Getting leech neurons to "think" is the basis of a new computational system being developed by a team led by Georgia Institute of Technology professor of Bioengineering Bill Ditto. By adapting the concepts of chaotic computing, Ditto and his colleagues plan to create a computer that uses leach neurons to perform nonlinear calculations. So far, the team has successfully prompted two leech neurons in a petri dish to add six plus two.

Leech neurons are being used because they are large and relatively well characterized. Ditto and Ron Calabrese, a leech expert at Emory University, Atlanta, intend to have neurons doing computations on a silicon substrate within three years, followed by hybrid neural–silicon chips performing complicated computing problems such as pattern recognition.

Ditto's goal is to develop computers many times smarter than current digital machines. "Why not take advantage of the millions of years of evolution of neurons and see if we can incorporate them into a dynamical computer," he asks. Dynamical computing arrives at answers by organizing itself, under reproducible conditions. "If the brain needs to solve a complicated problem, it doesn't go out and buy faster neurons, it just throws more neurons at the problem and connects them in more creative ways," he explains.

The team accomplished the simple 6 + 2 addition by attaching electrodes and introducing a current that changes the resting voltage of the cell. By varying voltage, the firing pattern of the neuron can be changed, and numbers can be associated with these changes.

By mimicking the operation of the brain, Ditto hopes to come up with a system that, much like a person, is adaptable. Yet introducing the human trait of adaptability may also introduce the frailty of temperament, and controlling a feisty leech computer could be a problem.