New to LessWrong?

New Comment
5 comments, sorted by Click to highlight new comments since: Today at 10:56 PM

A corollary of Chapman's 'if something seems hard the representation is probably wrong': if your working memory is overloaded, try changing what you pay attention to. In this particular case, paying attention to each monster on the screen in down well might be expensive compared to paying attention to the properties of open/safe spaces or paths. This figure ground inversion is also the point of bullet hells.

Strongly agree with the description of how transfer might actually work (relative to the Latin->math magical thinking). It's especially funny because It's a self referential error, the person who glosses over just how that's supposed to work is also the person who will gloss over the things that make transfer learning work.

I also think there's something really weird going on in terms of development psychology where kids are much better at transfer and then it gets broken by our education model or something. I definitely felt like noticing started being penalized at some point and most of the adults were in on it. It occurs to me that in addition to the concept of a 'summon sapience' spell, it's also useful (but harder) to notice the 'banish sapience' spell.

Reading this inspired me to go back and try Downwell on the PC again, after trying to beat it for a while and giving up several years ago. I'm attempting to use the skill "when things seem grindy and difficult, try identifying the simplest, most obvious improvement you can make to your strategy instead of just trying harder". So far this has included:

  • Using the "floating" playstyle instead of the default

  • Playing slowly and carefully on the first two zones, then rushing through the last two as quickly as possible.

  • Getting apples and other HP restoration items when at full health(I didn't realize they contributed to upgrading your max health)

Despite playing for much less time, I've gotten further than I did a few years ago! I've made it to the boss(standard difficulty) but died halfway through. Next I'm going to try getting combos in the first two zones, to accumulate more gems early.

Have you tried playing Spelunky? I think it might be better than Downwell as a rationality testing ground since it relies less on twitch reaction skills.

Congrats!

I have also played (and beaten) Spelunkey. I agree it's a pretty good option. (I also think puzzle games like Baba Is You are good in a somewhat more direct way, as are multiplayer games that require modeling incentives and coordination)

I'm spending a lot of effort these days on "training rationality qua rationality". But I actually somewhat like that Downwell is reflex-based, because it forces to me to think "okay, how do I deliberate practice something that requires a lot of skills at once, in a domain where it's hard to cleanly separate them?". 

The mechanism by which Downwell is hard to train is different from the mechanism by which, say, estimating the value of X-risk reduction projects is hard to train. But (I think) there is some shared structure of "man, it's really hard to train this" and then finding a way to push through and train it anyway.

Downwell is particularly significant to me because I know I plateau'd at it previously, so it's easier to see the diff of whether an intense deliberate practice focus can help. (This is a fairly unique fact about me+Downwell and not relevant to other people though)

Update: I beat standard mode. Getting lots of combos made the game much easier(and more fun)

Robin taught me the tool of "take lots of liveblogging notes".

aw, a little h/t to me. 

Just to note, I'd describe the thing I suggested differently: 

(mostly I'm going to write this for my own clarity)

your goal at the time of my suggestion was [learning about transformers], and you were primarily just reading and occasionally taking notes seemingly-for-memory (like writing down key concepts). It seemed to me that to "be in touch with the territory" of learning,  you should be watching your learning process, ie the felt-sense of the building of the schema as you go over the material, the sensations of uncertainty/ambiguity/confusion, the sensations of insight, etc. To see what you're attending to and how you're attending to it within the constraints of your sequential, moment-to-moment experience. And yeah, a great way to do that is write down what's going on internally while you interact with the material.

(perhaps you didn't hear it this way, but that's a bit of the generator / what I meant~)