I am seriously warning you: it is important that you become very skillful—fast, thorough, reflective, self-sharpening—at finding or building various decently-motivated-if-imperfect models of the same process/concept/thing so as to form a constellation of useful perspectives on different facets of it, and different ways of carving its joints, and why different facets/carvings might seem differentially important to various people or groups of people in different memetic or psychological contexts, et cetera.
Why do you think this is so important? As far as I can tell, this is not how humanity made progress in the past. Or was it? Did our best scientists and philosophers find or build “various decently-motivated-if-imperfect models of the same process/concept/thing so as to form a constellation of useful perspectives on different facets of it”?
Or do you claim that humanity made progress in the past despite not doing what you suggest, and that we could make much faster progress if we did? If so, what do you base your claim on (besides your intuition)?
Why do you think this is so important? As far as I can tell, this is not how humanity made progress in the past.
This actually seems to me exactly how humanity has made progress—countless fields and paradigms clashing and putting various perspectives on problems and making progress. This is a basic philosophy of science perspective, common to views as dissimilar as Kuhn and Feyerabend. There’s no one model that dominates in every field (most models don’t even dominate their field; if we look at the ones considered most precise and successful like particle physics or mathematics, we see that various groups don’t agree on even methodology, much less data or results).
But I think the individuals who contributed most to progress did so by concentrating on particular models that they found most promising or interesting. The proliferation of models only happen on a social level. Why think that we can improve upon this by consciously trying to “find or build various decently-motivated-if-imperfect models”?
None of that defends the assertion that humanity made progress by following one single model, which is what I was replying to, as shown by a highly specific quote from your post. Try again.
I didn’t mean to assert that humanity made progress by following one single model as a whole. As you point out, that is pretty absurd. What I was saying is that humanity made progress by (mostly) having each individual human pursue a single model. (I made a similar point before.)
I took Will’s suggestion to be that we, as individuals, should try to pursue many models, even ones that we don’t think are most promising, as long as they are “decently motivated”. (This is contrary to my intuitions, but not obviously absurd, which is why I wanted to ask Will for his reasons.)
I tried to make my point/question clearer in rest of the paragraph after the sentence you quoted, but looking back I notice that the last sentence there was missing the phrase “as individuals” and therefore didn’t quite serve my purpose.
I think you’re looking at later stages of development than I am. By the time Turing came around the thousands-year-long effort to formalize computation was mostly over; single models get way too much credit because they herald the triumph at the end of the war. It took many thousands of years to get to the point of Church/Goedel/Turing. I think that regarding justification we haven’t even had our Leibniz yet. If you look at Leibniz’s work he combined philosophy (monadology), engineering (expanding on Pascal’s calculators), cognitive science (alphabet of thought), and symbolic logic, all centered around computation though at that time there was no such thing as ‘computation’ as we know it (and now we know it so well that we can use it to listen to music or play chess). Archimedes is a much earlier example but he was less focused. If you look at Darwin he spent the majority of his time as a very good naturalist, paying close attention to lots of details. His model of evolution came later.
With morality we happen to be up quite a few levels of abstraction where ‘looking at lots of details’ involves paying close attention to themes from evolutionary game theory, microeconomics, theoretical computer science &c. Look at CFAI to see Eliezer drawing on evolution and evolutionary psychology to establish an extremely straightforward view of ‘justification’, e.g. “Story of a Blob”. It’s easy to stumble around in a haze and fall off a cliff if you don’t have a ton of models like that and more importantly a very good sense of the ways in which they’re unsatisfactory.
Those reasons aren’t convincing by themselves of course. It’d be nice to have a list of big abstract ideas whose formulation we can study on both the individual and memetic levels. E.g. natural selection and computation, and somewhat smaller less-obviously-analogous ones like general relativity, temperature (there’s a book about its invention), or economics. Unfortunately there’s a lot of success story selection effects and even looking closely might not be enough to get accurate info. People don’t really have introspective access to how they generate ideas.
Side question: how long do you think it would’ve taken the duo of Leibniz and Pascal to discover algorithmic probability theory if they’d been roommates for eternity?
If so, what do you base your claim on (besides your intuition)?
I think my previous paragraph answered this with representative reasons. This is sort of an odd way to ask the question ’cuz it’s mixing levels of abstraction. Intuition is something you get after looking at a lot of history or practicing a skill for awhile or whatever. There are a lot of chess puzzles I can solve just using my intuition, but I wouldn’t have those intuitions unless I’d spent some time on the object level practicing my tactics. So “besides your intuition” means like “and please give a fine-grained answer” and not literally “besides your intuition”. Anyway, yeah, personal experience plus history of science. I think you can see it in Nesov’s comments from back when, e.g. his looking at things like game semantics and abstract interpretation as sources of inspiration.
I think you’re looking at later stages of development than I am.
You’re right, and perhaps I should better familiarize myself with earlier intellectual history. Do you have any books you can recommend, on Leibniz for example?
