Oddhead Blog

Until just reading about it in Wired, I knew little¹ of the apparent suicide of Push Singh, a rising star in the field of artificial intelligence.

Singh seemed to have everything going for him: brilliant and driven, he became the protégé of his childhood hero Marvin Minsky, eventually earning a faculty position alongside him at MIT. Professionally, Singh earned praise from everyone from IEEE Intelligent Systems, who named Singh one of AI’s Ten to Watch (apparently revised), to Bill Gates, who asked Singh to keep him appraised of his latest publications. Singh’s social life seemed healthy and happy. The article struggles to uncover a hint of why Singh would take his own life, mentioning his excruciating chronic back pain (and linking it to a passage on the evolutionary explanation of pain as “programming bug” in Minsky’s new book, a book partly inspired by Singh).

The article weaves Push’s story with the remarkable parallel life and death of Chris McKinstry, a man with similar lofty goals of solving general AI, and even a similar approach of eliciting common sense facts from the public. (McKinstry’s Mindpixel predated Singh’s OpenMind initiative.) McKinstry’s path was less socially revered, and he seemed on a never ending and aching quest for credibility. The article muses whether there might be some direct or indirect correlation between the eerily similar suicides of the two men, even down to their methods.

For me, the story felt especially poignant, as growing up I was nourished on nearly the same computer geek diet as Singh: Vic 20, Apple II, Star Trek, D&D, HAL 9000, etc. In Singh I saw a smarter and more determined version of myself. Like many, I dreamt of solved AI, and of solving AI, even at one point wondering if a neural network trained on yes/no questions might suffice, the framework proposed by McKinstry. My Ph.D. is in artificial intelligence, though like most AI researchers my work is far removed from the quest for general AI. Over the years, I’ve become at once disillusioned with the dream² and, hypocritically, upset that so many in the field have abandoned the dream in pursuit of a fractured set of niche problems with questionable relevance to whole.

Increasingly, researchers are calling for a return to the grand challenge of general AI. It’s sad that Singh, one of the few people with a legitimate shot at leading the way, is now gone.

Push Singh Memorial Fund

Mathematicians, third graders, and talkative defense department computers alike all know that there is an infallible way to play tic tac toe. A competent player can always force at least a tie against even the most savvy opponent.

In the July issue of Science, artificial intelligence researchers from the University of Alberta announced they had cracked the venerable game of checkers in the same way, identifying an infallible strategy that cannot lose.¹

It doesn’t matter if the strategy is unleashed against a bumbling novice or a flawless grandmaster, it can always eke out at least a tie if not a win. In other words, any player adopting the strategy (a computer, say) makes for the most flawlessy grandmasterest checkers player of all time, period.

The proof of correctness is a computational proof that took six years to complete and was twenty-seven years in the making.

Tic tac toe and checkers are examples of deterministic games that do not involve dice, cards, or any other randomizing element, and so “leave nothing to chance”. In principle, every deterministic game, including chess, has a best possible guaranteed outcome² and a strategy that will unfailingly obtain it. For chess, even though we know that an optimal strategy exists, the game is simply too complex for any kind of proof — by person or machine — to unearth it as of yet.

The UofA team’s accomplishment is significant, marking a major milestone in artificial intelligence research. Checkers is probably the first serious, popular game with a centuries-long history of human play to be solved, and certainly the most complex game solved to date.

Next stop: Poker

Meanwhile, the UofA’s poker research group is building Poki, a computer player for Texas Hold’em poker. Because shuffling adds an element of chance, poker cannot be solved for an infallible strategy in the same way as chess or checkers, but it can in principal still be solved for an expected-best strategy. Although no one is anywhere near solving poker, Poki is probably the world’s best poker bot. (A CMU team is also making great strides.)

Poki’s legitimate commercial incarnation is Poker Academy, a software poker tutor. An unauthorized hack of Poker Academy [original site taken down; see 2006 archive.org copy] may live an underground life as a mechanical shark in online poker rooms. (Poki’s creators have pledged not to use their bot online unidentified.)

Poker web sites take great pains to weed out bots — or at least take great pains to appear to be weeding out bots. Then again, some bot runners take great pains to avoid detection. This is a battle the poker web sites cannot possibly win.