Cambridge scientists have shown that placing physical constraints on an artificially-intelligent system – in much the same way that the human brain has to