This calls for a link to simulated annealing, an optimization heuristic. Here, initial sampling is “provocation” and the jumps later in the process of cooling are “movement”.
It shouldn’t be so surprising to me, seeing simulated annealing in so many places. It’s a good approximation for optimising NP problems, it should develop organically in places dealing with NP problems.
How do you set your prior for initial sampling? This seems like the fundamental problem to me (with regards to creativity). Suppose I want to brainstorm about a scientific problem. I don’t want to start with “cars have square wheels” because that has nothing to do with my problem. But, if I start with a statement of the problem, I have already “primed” myself in a certain direction, even if I just negate the problem statement.
I guess the analogy breaks down a little. For SA, I would use a uniform prior, and I’m guaranteed eventually to explore every part of the space. Idea space is too big for that. On the other hand, the trade-off is similar: more random starting points that are random, or fewer random starting points using a better prior.
Isn’t that the purpose of the 5 provocation techniques? You pick a few statements which are taken for granted related to the area of the problem, and then reverse/negate/distort them in some way. This isn’t a uniform distribution over idea space, since the taken-for-granted statements are related to the problem area.
This raises interesting off-topic question: does ‘intelligence’ itself confer significant advantage over such methods (which can certainly be implemented without anything resembling agent’s real world utility)?
We are transitioning to being bottlenecked (in our technological progress, at least) by optimization software implementing such methods, rather than being bottlenecked by our intelligence (that is in part how the exponential growth is sustained despite constant human intelligence); if the AI can’t do a whole lot better than our brainstorming, it probably won’t have upper hand over dedicated optimization software.
This calls for a link to simulated annealing, an optimization heuristic. Here, initial sampling is “provocation” and the jumps later in the process of cooling are “movement”.
It shouldn’t be so surprising to me, seeing simulated annealing in so many places. It’s a good approximation for optimising NP problems, it should develop organically in places dealing with NP problems.
How do you set your prior for initial sampling? This seems like the fundamental problem to me (with regards to creativity). Suppose I want to brainstorm about a scientific problem. I don’t want to start with “cars have square wheels” because that has nothing to do with my problem. But, if I start with a statement of the problem, I have already “primed” myself in a certain direction, even if I just negate the problem statement. I guess the analogy breaks down a little. For SA, I would use a uniform prior, and I’m guaranteed eventually to explore every part of the space. Idea space is too big for that. On the other hand, the trade-off is similar: more random starting points that are random, or fewer random starting points using a better prior.
Isn’t that the purpose of the 5 provocation techniques? You pick a few statements which are taken for granted related to the area of the problem, and then reverse/negate/distort them in some way. This isn’t a uniform distribution over idea space, since the taken-for-granted statements are related to the problem area.
This raises interesting off-topic question: does ‘intelligence’ itself confer significant advantage over such methods (which can certainly be implemented without anything resembling agent’s real world utility)?
We are transitioning to being bottlenecked (in our technological progress, at least) by optimization software implementing such methods, rather than being bottlenecked by our intelligence (that is in part how the exponential growth is sustained despite constant human intelligence); if the AI can’t do a whole lot better than our brainstorming, it probably won’t have upper hand over dedicated optimization software.
incredible comment. that was very insightful, thanks