Will Tassilo Think Singular Learning Theory Isn’t Useful for Alignment by the End of 2024?
Dec 31

Inspired by this quote from Evan Hubinger on Manifund:

Though I have high hopes for Singular Learning Theory, my modal outcome is that it's mostly just wrong and doesn't explain machine learning inductive biases that well. Inductive biases are very complex and most theories like this in the past have failed. Though I think this is a better bet here than most, I don't think I expect it to succeed.

I was surprised by this after everyone else I respect either thought Singular Learning Theory was the best thing since sliced bread or didn't feel able to evaluate what it is about. I couldn't find an already existing market about it, so I made this one.

About me: I Personally found the posts on lesswrong and the presentations on the singular learning theory seminar summit quite intruiging, but I still have a hard time wrapping my head around the material. I think I haven't engaged with enough math theories that are "technically correct" but overselling themselves to be sure I wouldn't be fooled by them.

I will probably ask some of the alignment researchers I most respect (such as John Wentworth or Steven Byrnes) for advice about the assessment unless it's dead-obvious. (Yeah I straightup copied this paragraph from tailcalled's post, because it's content fit perfectly for me as well).

Get Ṁ600 play money
Sort by:

Changed the title because it is probably less confusing and doesn't change the outcome much in my expectation. I also put the resolution date closer, because @JesseHoogland is confident they will have results by then. I left some leeway to make this question more about the idea than when they publish their results.

There hasn't been a lot of price movement. If the problem is that this is the wrong kind of market or that it should be resolved by someone who actually understands the material well, let me know.

More related questions