Another thing to consider is reversible computing. If I’m run forward and then backward, do they cancel out or what? I don’t find this that interesting but you might.
David Joshua Sartor
Karma: 206
No, people will be dissatisfied with the answer to that too. The Easy Problem of Consciousness 2.
it’s a mysterious force permeating the entirety of the universe which lies in all things, but doesn’t have any detectable causal impact on the world beyond that described entirely by mundane theories
(Jeopardy! answer) What is, temperature?
My understanding is that centaurs beat unassisted Go AIs at this time. Not sure if they do noticeably better against humans.
That is, to the extent that the community can distinguish between good and bad criticisms on the merits,[24]determined critics of a particular author are just providing free verification labor, not conducting a DDoS attack: as it is written of the fifth virtue, “Do not believe you do others a favor if you accept their arguments; the favor is to you.” (And to the extent that the community can’t distinguish between good and bad criticisms on the merits, we have bigger problems.)
In fact, obviously to some extent we can’t distinguish between good and bad criticisms on the merits; we have limited time.
We have many big problems to solve, that doesn’t mean we should ignore smaller ones.
The reason this 2023 comment is compatible with the controversial 2020 remarks is because Achmiz believes that (what he considers) simple requests for definitions and examples are productive, even if not all comments are productive. According to this view, a request for examples of a phenomenon should not be hard to satisfy if the phenomenon is real; a request for a definition of an niche term should not be hard to satisfy if term means anything. Ideally, the answers should have already been in the post; the fact that commenters have to ask is a manifestation of the post not being complete.
My understanding was that the two conversations just used ‘obligation’ in different ways. One was assuming Heroic Responsibility, and the other was using prosaic responsibility.
“Perfect” is a misnomer. 1 book per question always is “perfect”.
It is a less ambitious alternative to ‘natural abstractions hypotheses’ and, in my view, more likely to be true.
I think that “less ambitious” deserves at least as much qualification as “more likely” (which is probably about none).
Of course, and you also see diminishing marginal costs.
There are really a lot of things that might’ve been hidden, and some of those would’ve been around a long time.
It’s been proven that, with sufficient compute, two FairBots cooperate even if they’re written in different formal-systems/programming-languages, and even if one has much more compute than the other; and FairBot isn’t exploitable. I don’t see why legibility couldn’t be found in some efficient agents too.
MIRI has also done work on decision problems outside LDT’s fair problem class, like Open-Source Prisoner’s Dilemma.
FairBot cooperates if it can prove you cooperate, defects otherwise. In this case, being too hard to predict gets you defected against.
the more quickly computers developed, the less history you’d have to bubble to hide the fact that they develop quickly
If you hide a small portion of history, people can still see that there are computers after and not before.
Where can I find the LessOnline album?
Where can I download these?
In particular I don’t know how to download “The Ninth Night of November”.
Friday’s far enough for milk to go bad, but it’s near enough for those other considerations.
I naturally read the title as a (shallow/nonsubstantive) reference to the video-game Undertale. It’s at the least a funny coincidence.
Here’s a quote, including spoilers.‴
* LOVE, too, is an acronym.
* It stands for “Level of Violence.”
* A way of measuring someone’s capacity to hurt.
* The more you kill, the easier it becomes to distance yourself.
* The more you distance yourself, the less you will hurt.
* The more easily you can bring yourself to hurt others.‴
Yeah, leading hypothesis changes.
The agent (at least effectively) has access to RNG that Murphy knows. I’m not sure if it’s supposed to be allowed private RNG as well.
I think the hypothesis class contains “square-free sequence until n”, for any finite n.
Set n = current index + 2.It’s very poorly worded, I might be misunderstanding too.
I have inside a thing I can choose to feel confused about in (what I think is) the same way as other people that speak of qualia. I wouldn’t think to do this if other people didn’t tell me, and I don’t have a strong opinion on if it’s a bad or good idea.
Part of me feels that that confusion is silly, but it’s not (yet) clearly-silly like being confused about free-will.
Free-will is a brain-concept that most people use, which can be used well or poorly, or perhaps be discarded; very similar is the sense of self. Also nearby (though less arbitrary) are the distinction between predicting and planning, and similarly the division of expected-utility into expectation and utility.
Maybe the perception of qualia is a similar feeling, or something?