> I have a feeling that this is basically the generic talking point to use when your opponent is more radical than you.
EA people would probably phrase it as something about how updating strong priors in response to weak evidence needs to happen slowly, but I feel the Bayesian formulation is a bit toothless when it comes to practical applications.
The broader point is that when your opponent is more radical than you on a factual issue[0], but they don't present any evidence for why, they're probably wrong. This isn't good enough in a debate but it's a fine heuristic for deciding whether to use opioids as performance enhancers.
> I think you're painting effective altruism with too broad a brush and giving them too little credit. I'm very skeptical that the typical effective altruist is ordering semaglutide from china
This is a fair criticism, but I didn't mean to apply it to the movement as a whole, only to the particular failure mode where some effective altruists (or more generally, rationalists) talk themselves into doing bizarre and harmful things that equivalently smart non-EAs would not. It's easy to talk about Chesterton's Fence but it's not so easy to remember it when you read about something cool on Wikipedia or HN.
> Hold on, all it takes to turn over "conventional wisdom" on nutrition is "many nutritionists" and "new research"?
I'm just looking for a heuristic that stops you doing weird rationalist stuff, not a population-wide set of dietary recommendations. It's okay if some low-risk experimentation slips through, even if it's not statistically rigorous and even if it's very slightly harmful.
The point is that there are two requirements being met: first, no strong expert consensus ("many nutritionists" was too weak a phrasing and I apologise), and second, if you ask a few random strangers (representing conventional wisdom) whether eating more vegetables and less bread is good for you they'll tell you to go for it if you want to, while if you ask about using non-prescription opioids they'll be against it.
EA people would probably phrase it as something about how updating strong priors in response to weak evidence needs to happen slowly, but I feel the Bayesian formulation is a bit toothless when it comes to practical applications.
The broader point is that when your opponent is more radical than you on a factual issue[0], but they don't present any evidence for why, they're probably wrong. This isn't good enough in a debate but it's a fine heuristic for deciding whether to use opioids as performance enhancers.
> I think you're painting effective altruism with too broad a brush and giving them too little credit. I'm very skeptical that the typical effective altruist is ordering semaglutide from china
This is a fair criticism, but I didn't mean to apply it to the movement as a whole, only to the particular failure mode where some effective altruists (or more generally, rationalists) talk themselves into doing bizarre and harmful things that equivalently smart non-EAs would not. It's easy to talk about Chesterton's Fence but it's not so easy to remember it when you read about something cool on Wikipedia or HN.
> Hold on, all it takes to turn over "conventional wisdom" on nutrition is "many nutritionists" and "new research"?
I'm just looking for a heuristic that stops you doing weird rationalist stuff, not a population-wide set of dietary recommendations. It's okay if some low-risk experimentation slips through, even if it's not statistically rigorous and even if it's very slightly harmful.
The point is that there are two requirements being met: first, no strong expert consensus ("many nutritionists" was too weak a phrasing and I apologise), and second, if you ask a few random strangers (representing conventional wisdom) whether eating more vegetables and less bread is good for you they'll tell you to go for it if you want to, while if you ask about using non-prescription opioids they'll be against it.
[0] Values-driven stuff is different.