3 ms·

The idea that a static world implies monotonically increasing confidence in your beliefs is really shaky, logically. Hill-climbing in a landscape of belief-fitn

by theoh 12y ago

The idea that a static world implies monotonically increasing confidence in your beliefs is really shaky, logically. Hill-climbing in a landscape of belief-fitness could easily arrive at a local maximum. To further improve your beliefs you might need to become less certain. Apart from that, how can something (confidence) increase in a static world? PG has clearly tossed this essay off without serious thought or discussion. A waste of time.

nl 12y ago
The idea that a static world implies monotonically increasing confidence in your beliefs is really shaky, logically. Hill-climbing in a landscape of belief-fitness could easily arrive at a local maximum. Exactly, and it's common to see that in many fields. People are persuaded by things that were bad ideas when they were new, and aren't swayed by new evidence to the contrary. To further improve your beliefs you might need to become less certain. Isn't that sort of the point of the essay? Apart from that, how can something (confidence) increase in a static world? I don't understand. Didn't your previous point argue exactly how hill-climbing could happen, and isn't that a (bad) example of confidence changing? Or are you really arguing over the semantics changing beliefs in a static world? I hope you aren't; otherwise, like you said: A waste of time.
- theoh 12y ago
  This is definitely a waste of time but my point is that you don't need to have a changing (dynamic) fitness landscape for monotonicity to be insufficient in the long term - you could get stuck at a local maximum _even_ in a static world. The second point is that drawing a line between the "world" and your beliefs about it is a pretty dopey dualistic way of thinking. But hey, Graham doesn't care that there might be real knowledge and valid experience of these matters in the accumulated history of philosophy or psychology, he just shoots from the hip with some flaky positivistic analogy with computation... Ridiculous.
  - ncallaway 12y ago
    This assumes you can only hold beliefs that are adjacent to your current beliefs (i.e. the graph of your beliefs over time must be continuous). I disagree with that assumption. You can explore ideas that you don't hold, and these can be quite varied from your current beliefs. Once you find an idea that you put more confidence into that your current beliefs, then that idea becomes your belief. The local optimum from high-climbing doesn't constrain you in a world where your beliefs can teleport from one hill to another.

by theoh 12y ago

The idea that a static world implies monotonically increasing confidence in your beliefs is really shaky, logically. Hill-climbing in a landscape of belief-fitness could easily arrive at a local maximum. To further improve your beliefs you might need to become less certain. Apart from that, how can something (confidence) increase in a static world? PG has clearly tossed this essay off without serious thought or discussion. A waste of time.

nl 12y ago

The idea that a static world implies monotonically increasing confidence in your beliefs is really shaky, logically. Hill-climbing in a landscape of belief-fitness could easily arrive at a local maximum. Exactly, and it's common to see that in many fields. People are persuaded by things that were bad ideas when they were new, and aren't swayed by new evidence to the contrary. To further improve your beliefs you might need to become less certain. Isn't that sort of the point of the essay? Apart from that, how can something (confidence) increase in a static world? I don't understand. Didn't your previous point argue exactly how hill-climbing could happen, and isn't that a (bad) example of confidence changing? Or are you really arguing over the semantics changing beliefs in a static world? I hope you aren't; otherwise, like you said: A waste of time.

theoh 12y ago
This is definitely a waste of time but my point is that you don't need to have a changing (dynamic) fitness landscape for monotonicity to be insufficient in the long term - you could get stuck at a local maximum _even_ in a static world. The second point is that drawing a line between the "world" and your beliefs about it is a pretty dopey dualistic way of thinking. But hey, Graham doesn't care that there might be real knowledge and valid experience of these matters in the accumulated history of philosophy or psychology, he just shoots from the hip with some flaky positivistic analogy with computation... Ridiculous.
- ncallaway 12y ago
  This assumes you can only hold beliefs that are adjacent to your current beliefs (i.e. the graph of your beliefs over time must be continuous). I disagree with that assumption. You can explore ideas that you don't hold, and these can be quite varied from your current beliefs. Once you find an idea that you put more confidence into that your current beliefs, then that idea becomes your belief. The local optimum from high-climbing doesn't constrain you in a world where your beliefs can teleport from one hill to another.