Genetic Tic Tac Toe With Bacteria

Researchers used the programmable DNA techniques of DNA origami to fashion a microscale game board out of X's and O's. They then played a microscale match of tic-tac-toe with bacteria. The scientists - including Caltech postdoc Grigory Tikhomirov BS '94 and his colleagues - genetically modified E. coli to act as a simple neural network, a form of artificial intelligence, to play the game against a human player.

The system modeled the evolution of an organism using how well it plays tic-tac-toe as a fitness function. Essentially the algorithm punishes invalid moves (moving where there is already a mark) and rewards winning moves by rewarding the parent with more genes, and redistributing the remaining genes amongst the offspring. The genes are determined by a weighing system that determines the optimal move to make depending on the odds of the two opposing players making an L-shaped fork, or the odds of both players getting an X and an O in the same row or column.

The program included a DNA library, a set of rules for playing the game and rules for life and reproduction, a display mechanism and an accounting/evaluation system to track the gene pool of the evolving automatons and their win/loss record. The system also included a decision tree to find the best move to make based on an evaluation of all the possible combinations of moves, a weighted score that sums up the probabilities of each move being good or bad, and a GeniePop class that contained all of the data for the Genie population including the geneCount, pWeighting, leDorVaDor(), maybe() and their LifeScores.