you think improving human meta-philosophy would improve computational meta-philosophy?
Perhaps it’d be useful if there was a group that took more of a dialectical approach, such as in a philosophy class? For example, it could collect different perspectives on what needs to happen for AI to go well and try to help people understand the assumptions underlying the project they are considering being valuable.
Yeah, this is better than my example of food & medicine.
Unfortunately, I didn’t see this comment and now I can’t remember, so I just removed it and we’ll never know.
This comment didn’t age well
“Noting that there is a certain level of verbal confusion does not imply that there is nothing going on except verbal confusion”—I’m not claiming that verbal confusion is all that is going on, but I will admit that I could have been clearer about what I meant. You are correct that Chalmer’s aim was to highlight something about consciousness and for many people discussion of zombies can be a useful way of illustrating how reductive the materialist theory of consciousness is. But from a logical standpoint, it’s not really any different from the argument you’d make if you were discussing consciousness directly. So if the zombie argument is easier to grasp for you, great; otherwise you can ignore it and focus on direct discussion of consciousness instead.