How are you pursuing the creation of conscious artifacts in your work at the Neurosciences Institute?
We construct what we call brain-based devices, or BBDs, which will be increasingly useful in understanding how the brain works and modeling the brain. They may also be the beginning of the design of truly intelligent machines.
What exactly is a brain-based device?
It looks like maybe a robot, R2-D2 almost. But it isn’t a robot, because it’s not run by an artificial intelligence [AI] program of logic. It’s run by an artificial brain modeled on the vertebrate or mammalian brain. Where it differs from a real brain, aside from being simulated in a computer, is in the number of neurons. Compared with, let’s say, 30 billion neurons and a million billion connections in the human cortex alone, the most complex brain-based devices presently have less than a million neurons and maybe up to 10 million or so synapses, the space across which nerve impulses pass from one neuron to another.
What is interesting about BBDs is that they are embedded in and sample the real world. They have something that is equivalent to an eye: a camera. We give them microphones for the equivalent of ears. We have something that matches conductance for taste. These devices send inputs into the brain as if they were your tongue, your eyes, your ears. Our BBD called Darwin 7 can actually undergo conditioning. It can learn to pick up and “taste” blocks, which have patterns that can be identified as good-tasting or bad-tasting. It will stay away from the bad-tasting blocks, which have images of blobs instead of stripes on them —rather than pick them up and taste them. It learns to do that all on its own.
Why is this kind of machine better than a robot controlled by traditional artificial intelligence software?
An artificial intelligence program is algorithmic: You write a series of instructions that are based on conditionals, and you anticipate what the problems might be. AI robot soccer players make mistakes because you can’t possibly anticipate every possible scenario on a field. Instead of writing algorithms, we have our BBDs play sample games and learn, just the way you train your dog to do tricks.
At the invitation of the Defense Advanced Research Projects Agency, we incorporated a brain of the kind that we were just talking about into a Segway transporter. And we played a match of soccer against Carnegie Mellon University, which worked with an AI-based Segway. We won five games out of five. That’s because our device learned to pick up a ball and kick it back to a human colleague. It learned the colors of its teammates. It did not just execute algorithms.
It’s hard to comprehend what you are doing. What is the equivalent of a neuron in your brain-based device?
A biological neuron has a complex shape with a set of diverging branches, called dendrites, coming from one part of the center of the cell, and a very long single process called an axon. When you stimulate a neuron, ions like sodium and potassium and chloride flow back and forth, causing what’s called an action potential to travel down the neuron, through the axon, to a synapse. At the synapse, the neuron releases neurotransmitters that flow into another, postsynaptic neuron, which then fires too. In a BBD, we use a computer to simulate these properties, emulating everything that a real neuron does in a series of descriptions from a computer. We have a set of simple equations that describe neuron firing so well that even an expert can’t tell the difference between our simulation spikes and the real thing.
All these simulations and equations sound a lot like the artificial intelligence ideas that haven’t been very successful so far. How does your concept for a conscious artifact differ?
The brain can be simulated on a computer, but when you interface a BBD with the real world, it has the same old problem: The input is ambiguous and complex. What is the best way for the BBD to respond? Neural Darwinism explains how to solve the problem. On our computers we can trace all of the simulated neuronal connections during anything the BBD does. Every 200 milliseconds after the behavior, we ask: What was firing? What was connected? Using mathematical techniques we can actually see the whole thing converge to an output. Of course we are not working with a real brain, but it’s a hint as to what we might need to do to understand real brains.
When are we going to see the first conscious artifact emerge from your laboratory?
Eugene Izhikevitch [a mathematician at the Neurosciences Institute] and I have made a model with a million simulated neurons and almost half a billion synapses, all connected through neuronal anatomy equivalent to that of a cat brain. What we find, to our delight, is that it has intrinsic activity. Up until now our BBDs had activity only when they confronted the world, when they saw input signals. In between signals, they went dark. But this damn thing now fires on its own continually. The second thing is, it has beta waves and gamma waves just like the regular cortex—what you would see if you did an electroencephalogram. Third of all, it has a rest state. That is, when you don’t stimulate it, the whole population of neurons stray back and forth, as has been described by scientists in human beings who aren’t thinking of anything.
In other words, our device has some lovely properties that are necessary to the idea of a conscious artifact. It has that property of indwelling activity. So the brain is already speaking to itself. That’s a very important concept for consciousness.