Yes, though more intuitively/implicitly than intentionally/explicitly. I actually made an effort to avoid it, at times, in order to increase variance among the solutions—otherwise I’d probably have a lot more solutions in the cluster of communication #8/9.
In general, that motion is good for “deep” thinking, in the sense of generating plans which require multiple intermediate steps (and where it may not be obvious what the “right” intermediate steps are). In that case, coming up with actually good intermediate steps is what matters, and the barrier/path mental motion is useful for that. In this exercise, on the other hand, we’re just going for lots of variance on a “shallow” problem (i.e. just one or two steps), and making the plan “good” is explicitly not a priority.
I mean, ultimately the hope is that the skill could be chunked and then transfer to deep problems.
One mechanism is that deep problems require a few little, intermediate steps of pure creativity, where this would be helpful.
Though I am curious if you’d claim that there’s actually a qualitatively different kind of creativity at play for deep problem, compared to shallow ones, and as a result that things won’t transfer. (And, if you claim that, what sort of babble challenge prompt or game format would help with that.)
Though I am curious if you’d claim that there’s actually a qualitatively different kind of creativity at play for deep problem, compared to shallow ones, and as a result that things won’t transfer.
I do think that, although it’s tricky to explain the difference in concrete terms. At a theoretical level, it’s the difference between solving a problem where lots of solutions work and it’s easy to tell that they work, vs a problem where most solutions don’t work and it’s hard to tell which ones don’t work without further thought/effort.
Here’s an example which I used to use as an interview problem: fill in a blank Sudoku board. There’s a lot more than 50 ways to do it, but if you just start writing in random numbers then you’ll probably get stuck.
(The following is pre-rigorous and tries to make legible some vague intuitions.)
My current toy model is that there’s a kind of cognition where you’re loading up all the constraints in working memory, and then aim your babble to fire given the constraints. But the basic “babble firepower” is really the same.
You also need the to learn to integrate new gears into your babble. For example, formulating the right maths theorems requires learning and manipulating a bunch of unfamiliar abstractions. And you need to have those abstractions be readily accessible by your babble muscle, as opposed to painstakingly loaded up with deliberate, S2 reflection.
But it still seems to that “figuring out which solutions work” is still a separate pruning step, and the babble component is largely similar?
I’m not saying the babble goes all the way, but that there’s some micro-level, elementary cognitive motion that’s the same.
fill in a blank Sudoku board. There’s a lot more than 50 ways to do it, but if you just start writing in random numbers then you’ll probably get stuck.
Interesting! I guess you also have “Send a satellite into orbit without using a rocket?”
Yes, though more intuitively/implicitly than intentionally/explicitly. I actually made an effort to avoid it, at times, in order to increase variance among the solutions—otherwise I’d probably have a lot more solutions in the cluster of communication #8/9.
In general, that motion is good for “deep” thinking, in the sense of generating plans which require multiple intermediate steps (and where it may not be obvious what the “right” intermediate steps are). In that case, coming up with actually good intermediate steps is what matters, and the barrier/path mental motion is useful for that. In this exercise, on the other hand, we’re just going for lots of variance on a “shallow” problem (i.e. just one or two steps), and making the plan “good” is explicitly not a priority.
I mean, ultimately the hope is that the skill could be chunked and then transfer to deep problems.
One mechanism is that deep problems require a few little, intermediate steps of pure creativity, where this would be helpful.
Though I am curious if you’d claim that there’s actually a qualitatively different kind of creativity at play for deep problem, compared to shallow ones, and as a result that things won’t transfer. (And, if you claim that, what sort of babble challenge prompt or game format would help with that.)
I do think that, although it’s tricky to explain the difference in concrete terms. At a theoretical level, it’s the difference between solving a problem where lots of solutions work and it’s easy to tell that they work, vs a problem where most solutions don’t work and it’s hard to tell which ones don’t work without further thought/effort.
Here’s an example which I used to use as an interview problem: fill in a blank Sudoku board. There’s a lot more than 50 ways to do it, but if you just start writing in random numbers then you’ll probably get stuck.
(The following is pre-rigorous and tries to make legible some vague intuitions.)
My current toy model is that there’s a kind of cognition where you’re loading up all the constraints in working memory, and then aim your babble to fire given the constraints. But the basic “babble firepower” is really the same.
You also need the to learn to integrate new gears into your babble. For example, formulating the right maths theorems requires learning and manipulating a bunch of unfamiliar abstractions. And you need to have those abstractions be readily accessible by your babble muscle, as opposed to painstakingly loaded up with deliberate, S2 reflection.
But it still seems to that “figuring out which solutions work” is still a separate pruning step, and the babble component is largely similar?
I’m not saying the babble goes all the way, but that there’s some micro-level, elementary cognitive motion that’s the same.
Interesting! I guess you also have “Send a satellite into orbit without using a rocket?”