I think most people stay in “off-the-cuff” territory most of the time. Getting past that usually requires putting in some effort, which requires motivation. That motivation could be internal—that you find the problem very interesting or very bothersome/worrisome on a personal level. Or external—you’re getting paid to work on it. If you aren’t getting paid and the topic has a primarily academic/abstract feel to it [which is often the case here], you will likely come up with some easy off-the-cuff ideas and stop at precisely that point at which the difficulty of thinking more about it becomes higher than your interest in it.
So the sequences don’t provide enough material to do work in things like Safety/Game Theory/Decision Theory/etc. - people also have to read some textbooks or research stuff?
I’ve read the (original?) Sequences, and I definitely do not feel qualified to do work in AI Safety, game theory, or decision theory. There are many posts on Less Wrong about those topics that I don’t even understand enough to follow, much less enough to contribute or critique them. So yes, I think most people would have to read textbooks on the subject or otherwise do a lot more learning work to significantly contribute. This is not too surprising; enough people have been doing enough work on those topics for enough years that I should not expect to be able to jump into it without some effort to cover what they’ve done and catch up.
I think most people stay in “off-the-cuff” territory most of the time. Getting past that usually requires putting in some effort, which requires motivation. That motivation could be internal—that you find the problem very interesting or very bothersome/worrisome on a personal level. Or external—you’re getting paid to work on it. If you aren’t getting paid and the topic has a primarily academic/abstract feel to it [which is often the case here], you will likely come up with some easy off-the-cuff ideas and stop at precisely that point at which the difficulty of thinking more about it becomes higher than your interest in it.
So the sequences don’t provide enough material to do work in things like Safety/Game Theory/Decision Theory/etc. - people also have to read some textbooks or research stuff?
I’ve read the (original?) Sequences, and I definitely do not feel qualified to do work in AI Safety, game theory, or decision theory. There are many posts on Less Wrong about those topics that I don’t even understand enough to follow, much less enough to contribute or critique them. So yes, I think most people would have to read textbooks on the subject or otherwise do a lot more learning work to significantly contribute. This is not too surprising; enough people have been doing enough work on those topics for enough years that I should not expect to be able to jump into it without some effort to cover what they’ve done and catch up.