Jess Riedel — LessWrong

LESSWRONG
LW

Reinforcement Learning from Information Bazaar Feedback, and other uses of information markets

the problem is an inability to reliably commit to forgetting....A comment by John Wentworth mentioned that you can use amnestics to overcome this: make your purchase decision while under the influence of an amnestic that blocks your mind’s ability to write to long-term memory....When you read this, it’s hard not to immediately think of LLMs, which can make purchase decisions without committing anything to long-term memory

Robin Hanson also discussed using ems to do this in Age of Em. Probably the idea has been around even longer.

Proceedings of ILIAD: Lessons and Progress

Jess Riedel1y10

I think this is something we may do, with the caveat that we would make it an option that the author can choose (and that that would be clear to the readers.) We don’t want to get in the business of deciding which online discussions are good or bad enough to be worth endorsing as worth the reader’s time.

Thanks for raising this.

Proceedings of ILIAD: Lessons and Progress

Jess Riedel1y20

Thanks, we really appreciate the questions.

Our general approach to scope is to ask (1) if the topic is worth studying, and (2) if there are no other venues that can offer a substantially better review. If so, we’ll probably say yes. (We generally want to avoid reviewing manuscripts where there are already good existing journals who accept submissions on the topic, e.g., almost all interpretability.) We are willing to go outside our comfort zone to get worthwhile manuscripts reviewed imperfectly if the alternative is they get reviewed nowhere. One advantage... (read more)

Vacuum Decay: Expert Survey Results

Jess Riedel1y129

The random fluctuations in macroscopic chaotic systems, like Plinko or a well-flipped coin in air, can be just as fundamentally quantum as vacuum decay through tunneling. So by this argument you’d be unconcerned getting into a machine that flips a coin and shoots you if tails. Bad idea.

Abstractions as Redundant Information

Jess Riedel2y10

Suppose I sample the genomes of two random humans, $G_{1}$ and $G_{2} .$ What information is redundant across these two random variables?...So, for instance, I throw away $G_{i}$ , then I look at all the other genomes and see that in most places they’re the same - so when I sample my new $G_{i},$ I know that it should match all the other genomes in all those places.

I can't really tell what distribution(s) you're talking about here. You describe G_1 and G_2 as two random humans; wouldn't these then just be two draws from the same distribution o... (read more)