Human vs Computer chess contest

A simple way to "handicap" the computer in a human versus computer chess match (while still maintaining the appearance of fair playing conditions) is to constrain the computer to be deterministic. That is, if in one game it calculates a certain move, then if a successive game in the match reaches the same state, it must make the same move as it did previously. In other words, the computer will never choose to deviate from a previous game. A simple way to implement this would be to add each move to the computer's opening book that it will use in successive games. The human may, and in fact ought to, deviate.

Let this be the basis of a chess contest with a prize. To enter the contest, one pays $10 to set up an account. Using that account, one plays a series of games against the computer. Whoever defeats the computer first, wins. The computer computes each response to the many simultaneous games it is playing in some fair queueing round-robin manner. Multiple computers may be used if there are lots of contestants. The computer maintains a global "opening book" for all the moves it has made across all games. If a contestant makes a move that already has a book response (possibly because some other contestant has already tried that move, or it was in the original opening book), the response is returned immediately. Otherwise the move is placed on the queue and an estimated time the computer's response will be ready is printed, based on the current length of the queue.

Salt dust

How finely can sodium chloride be ground? Two atoms?

Text messaging

The names in my contacts list ought to be added to the predictive text dictionary.

LOTR story idea

The One Ring has been unearthed in the modern day, having been encased in volcanic rock for thousands of years.

Geodesic Dome

I finally realized why a geodesic dome is called that. The points inside each triangular icosahedral face are constructed on the intersections of geodesics connecting equally spaced points on the edges of each icosahedral face. It is not obvious that three geodesics intersect at a point. These geodesics cause forces to be purely compressive along the edge structural members, probably explaining the dome's strength.

Could a geodesic dome be constructed from an octahedron instead of an icosahedron?

Milk

I wish a device that will tell me if my milk is spoiled. And a way to test it at the supermarket.

Cell music composer

I'd like to see a cellphone app for arranging music, combining an MP3 player that gives control of what segment of a song to play, and a Lilypond or Finale-like musical notation input editor and synthesizer.

A cell phone piano could be part of it. Seven number buttons are used for the scale, two for accidentals, the remaining three for changing octaves or re-centering the scale on a certain key.

Super ko

The various different rulesets regarding how to deal with super-ko in the game of go 囲碁 bother me. No-result or replay-game is bothersome because the game might be decided regardless of the outcome of the super ko. In positional super ko, if a different player is to move in the repeated position, the game is guaranteed to become different on the next move. The amount of hidden state theoretically required to implement situational or positional super ko is infeasibly enormous and detracts from the concept that the state of the game should be mostly encoded in the visible board position, not some potentially enormous list of forbidden previous positions.

My radical solution is, simply stated, let the game never end. Practically, a game is scored as follows: When both players agree that they are trapped in a super ko situation that no further progress may be made, the game is scored as the average score of all positions being visited in the super ko. If it is a loop, it is simply the average of all the positions in the loop. Theoretically, the players should always be able to agree on what the loop is (one does not need to play mixed strategies in games of perfect information), but if they cannot agree, Markov-chain Monte Carlo may be used to average over a graph.

As a shortcut, one need only bound the average enough to determine the outcome of the game. The average need not be calculated exactly.

A group left on the board that would normally be marked as dead achieves eternal life as players trapped in a loop never get a chance to kill it off.

The average can conceivably be any arbitrary fractional number, so komi should be an irrational number to avoid a tied score.

fMRI rap

Take an fMRI of someone improvisationally rapping. The brain is simultaneously creating rhyme and reason (meaning).

--
==================================================================
This mobile text message is brought to you by AT&T