A twist on artificial intelligence: DNA

Milan Stojanovic has a new tic-tac-toe adversary in his Columbia University laboratory, and after a hundred matches he has yet to beat it. "There's no way," he sighs.

Who is this tic-tac titan? A computer, of course. Only here's the twist: its brain is based not on silicon microchips but on molecules of DNA.


Strange as it seems, the double helix is turning into a growing draw for computer scientists. DNA is tiny - a trillion molecules squeeze into a single drop of water. It performs chemical reactions at blinding speed. And it holds enormous amounts of information. In fact, 1 gram of the stuff can store as much as 1 trillion compact discs, scientists estimate.

But since the first DNA computer was devised in 1994, the devices have remained scientific sideshows - error-prone, unresponsive and capable of solving only obscure problems. Now, Stojanovic and computer scientist Darko Stefanovic of the University of New Mexico have licked at least one of those issues by creating the first interactive biocomputer. MAYA, as they've dubbed their tic-tac-toe device, is described in the latest edition of Nature Biotechnology.


The straightforward logic of tic-tac-toe has made it a favorite of computing pioneers for more than 150 years. British mathematician Charles Babbage, often called the father of modern computing, dreamed up a mechanical tic-tac-toe player in the 1840s. The creators of EDSAC, an early stored-program computer, coded the first electronic version in 1949.

Nobody who sees MAYA would mistake it for a traditional computer or video game, Stojanovic said, chuckling. More glassware than Game Boy, the board consists of nine wells, arranged in the classic three-by-three pattern. Snippets of custom-designed DNA float in each well. These, says the scientist, are the secret to MAYA's smarts: The strands are crafted to anticipate each possible move a player can make.

"Now, let's say you want to play," Stojanovic begins. Players first select from a series of test tubes, each containing DNA designated for a specific square. Then they dribble the tube's contents into all nine wells, not just the one they intended to mark. As the strands chemically combine, one of the squares will gradually glow green. This is MAYA's move.

To simplify the experiment, the scientists have set ground rules. MAYA always gets to go first and is always assumed to pick the middle square. Humans, meanwhile, must select as their first move either the upper left-hand corner or the square below. After that, anything goes.

Of course, as tic-tac-toe logicians know, these restrictions ensure that MAYA can't lose - and that human players quickly grow bored.

"The first time it was fun," Stojanovic says. Afterward, "it requires some focus."

But Ehud Shaprio, a computer scientist at the Weizmann Institute of Science in Rehovot, Israel, says games are not the ultimate goal of scientists pursuing biocomputing. "The next thing is not to scale up to checkers," he says.

One of Stojanovic's underwriters is NASA. The space agency, he says, wants to develop molecule-sized biocomputers capable of diagnosing and perhaps even repairing disease at the cellular level - perfect for astronauts on long spaceflights and something silicon devices could never do.


"This is just a small first step," Stojanovic says.