Tuesday, 14 June 2011

Mind Charting II & Eliezer's Exercises

Having described the problem yesterday, I made a concerted effort today to throw myself into the mindcharting. Harnessing some of the techniques Academian described last Friday, with some help from my girlfriend, who it seems has already mastered most of those techniques for herself, I was able to make a lot of progress on the development and classification of the chart. I'm hopeful it will be useful when we get to chart-surgery in the next few days.

There were two particular pieces of advice that I found useful - first, the idea of simply noticing when my brain wanted to argue with me about the theory behind the exercise and letting it know that I'd noticed, and would let it express it's qualms later: I have a piece of paper with 25 tally marks on it, and made quite a lot of progress on my chart.

The other useful exercise was trying to associate the work with more pleasant sensations, not too much progress so far, but I'm getting there.

The afternoon's session was mostly Eliezer abusing the RBCers for material for his book. I took a few notes, and here is what I remember: it's quite bullet-pointy, partly because the talk was quite bullet-pointy and partly because that's what happens when you try to take notes on half a side of A4.

The first pearl of wisdom I remember was the idea that, in real life, Munchkinism is a good thing: the basic principle is that we want to avoid the error committed by the pigeons in the experiment referred to in this experiment - the fact that something requires more work to achieve does not make it more valuable.

We then each spent 30 seconds trying our hardest to come up with an airtight argument proving that the sky is green (the point being that if know arguments can be used to argue that the sky is green, then we should be more suspicious of them). Arguments included:
The languages which have sensible colour words only have one word for the two concepts of "blue" and "green" - so there's clearly no difference, and the sky is green.
The sky must be green, because plants are green, and they would have evolved to camouflage against the sky.
The sky is green because somewhere in the multiverse there is something that more closely fits our definition of "sky" than the thing above our heads now does, and is green.
And finally...
We are in a room of intelligent people who have all come to the conlcusion that the sky is green, the sky must be green

We then spent another 5 minutes coming up with justifications for the policy that we should eat babies... reasons included "life is good, eating more babies would allow us to create more of them, so increase the total amount of life" and the Umeshism "If you've never eaten a baby, then you haven't been adventurous enough in your food choices".

Another random pearl of wisdom came up somewhere in the next part of the discussion: When choosing a job, try to choose in which your progress at any given point is easily measured, as that's more like World of Warcraft.

We then spent time coming up with some exercises for the upcoming book - a list of questions on which we should hold off on proposing solutions, which ended up looking a lot like a list of Microsoft Interview Questions, including "what would I do with an empty skyscraper in New York if someone gave me one and I couldn't just sell it?".

The other two types of exercises were: ways to abuse rationality techniques, in order to try and see where they can be misused. Example: ask yourself what your beliefs predict; abuse - if there were a Fifth Column then there would be no evidence of a Fifth Column and questions on which lack of self-honesty may well deceive us, ranging in difficulty from "should I eat carrots or dirt?" through "will I actually stop after two oreos?" to "Am I a more moral person than average?".

After this, the discussion became a lot more general: from why Atlas Shrugged is bettre than Das Kapital because it's more obviously wrong to whether or not we can visualise 5 dimensional solids. All in all, a fun afternoon, and at least a few good ideas for exercises to figure out what rationality failure modes feel like.

No comments:

Post a Comment