It seems (to me) to be analogous to a lot of fairly technical pursuits:
Seismic analysis from purcussion events for finding oil.
Tracking the impact of an object on the moon to detect water.
Looking for the decay of particles produced and collided by accelerators.
Pitching to a batter, over time, will reveal the best way to pitch to that batter (what are his/her strenghts and weaknesses).
Haggling.
Approaching its most distilled form:
If a system is not giving you information, affect the system in some way [doesn’t have to be an “attack” per se]. How the system changes based on your input is instructive, so absorb all of that data.
Exactly. Poke the confusing-thing, make it give up evidence about how it works. And pay attention to that evidence.
I like your last two examples because they involve situations where you don’t have all the time in the world to approach a phenomenon, like we do (or at least feel like we do) when studying the fundamental and unchanging laws of Nature. You have to learn and adapt in real time.
Of course in a sports game you’re already going to “attack” because it’s part of the game. So the virtue lies in noticing the evidence it produces. It might seem like an obvious thing to say, but then you see people/teams repeating the same failed strategy over and over.
Haggling, negotiation, is pretty much the original context of the quote, and I think the immediate point was to avoid playing defensively and giving up initiative. Waff was trying to tell himself something like, “don’t just sit there intimidated by this powerful and mysterious woman, letting her frame the conversation to her advantage. Look for a way to learn more. Probe. Evoke a response.”
I’m also thinking of strategy games like, say, Starcraft. You want to commit some resources (units, time, your own attention) to scouting, in order to find out what types of units the enemy is relying on (so as to best counter them), which patches of valueable resource he has covered and how vulnerable they are to attack, how well he responds to raiding/harassment etc.
Another way of looking at it that (I think!) makes it obvious: it’s advice for AI, in the AI-in-the-box experiment.
I imagine it’s what Eliezer means when he says he won the experiment “the hard way”. He just kept poking for a psychological mechanism to exploit until he found one that happened to work with the given person. And the intermediate, failed attempts, probably helped him model the person and narrow down on whatever actually might work.
I don’t understand this one. Anyone want to explain it?
Attack. Then based upon the results of the attack modify your behaviour. Or attacck then update your model of the enemy.
It seems (to me) to be analogous to a lot of fairly technical pursuits: Seismic analysis from purcussion events for finding oil. Tracking the impact of an object on the moon to detect water. Looking for the decay of particles produced and collided by accelerators. Pitching to a batter, over time, will reveal the best way to pitch to that batter (what are his/her strenghts and weaknesses). Haggling.
Approaching its most distilled form: If a system is not giving you information, affect the system in some way [doesn’t have to be an “attack” per se]. How the system changes based on your input is instructive, so absorb all of that data.
Exactly. Poke the confusing-thing, make it give up evidence about how it works. And pay attention to that evidence.
I like your last two examples because they involve situations where you don’t have all the time in the world to approach a phenomenon, like we do (or at least feel like we do) when studying the fundamental and unchanging laws of Nature. You have to learn and adapt in real time.
Of course in a sports game you’re already going to “attack” because it’s part of the game. So the virtue lies in noticing the evidence it produces. It might seem like an obvious thing to say, but then you see people/teams repeating the same failed strategy over and over.
Haggling, negotiation, is pretty much the original context of the quote, and I think the immediate point was to avoid playing defensively and giving up initiative. Waff was trying to tell himself something like, “don’t just sit there intimidated by this powerful and mysterious woman, letting her frame the conversation to her advantage. Look for a way to learn more. Probe. Evoke a response.”
I’m also thinking of strategy games like, say, Starcraft. You want to commit some resources (units, time, your own attention) to scouting, in order to find out what types of units the enemy is relying on (so as to best counter them), which patches of valueable resource he has covered and how vulnerable they are to attack, how well he responds to raiding/harassment etc.
Another way of looking at it that (I think!) makes it obvious: it’s advice for AI, in the AI-in-the-box experiment.
I imagine it’s what Eliezer means when he says he won the experiment “the hard way”. He just kept poking for a psychological mechanism to exploit until he found one that happened to work with the given person. And the intermediate, failed attempts, probably helped him model the person and narrow down on whatever actually might work.