The most recent example is truly gratifying because there's an organ in your brain called the hippocampus – which is right here on the lateral margins and inferior portion of your so-called temporal lobe, and this is called the hippocampus or the seahorse because if you cut it and section it half across, it looks like a seahorse in its shape. It is the part of your brain necessary for long-term memory, and this particular... you may have seen a film called Memento. Memento is about a chap who had what these people have – an operation to keep him from fatal epilepsy, in which both hippocampus or hippocampi were removed. Now what happens to a patient who has that is very interesting; first shown by Brenda Milner at McGill University in Canada a long time ago, the patient is freed from his epilepsy to a large extent but the patient can no longer incorporate short-term into long-term memory. So up to the time of the operation they're fine. After the operation you come in and talk to them, they're conscious, they deal with you, you interact with them, but when you move out of the room and come back a minute later they don't know who you are; they cannot put down long-term memories. So we actually simulated a hippocampus in one of our Darwin automata, Darwin 10, and what we did is we had it perform a task in which it located a target in a room, according to what was in the room.

Now, what a patient with no hippocampi can't do is have episodic memory. It cannot remember sequences of signals or sequences of recognizable events. What we did is we, based on the neuroscientific research done in many laboratories, made a model of the hippocampus, put it into this brain-based device and had it challenged by a task which is based on something called the Morris water maze. Morris is a Scottish psychologist who's a member of NRP, by the way, who invented the following idea. He put a rat inside milky water underneath which was a hidden platform. Rats do not like water and so they swim around a while and when they find the platform they stay there. Meantime, however, they're registering everything they see, so the next time you put the rat in he goes straight for the platform on the basis of his episodic memory of those cues. Well, we made a dry version – we don't want to put our Darwin automaton into water. We made a dry version in the pen of a round area which has the same black color but different reflectance, and we allowed the value system of Darwin 10 to respond when it's over that. We then put it in and at random it explored, seeing these signals from outside – different colors on the wall, different stripes – and when it finally found the platform and stayed there and got a positive value, it would remember all of these things, so no matter where you put it, it would go in a very much more direct fashion for the podium. This is really quite exciting because it said to us, aside from computer power, there's probably no limit to which we couldn't bring brain-based devices.

Now, what did we learn? The brain-based device is not a real brain. However it is based on the rules that we understand brains to be and we look at every single thing that happens, from the molecule to the behavior, and so this is, I think, a very definite advance. What it's also advanced to... will be is in computing. Computers cannot be pattern recognizers. Let me open that up a little bit. Computers can do theorems – they can even prove theorems – but they can't select axioms. Computers cannot deal with novelty. You have to program them on the basis of what you know; they work by logic and they work by a clock. But these Darwin devices are not that at all. They learn by making mistakes. They don't crash by making mistakes; they're more like us. And so imagine that some day we may be able to construct a new kind of hybrid machine which consists of a Turing machine or a computer which has all this data and all this logic, hooked to a brain-based device which is exploring a set of novelties, so the two speak to each other the way we speak to computers. That would be I think a very, very important jump ahead in our ideas about how you learn things and how you make patterns and how you relate patterns and creativity to logic.