A better yodeling contest

I really pity
Schroedinger's kitty

Paul McEvoy

One of my current side projects is an over-the-top article on a kind of anthropic computing problem on sending information back in time. In a nutshell, anthropic computing involves computing a random piece of data - if this meets our specifications, then good! Otherwise we destroy the universe (or perhaps just ourselves). Therefore an experimenter never observes any world-state in which the "wrong" random choice was chosen. In most formulations of this concept (see the Jargon File for example) the device is assumed to be "perfect" and so an infinite amount of information can be sent back in time. However this doesn't seem physically feasible since it seems it assumes the device has zero probability of error, whereas common failure modes (such as accidental deactivation or nuclear war) have probabilities relatively high compared to many computational problems, and therefore despite the canonical paper's opinion on the subject, it doesn't look like this kind of computing provides a loophole to compute NP-complete problems in polynomial time.

This is just a roundabout way of explaining why I happened to run across the cool webcomic/puzzle Meanwhile, which is an exploration of sorts into time travel, mindreading, and destroying the universe. Definitely worth a look (even if you didn't get that last paragraph).

Oh, and FYI the amount of information that can be sent back to the present is
\[\frac{1}{\log2}\left(\frac{\log\alpha}{\alpha-1}-1-\log\left(\frac{\log\alpha}{\alpha-1}\right)\right)\]
bits, where $\alpha$ is the ratio between the probability that the universe survives when our "gun" is fired, vs. the probability the universe survives if the gun is not fired.

Information given survival ratio $\alpha$

…although this concept doesn't seem to catch when I try explaining it to people. I've given the following explanation many times:

• game theory = economics
• combinatorial game theory = games

Granted, this is a somewhat simplistic view, but it does provide a good general idea. In any case, it doesn't seem to help.

Today, I was accosted in the hallway near my office to help someone work through a mixed-strategy (simultaneous move) game. Fortunately, it was simple enough I eventually knew how to do it, but I'm beginning to be worried people will start taking me as a fraud due to my inexpertise in the area, despite it not actually being my field. On a separate occasion this afternoon, someone assured me that a Nobel Prize in Economics had been awarded for "combinatorial game theory"; and again today, another individual emailed me a payoff matrix expecting my familiarity with it.

I suppose the problem of explaining your technical field is fairly common. Perhaps I needn't worry too much, since I probably know more about standard game theory than anyone prone to making this kind of mistake; however, seeing the same confusion three times in one day suggests I need to adopt a new strategy to deal with the situation.

I wouldn't mind learning enough game theory in the economic sense to keep up appearances, but I probably don't have time for that in the near future. So, I think I have no choice but to lie about the name of my field. Any mention of "games" will immediately be linked to popular game theory, no matter how much explaining I do. CGT could be seen as a subfield of "recreational mathematics", but then I might need to explain that it is, in fact, professional research. I might be able to get away with "combinatorics", or make up a name like "combinatorial, two-actor decision-tree analysis". What do you think?

(note: post written when tired, explaining the lacks of sense made)

Sam had apparently been studying shogi, and it took me a few games to figure out a response to his opening last night. When I did, it was a long, slow, crushing victory as my forces eliminated his rook-side before dropping his captured rook in for the kill. So, it was an encouraging end, even if I did lose most of my games.

Anyway, I just ran across a page describing Tai Shogi, probably the largest natural chess variant. The board measures 25×25, and each player starts with 177 pieces of more than 100 different types, including the Drunk Elephant and the Reclining Dragon. The royal piece is the "Emperor", who is also the most powerful, being able to move to any undefended square. Now, there's a game.

The famous "ear reddening move"

Originally uploaded by Kevin Saff.

There were something like eight new people who showed up to go club tonight, and Jason had them going on the first-capture game, an introduction that I somehow skipped when I was starting up. The idea is to demonstrate the basic concepts of go in a simpler game - just try to make the first capture on the board - passing is not allowed. So, that was some fun.

Better were two great games I played even, as black, and won. It seemed like it's been forever since I've won a game, or even been close. I still have a lot to understand about strength of groups in the opening, keeping my eye on the big picture, and not being careless in my reading. Tonight, I felt like I did reasonably well.

Oh, that picture is not one of my games. It's the "ear reddening" game won by the 17 year old Shusaku. You will probably never see me play moves anywhere near as amazing as in that game. Click on the picture for a brief recounting of the story behind this game.