interstice
3.1K posts
@an_interstice @CAIS Anthropic is the ruling party and EAs are the controlled opposition?
English
To clarify, the Center for AI Safety has not taken funding from Coefficient Giving / Open Philanthropy for years.
We believe the effective altruism movement is, unfortunately, controlled opposition. The less influence it has on AI safety, the better.
Dan Hendrycks@hendrycks
EA ≠ AI safety AI safety has outgrown the EA community The world will be safer with a broad range of people tackling many different AI risks
English
@tszzl if there's a software intelligence explosion we might get a very large amount of progress in a short amount of wallclock time
English
modern alignment methods seem to work reasonably well across orders of magnitude of model scaling, survived the transition to verifiable rewards and that should at least inform your decision making
Brangus🔍⏹️@RatOrthodox
I have heard that some anthropic safety leadership are going around telling people that alignment is a solved problem. This seems like a predictable failure to me, and I would like people who thought that funneling talent towards anthropic was a good idea to think about it.
English
interstice retweetledi
I never know how much these things are already in the water supply
Nevertheless, new post
"LLMs as Giant Lookup-Tables of Shallow Circuits"
lesswrong.com/posts/a9KqqgjN…
English
@bstract_thot guy who appears to be randomly fucking around in life in order to setup an arbitrary code execution exploit.
English
@Oszillate I have a similar plan. would you say the bottleneck is mostly in phase 1 or phase 2?
English
the returns on my former longterm plan of "be an interesting ~ impressive person and people will appear to be your friends" (tho effectively w/o further work already since after meeting my ex) have been in general bad for long and I should clearly shake up things about this
English
@VesselOfSpirit the problem is you're spelling out too much of your objection in the question so it's obvious you think they're dumb, it's better to ask a question that *implies* your objection 1 or 2 steps out.
English
like "i'm curious to explore your mental model of how sydney, a city in the southern hemisphere, can be the capital of poland" basically sounds like "sydney can't be the capital of poland because it's in the southern hemisphere, you absolute moron"
English
when someone says something that's clearly worth criticizing, basically every polite and constructive thing you can say in response ends up sounding like snarky veiled aggression, because of echoes of past conversations. it sucks
English
@nickcammarata in the compression frame couldn't a single datapoint rule out a large class of particularly simple compressors? that seems like what's going on in the double slit case.
although I agree compression probably isn't the final idea of what science is.
English
apparently my habit of consuming 6 apples a day is trad and #lindy
Gift of Trees of Draught of Barrel@IntractableLion
Apples were such a staple starch calorie source that the French named the potato after them. Actual data is hard to find for the 14-1600s but I've seen claims on the order of 1 pound/person/day for as long as the fruit would keep
English
@MatthewJBar @slatestarcodex hmm OK it sounds like you misread it in the OP but are correctly reading it in this reply chain, fair enough
English
@an_interstice @slatestarcodex Did you misread my last tweet? I completely understood that point.
English
@MatthewJBar @slatestarcodex you're totally misreading the chimp analogy. the chimps are meant to be the stand-in for *humans* and the *humans* are analogous to AI. Humans are depicted as the scary thing, not chimps
English
@slatestarcodex Chimpanzees didn't breed humans to be friendly, so the original analogy creates a misleading picture of how worried we should be. The golden retriever analogy is clearly closer to the actual situation.
English
@snigus @_AashishReddy riddle is depicted as correct though. it's implied the world would be destroyed by harry's soying out unless he swears an unbreakable vow to be paranoid about that happening.
English
@_AashishReddy Is that true? IIRC in HPMOR riddle can be read as skeptical of technology, but Harry is constantly soying out about technology.
English
I think there is an element in Yudkowskian thought which does sort of inexorably lead one to be sceptical of technology (technological civilisation?) in general. As opposed to the usual view that AI is a unique exception. One sees hints of this in HPMOR and The Sequences.
Eliezer Yudkowsky@allTheYud
It looks to me like telegrams were the optimal level of social communication technology, and society began to disintegrate after voice radio was invented, though not to the same extent as it began to collapse after television.
English
@Algon_33 trivial, def'n(mathematics): adj. applied to a result whose proof is known to the author
English
@an_interstice Thinking about this a bit more, I feel like "it's just a trivial combination of X,Y,Z" is a necessary outcome. If you look at some axioms, say Peano arithmetic, then every theorem is already contained in the axioms. The Prime Number Theorem, FLT, whatever.
English
Mulling over the fact that everything popular I've written was just rehashing what someone else said.
English
@Algon_33 OK so he had *one* original idea. that's apparently the standard for being a field-defining genius
English
@Algon_33 "Raussen and Skau: Is there one particular theorem or result you are the most proud of?
Gromov: Yes. It is my introduction of pseudoholomorphic curves, unquestionably. Everything else was just understanding what was already known and to make it look like a new kind of discovery."
English
@XshkpbrvjdLA8j @teortaxesTex so first you say consciousness is biology, then that it's existence itself? make up your mind!
English
@an_interstice @teortaxesTex Yes, we have no reasons to believe anything other than biology provides consc (say, from insects to humans), "computations" are made up
Also consc is probably identical to Being, so it's not an existent (a thing, process, phenomenon *in* the world), but rather existence itself
English
Well WELL WELL my strategy of sitting on my ass until *checks notes* a Google DeepMind research scientist publishes a paper that confirms my conclusions is productive as ever.
I don't expect mea culpas, but it's neat I have a citation to point to now.
Alexander Lerchner@AlexLerchner
🧵1/4 The debate over AI sentience is caught in an "AI welfare trap." My new preprint argues computational functionalism rests on a category error: the Abstraction Fallacy. AI can simulate consciousness, but cannot instantiate it. philpapers.org/rec/LERTAF
English
@teortaxesTex So consciousness depends on dynamics implemented in the "hardware" of a system, not the "software"?
My next question would be how what is hardware VS software is defined in general
English
@an_interstice > can't we just say that the physical system of the computer implementing the LLM is conscious
He's clearly saying this is not about the substrate. Maybe the server is conscious — with regard to workload scheduling or something like that. But it has no "LLM quale" per se.
English
@teortaxesTex regardless of how we choose to view it? Viewing it as performing computation is just a convenient way to interface with it for certain purposes.
English
@teortaxesTex so he's saying that computations can't be conscious because what it means for a thing to be a computation is observer-dependent, presupposing consciousness.
well OK, but then can't we just say that the physical system of the computer implementing the LLM is conscious...
English