INTERVIEW
Human, machine and evolutionary learning
Dr Richard Watson is an associate professor at the University of Southampton. He works at the interface between computer science and theoretical evolutionary biology
Just because we’re made out of different stuff doesn’t mean that humans learn in a different way to computers. Dr Watson’s work on expanding the theory of evolution has given him great insight into the way animals, machines and even evolution itself learns and adapts – and he has come to some quite staggering conclusions.
Can you explain how learning works?
One of the key insights of neural models of brain learning is that intelligence is not in the neurons – it’s in the organisation of the connections between them. The way that they learn is by changing the strengths of these connections.
There’s one particular model of how connections change which is sufficient to do some really interesting behaviours, and that is Hebbian learning. It says that if two neurons fire at the same time, or in quick succession, then the strength of the connection between them is increased. But if one fires without the other, the connection is decreased.