and Carl Shulman notes that his approach “has lead [him] astray repeatedly, but I haven’t seen as many successes.”
That quote may not convey my view, so I’ll add to this. I think Eliezer has had a number of striking successes, but in that comment I was saying that it seemed to me he was overshooting more than undershooting with the base rate for dysfunctionality in institutions/fields, and that he should update accordingly and check more carefully for the good reasons that institutional practice or popular academic views often (but far from always) indicate. That doesn’t mean one can’t look closely and form much better estimates of the likelihood of good invisible reasons, or that the base rate of dysfunction is anywhere near zero. E.g. I think he has discharged the burden of due diligence wrt MWI.
If many physicists say X, and many others say Y and Z which seem in conflict with X, then at a high rate there will be some good arguments for X, Y, and Z. If you first see good arguments for X, you should check to see what physicists who buy Y and Z are saying, and whether they (and physicists who buy X) say they have knowledge that you don’t understand.
In the case of MWI, the physicists say they don’t have key obscure missing arguments (they are public and not esoteric), and that you can sort interpretations into ones that accept the unobserved parts of the wave function in QM as real (MWI, etc), ones that add new physics to pick out part of the wavefunction to be our world, and ones like shut-up-and-calculate that amount to ‘don’t talk about whether parts of the wave function we don’t see are real.’
Physicists working on quantum foundations are mostly mutually aware of one another’s arguments, and you can read or listen to them for their explanations of why they respond differently to that evidence, and look to the general success of those habits of mind. E.g. the past success of scientific realism and Copernican moves: distant lands on Earth that were previously unseen by particular communities turned out to be real, other Sun-like stars and planets were found, biological evolution, etc. Finding out that many of the interpretations amount to MWI under another name, or just refusing to answer the question of whether MWI is true, reduces the level of disagreement to be explained, as does the finding that realist/multiverse interpretations have tended to gain ground with time and to do better among among those who engage with quantum foundations and cosmology.
In terms of modesty, I would say that generally ‘trying to answer the question about external reality’ is a good epistemic marker for questions about external reality, as is Copernicanism/not giving humans a special place in physics or drastically penalizing theories on which the world is big/human nature looks different (consistently with past evidence). Regarding new physics for objective collapse, I would also note the failure to show it experimentally and the general opposition to it. That seems sufficient to favor the realist side of the debate among physicists.
In contrast, I hadn’t seen anything like such due diligence regarding nutrition, or precedent in common law.
Regarding the OP thesis, you could summarize my stance as that assigning ‘epistemic peer’ or ‘epistemic superior/inferior’ status in the context of some question of fact requires a lot of information and understanding when we are not assumed to already have reliable fine-grained knowledge of epistemic status. That often involves descending into the object-level: e.g. if the class of ‘scientific realist arguments’ has a good track record, then you will need to learn enough about a given question and the debate on it to know if that systemic factor is actually at play in the debate before you can know whether to apply that track record in assessing epistemic status.
In that comment I was saying that it seemed to me he was overshooting more than undershooting with the base rate for dysfunctionality in institutions/fields, and that he should update accordingly and check more carefully for the good reasons that institutional practice or popular academic views often (but far from always) indicate. That doesn’t mean one can’t look closely and form much better estimates of the likelihood of good invisible reasons, or that the base rate of dysfunction is anywhere near zero.
I offered that quote to cast doubt on Rob’s assertion that Eliezer has “a really strong epistemic track record, and that this is good evidence that modesty is a bad idea.” I didn’t mean to deny that Eliezer had some successes, or that one shouldn’t “look closely and form much better estimates of the likelihood of good invisible reasons” or that “the base rate of dysfunction is anywhere near zero”, and I didn’t offer the quote to dispute those claims.
Readers can read the original comment and judge for themselves whether the quote was in fact pulled out of context.
That quote may not convey my view, so I’ll add to this. I think Eliezer has had a number of striking successes, but in that comment I was saying that it seemed to me he was overshooting more than undershooting with the base rate for dysfunctionality in institutions/fields, and that he should update accordingly and check more carefully for the good reasons that institutional practice or popular academic views often (but far from always) indicate. That doesn’t mean one can’t look closely and form much better estimates of the likelihood of good invisible reasons, or that the base rate of dysfunction is anywhere near zero. E.g. I think he has discharged the burden of due diligence wrt MWI.
If many physicists say X, and many others say Y and Z which seem in conflict with X, then at a high rate there will be some good arguments for X, Y, and Z. If you first see good arguments for X, you should check to see what physicists who buy Y and Z are saying, and whether they (and physicists who buy X) say they have knowledge that you don’t understand.
In the case of MWI, the physicists say they don’t have key obscure missing arguments (they are public and not esoteric), and that you can sort interpretations into ones that accept the unobserved parts of the wave function in QM as real (MWI, etc), ones that add new physics to pick out part of the wavefunction to be our world, and ones like shut-up-and-calculate that amount to ‘don’t talk about whether parts of the wave function we don’t see are real.’
Physicists working on quantum foundations are mostly mutually aware of one another’s arguments, and you can read or listen to them for their explanations of why they respond differently to that evidence, and look to the general success of those habits of mind. E.g. the past success of scientific realism and Copernican moves: distant lands on Earth that were previously unseen by particular communities turned out to be real, other Sun-like stars and planets were found, biological evolution, etc. Finding out that many of the interpretations amount to MWI under another name, or just refusing to answer the question of whether MWI is true, reduces the level of disagreement to be explained, as does the finding that realist/multiverse interpretations have tended to gain ground with time and to do better among among those who engage with quantum foundations and cosmology.
In terms of modesty, I would say that generally ‘trying to answer the question about external reality’ is a good epistemic marker for questions about external reality, as is Copernicanism/not giving humans a special place in physics or drastically penalizing theories on which the world is big/human nature looks different (consistently with past evidence). Regarding new physics for objective collapse, I would also note the failure to show it experimentally and the general opposition to it. That seems sufficient to favor the realist side of the debate among physicists.
In contrast, I hadn’t seen anything like such due diligence regarding nutrition, or precedent in common law.
Regarding the OP thesis, you could summarize my stance as that assigning ‘epistemic peer’ or ‘epistemic superior/inferior’ status in the context of some question of fact requires a lot of information and understanding when we are not assumed to already have reliable fine-grained knowledge of epistemic status. That often involves descending into the object-level: e.g. if the class of ‘scientific realist arguments’ has a good track record, then you will need to learn enough about a given question and the debate on it to know if that systemic factor is actually at play in the debate before you can know whether to apply that track record in assessing epistemic status.
I offered that quote to cast doubt on Rob’s assertion that Eliezer has “a really strong epistemic track record, and that this is good evidence that modesty is a bad idea.” I didn’t mean to deny that Eliezer had some successes, or that one shouldn’t “look closely and form much better estimates of the likelihood of good invisible reasons” or that “the base rate of dysfunction is anywhere near zero”, and I didn’t offer the quote to dispute those claims.
Readers can read the original comment and judge for themselves whether the quote was in fact pulled out of context.
Please take my comment as explaining my own views, lest they be misunderstood, not condemning your citation of me.
Okay, thank you for the clarification.
[In the original version, your comment said that the quote was pulled out of context, hence my interpretation.]