The bionic chess interface

Brutus

Initially the goal was to build a very fast chess program, running on FPGA hardware. The acronym stands for "Field Programmable Gate Arrays", which is essentially a programmable chip. In 2001 Dr Christian ("Chrilly") Donninger began writing chess playing code for FPGA use. The advantage is that anything programmed this way will run very much faster than on a general purpose chip like the Pentium or Athlon.

An additional benefit of using FPGAs is that it can be implemented on very small systems, consisting essentially of a pea-sized chess chip with few electronic components surrounding it. But it will be more powerful than the fastest desktop computers

The bionic interface

All this led to a new idea. In 2002 Dr Donninger was contacted by the US National Science Foundation and other defence agencies who had been working on a project to interface the brain directly with computer equipment. Chess appeared to be an ideal testing ground, since the bandwith required for the communication between brain and computer is very small (a chess move can be theoretically encoded in a single byte).

The scientists working on the project decided that the hippocampus, a portion of the temporal lobe, was the ideal location for a "bionic interface". The hippocampus is named after the seahorse, because of its shape, and it is normally the main relay station that determines whether a new memory should go into long-term storage or be deleted after its short-term usefulness is over. In this respect it is in constant contact with the higher regions of the brain.

In order to build a hippocampus interface the scientists had to overcome three three major hurdles. First of all they had to devise a mathematical model of how the hippocampus operates under all possible conditions, build that model into a silicon chip, and then interface the chip with the brain.

Initially the scientists were not able to understands how the hippocampus encodes information. So the team decided simply to copy its behaviour. In experiments performed by Theodore Berger at the University of Southern California in Los Angeles the hippocampus of rats were stimulated with electrical signals, millions of times over, until the researchers could be sure which electrical input produced a corresponding output. The results yielded a mathematical model of the entire hippocampus. The work was funded by the US National Science Foundation, Office of Naval Research and Defense Advanced Research Projects Agency.

With Dr Donninger's assistance the scientists then programmed this model onto a chip, which in a human would initially sit on the skull rather than inside the brain. The chip communicates with the brain through two arrays of electrodes inserted into the hippocampus. One records the electrical activity coming in from the rest of the brain, while the other sends appropriate electrical instructions back out to the brain.

Schematics of the hippocampus interface

The chess chip that Dr Donninger is designing to interface with a human brain consists of two separate components: one electrode array registers signals coming from other parts of the brain, representing a chess position. These are transmitted through very thin wires to the (initially external) chess chip, which processes the position. The results, in form of a move and auxillary data, such as evaluation, search depth, next best move, are transmitted through the second pair of wires to the second electrode array, loctated is a different part of the hippocampus, from where it is transmitted to other parts of the brain.

The electrodes that are inserted into the hippocampus are microscopic in size and delivered to their final place through a thin injection needle. The skull and cranium cavities are not opened, the entire procedure is performed under local anasthetics.

Interacting with the computer

In principle the wearer of a hippocampus bionic interface should, with a little practice, be able to communicate quite efficiently with the external chess chip. Essentially only a very small amount of information needs to be communicated by the brain through the interface to the chip. "We expect the total information flowing through the interface for each 'seek' will be about 32 bits, at the very most 64," says Dr Donninger. "This is easily handled by the hippocampus interface."

The same applies to the response channel: after doing a prodigeous amount of computing the chess chip sends a very small package of information, which is translated by the receiver electrode array into a signal that the higher regions of the brain interpret as the visual image of a chess move.

Initial tests with rats

Tests with the chip in live rats have begun, with very encouraging results. "The real proof will be if the animal's behaviour changes or is maintained," said Sam Deadwyler of Wake Forest University in Winston-Salem, North Carolina, who conducted the tests. Deadwyler says the hippocampus has a similar structure in most mammals, so little will have to be changed to adapt the technology for human beings. But before human trials begin, the team will have to prove unequivocally that the brain-computer interface is safe.

Are there ethical issues involved in this "bionic" connection between man and machine? Bernard Williams, a philosopher at Britain's University of Oxford, points out that it may take time for people to accept the technology. "Initially people thought heart transplants were an abomination because they assumed that having the heart you were born with was an important part of who you are."

A full-fledged working experiment with human beings has not been attempted, so it is not clear that everything will go according to plan. But there is reason for optimism. In the UCLA laboratories the scientists have installed the first bionic hippocampus interface into the brains of rats. The prcedure caused only brief discomfort to the animals, who were running around their cages just an hour after the procedure was completed.

Naturally the rodents are not able to send any meaningful requests through the hippocampus interface to the chess chips. "What we receive is essentially noise," says Dr Donninger. But the chess chip tries to generate random positions that match the signal patterns received from the brains of the rats as closely as possible. At set intervals genuine chess moves are sent through the return channel to the brains of the rats.

So obviously the rats are not able to play chess, or even execute a single legal move. That is not the goal of the experiment. However it is interesting to note that rats that have the interface installed and are conneced with the external chess chip will show a clear interest in a chess board placed within their range of vision (see picture above). This does not apply to rats who have not be so treated.

So how long will it be before the first real experiments with human beings can begin? Dr Donninger admonishes to caution. "There is still a lot of work to be done, both on the interface itself, the communications protocols and the chess algorithms on the external chips. I don't believe we will have the first working prototypes for human subjects ready before the beginning of the next year."

Naturally the hippocampus chess interface will not require the user to be connected by wires to a computer. Donninger envisions a small package consisting of a chess engine running in matchbox-sized case. The interface itself will consist of a receiver that can be made smaller than a standard hearing aid. The two components will use the bluetooth protocol to communicate, making the whole setup completely unobstrusive. "Pacemakers and hearing implants are gigantic compared to the low-bandwidth systems we will be using," said Donninger. Since electrode arrays use in the hippocampus implants contain only microscopic amounts of metal there is no danger that they will cause medtal detectors like the ones used at airports to sound alarm.