Check out my website: thermontology.com
interstice
Karma: 3,023
Well as I mentioned you would need to acquire more degrees of freedom. Basically a civilization of sentients running for an unbounded amount of time.
If the cosmological constant is zero(not totally implausible, there exists some data consistent with the apparent cosmological constant decaying to zero), then it seems plausible that we could escape heat death and run computations forever, since the energy needed to erase a bit declines with temperature, which asymptotes to zero in such a universe.
Then there’s the problem of acquiring more degrees of freedom to avoid our computations looping. Haven’t really thought about this part but does not seem obviously impossible. Zero-cosmological constant universes at least allow for asymptotically infinite information capacity in the forward light-cone
A likely very important consideration is that we would be meeting aliens expanding into the same space as us.
Yeah maybe not the best terminology, but I just mean they end up practically optimizing for those things instead of their stated goals.
Yeah maybe not the best terminology, I just mean they essentially end up pursuing those things in addition to/instead of their purported values.
Hmm, maybe there was a miscommunication, I didn’t mean to suggest there were?(although maybe there are some, interesting question...)
I’m saying if you can predict that a secret research program is unlikely to succeed, you can do something different such as politics or a non-secret research program.
I’m not comfortable setting aside the goodness of the consequences of the work
Reasonable. But if your project is likely to be ~neutral, that’s good to know too, no? Even if you don’t want to do more flywheel-y research, you could always pivot to doing something else entirely.
There is absolutely a population of conceptual AI researchers with extremely commercially valuable ideas who have decided not to ply their skills in that arena
Hmm, interesting...I’m curious who you’re thinking of but maybe you’d rather not say.
I actually do agree that “conceptually skilled” people can be very valuable in charting an overall direction, so it’s good if such people refrain from helping public AI projects. But I think conceptual skill is most useful when coupled with some sort of powerful feedback loop.
So this can work really well in domains like math, e.g. Andrew Wiles. But even there the dangers of wireheading on your own impressions of success are great, which is why some sort of scarce good in the external world like status and money can be a good feedback signal(well, “good” from the perspective of getting stuff done anyway, setting aside the goodness of the consequences of the work)
I wouldn’t say strong counter-evidence, but some counter-evidence yes. It could also be a sign they’re pivoting to more practical directions ETA: some evidence for the latter possibility(maybe?) is that there are rumors of a recent mini-coup at SSI.
I was also thinking of MIRI’s secret research and Jonathan Blow’s programming language. It’s possible that SSI and J. Blow could still succeed.
While it does seem to be the case that people who “get stuff done in the world” are often basically wireheading on social status/money/power, people who don’t do this often end up wireheading on their own imaginations which is even worse from a “contact with reality” perspective. Thought inspired by SSI seemingly failing to achieve anything with their money and years of secret research(unless the rumors of them having CL are true, but I would guess not?). More generally I have a cached intuition that people who start a secrecy focused institution/research project usually fail.
(Not an insider, but)To me it seems like the EAs both defer more to expert consensus (and status-ranking) and more think they can accomplish their aims politically. And they also internalize the elite worldview and norms more.
And yes, Some Thoughts on Metaphilosophy is still my best take on what philosophy is
OT, but what do you think of my take that philosophy is about mesa-optimizers extracting patterns (relevant to their functioning) from their underlying statistical model into a legible format?
“Messiness” seems to pop up only when you try to interpret domain A through domain B, but you’re confused about how to do it (e.g. you don’t know statistics) or lack capabilities to do it (e.g. you don’t have precise enough instruments).
Hmmm, so I think of messiness as being like...essentially, since the world is complex, the best models of most phenomena have many parameters.
But this means that, for long-running disagreements between world-models, most pieces of evidence won’t be able to resolve them. Because each world-model will have a lot of free parameters to interpret a new piece of evidence in any given way. This can be especially bad for controversial topics since people will engage in motivated reasoning, which there’s plenty of room for due to the abundance of free parameters. It’s also computationally expensive.
Thus to distinguish between world models, it’s often actually easier if the data comes from a small, artificially restricted domain so that there’s less wiggle room.
As an example, consider the “pots VS peoples” theories of the indo-europeans. That was definitively resolved by DNA data. Though people had plenty of archaeological data, it was not as disambiguating.
your definition doesn’t define what is an interesting question/problem
I think of interesting problems as being those whose answers (you think will) most help you disambiguate your global world model.
Gettier cases....yes there’s some similarity in examining an edge case. Lots of philosophical thought experiments are like that(but to be honest I never really understood the intuition that the weird JTB cases were not “knowledge” or why this distinction was important)
I actually think of “questions” and “problems” as being fairly neat concepts. “Question” just being some thing you can query about the system, like “what is the position of Mars tonight?”, “what is the regulator of this elliptic curve?”, whereas “problems” would be searches over easily verifiable structures, like “does there exist a proof of this conjecture?”, “does there exist a way of synthesizing such and such a chemical element?”
Regarding ontology...I think of flywheels as being small, toy domains which give unusually clear, unambiguous answers, which can aid in the construction of theories applicable to messier domains. “High resolution” parts of reality, if you will. Take the movement of the planets: the laws of motion can be seen more easily when separated from the vagaries of Earthly existence such as air drag. Or particle accelerators are literally higher resolution than most things(they probe higher energy levels ~ smaller scales). This is connected to my overall schema of hot VS cold things which I’ll get into later.
I’d like to frame philosophical undertaking in a way that highlights practical applications and consequences
What do you think of my discussion of usefulness in 1.6? I agree my discussion is mindy but I feel like all philosophy is mindy(even the purportedly anti-mindy variety)
I think of a flywheel as being something that is easy to iterate on and get unambiguous results. So yes, basically “well defined measurements” + “smaller state-space”.
Combining two flywheels into one...hmm, maybe. I would be a bit more general and say that deep discoveries compress a lot of your worldview, which could include uniting two small well-defined domains(or showing a particular small domain well predicts a bigger domain)
Funnily enough I just wrote a post about empirical flywheels in math and philosophy which I think are ~the same thing you call “scrying”.
I think you maybe could model it in logical induction? Basically when you are “scrying” you’re working in some sort of small verifiable domain which can give you lots of unambiguous feedback quickly, in the hopes that the information you get will be applicable to some larger setting you care more about. Doing it well would be like, developing good intuition/heuristics about what sort of small verifiable problems tend to be useful to work on.
Sure. By “not explicitly pre-trained” I just mean to say that there’s nothing ‘special’ about the characters from the training algorithm’s point of view, so in this respect they’re not so different from a hypothetical general predictive algorithm in humans(although actually I guess the human brain attaches special salience to other people, but regardless...)
It has a general objective of next-token prediction for which modeling characters is a useful strategy. IMO it’s plausible that the human brain is “trained” on prediction to a large extent, for which modeling characters is also a useful strategy.
Yes I agree actually. Any part of our physics could be run on a Turing machine which could be run in GoL. So any precise theorem along these lines would need to take into account the naturality of the mapping.