This one perhaps. I haven’t read it but feel pretty guilty about that fact. Two FAI-minded people have recommended it to me, though I sort of doubt that they’ve actually read it either. Ah, the joys and sorrows of hypothetical CliffsNotes.
ETA: I think Vassar is the guy to ask about history of science or really history of anything. It’s his fault I’m so interested in history.
Why do you think this is so important? As far as I can tell, this is not how humanity made progress in the past. Or was it? Did our best scientists and philosophers find or build “various decently-motivated-if-imperfect models of the same process/concept/thing so as to form a constellation of useful perspectives on different facets of it”?
Or do you claim that humanity made progress in the past despite not doing what you suggest, and that we could make much faster progress if we did? If so, what do you base your claim on (besides your intuition)?
This actually seems to me exactly how humanity has made progress—countless fields and paradigms clashing and putting various perspectives on problems and making progress. This is a basic philosophy of science perspective, common to views as dissimilar as Kuhn and Feyerabend. There’s no one model that dominates in every field (most models don’t even dominate their field; if we look at the ones considered most precise and successful like particle physics or mathematics, we see that various groups don’t agree on even methodology, much less data or results).
But I think the individuals who contributed most to progress did so by concentrating on particular models that they found most promising or interesting. The proliferation of models only happen on a social level. Why think that we can improve upon this by consciously trying to “find or build various decently-motivated-if-imperfect models”?
None of that defends the assertion that humanity made progress by following one single model, which is what I was replying to, as shown by a highly specific quote from your post. Try again.
I didn’t mean to assert that humanity made progress by following one single model as a whole. As you point out, that is pretty absurd. What I was saying is that humanity made progress by (mostly) having each individual human pursue a single model. (I made a similar point before.)
I took Will’s suggestion to be that we, as individuals, should try to pursue many models, even ones that we don’t think are most promising, as long as they are “decently motivated”. (This is contrary to my intuitions, but not obviously absurd, which is why I wanted to ask Will for his reasons.)
I tried to make my point/question clearer in rest of the paragraph after the sentence you quoted, but looking back I notice that the last sentence there was missing the phrase “as individuals” and therefore didn’t quite serve my purpose.
I think you’re looking at later stages of development than I am. By the time Turing came around the thousands-year-long effort to formalize computation was mostly over; single models get way too much credit because they herald the triumph at the end of the war. It took many thousands of years to get to the point of Church/Goedel/Turing. I think that regarding justification we haven’t even had our Leibniz yet. If you look at Leibniz’s work he combined philosophy (monadology), engineering (expanding on Pascal’s calculators), cognitive science (alphabet of thought), and symbolic logic, all centered around computation though at that time there was no such thing as ‘computation’ as we know it (and now we know it so well that we can use it to listen to music or play chess). Archimedes is a much earlier example but he was less focused. If you look at Darwin he spent the majority of his time as a very good naturalist, paying close attention to lots of details. His model of evolution came later.
With morality we happen to be up quite a few levels of abstraction where ‘looking at lots of details’ involves paying close attention to themes from evolutionary game theory, microeconomics, theoretical computer science &c. Look at CFAI to see Eliezer drawing on evolution and evolutionary psychology to establish an extremely straightforward view of ‘justification’, e.g. “Story of a Blob”. It’s easy to stumble around in a haze and fall off a cliff if you don’t have a ton of models like that and more importantly a very good sense of the ways in which they’re unsatisfactory.
Those reasons aren’t convincing by themselves of course. It’d be nice to have a list of big abstract ideas whose formulation we can study on both the individual and memetic levels. E.g. natural selection and computation, and somewhat smaller less-obviously-analogous ones like general relativity, temperature (there’s a book about its invention), or economics. Unfortunately there’s a lot of success story selection effects and even looking closely might not be enough to get accurate info. People don’t really have introspective access to how they generate ideas.
Side question: how long do you think it would’ve taken the duo of Leibniz and Pascal to discover algorithmic probability theory if they’d been roommates for eternity?
I think my previous paragraph answered this with representative reasons. This is sort of an odd way to ask the question ’cuz it’s mixing levels of abstraction. Intuition is something you get after looking at a lot of history or practicing a skill for awhile or whatever. There are a lot of chess puzzles I can solve just using my intuition, but I wouldn’t have those intuitions unless I’d spent some time on the object level practicing my tactics. So “besides your intuition” means like “and please give a fine-grained answer” and not literally “besides your intuition”. Anyway, yeah, personal experience plus history of science. I think you can see it in Nesov’s comments from back when, e.g. his looking at things like game semantics and abstract interpretation as sources of inspiration.
You’re right, and perhaps I should better familiarize myself with earlier intellectual history. Do you have any books you can recommend, on Leibniz for example?
This one perhaps. I haven’t read it but feel pretty guilty about that fact. Two FAI-minded people have recommended it to me, though I sort of doubt that they’ve actually read it either. Ah, the joys and sorrows of hypothetical CliffsNotes.
ETA: I think Vassar is the guy to ask about history of science or really history of anything. It’s his fault I’m so interested in history.