drupal statistics module

Machines Like Us

Machines Like Us interviews: John Searle

Sunday, 15 March 2009

JS (continued): You point out that we simulate “weather and other physical processes.” You can do a computer simulation of any process that you can describe precisely. So we could do a computer simulation of the digestion of pizza down to the level of every molecule or even sub-atomic particles involved in digestion. Does anybody really believe that if we had done a computer simulation of the digestion of pizza on a computer that we can then rush out, buy a pizza, and stuff it into the computer and the machine will digest a pizza? Of course not. Or take another natural process you mention, the weather. We do computer simulations of the weather all the time, but nobody says that because we are going to do a computer simulation of a big rainstorm, that we should all bring umbrellas when we turn the computer on.

Once again, a model is not the real thing. The computer simulation is just that, it is a model or a picture. It is a simulation and simulation is not duplication. Now why not? What is the difference between the model and the real thing? Well, I mentioned that in my answer to question one. The computer simulation, whether it is rain, or whether it is rainstorms, digestion, or cognition, is all a matter of syntactical symbols, zeroes and ones, so the two fundamental principles underlying the answers that I have been giving you can be stated very clearly in two four word sentences. First, syntax is not semantics. And second, simulation is not duplication.

Years ago I baptized the view that computation by itself is sufficient for cognition as “Strong Artificial Intelligence” (Strong AI for short). It is important to realize how profoundly anti-scientific Strong AI is. The scientific approach is to treat the brain as a machine like any other; a biological machine to be sure, but all the same a machine. The machine processes in the brain are as much machine processes as the machine processes in the stomach and the digestive tract. The simulation of cognition on a computer stands to real cognition in exactly the same relation that the computation simulation of a rainstorm stands to a real rainstorm or the computational simulation of digestion stands to real digestion. There is a residual dualism in Strong AI because it does not treat cognition as a normal part of the natural biological world like any other biological phenomena. It treats it as computation and computation, to repeat a point made in answer to the earlier question, does not name a machine process defined in terms of energy transfer, it names an abstract, mathematical, algorithmic set of processes that we have found ways to implement on machines, but the process itself is not defined as a machine process.

MLU: You seem to be saying you have no problem with the idea that we could simulate a brain, and that the simulation could act like a human, at least in principle? I imagine this brain simulation shouting, "I really exist! Please don't turn me off!" and doing clever things to show it is conscious, like passing the Turing test, writing poetry, proving mathematical theorems, phoning people and fooling them into thinking it is actually the person from whose brain it was derived or debating against you and I can imagine you being utterly disinterested and turning it off anyway, if you had no practical use for it, because, in your view, it is a very good simulation of a mind but there is no mind really there: nobody is home.

Before we continue, some of our readers will know about Sir Roger Penrose's views on AI. Penrose, like you, is skeptical about AI, but we should be clear here that his views are very different to yours. Penrose thinks brains rely on "non-computable physics" and he would say that trying brain simulation like this would not even result in the correct behavior, because computation cannot even produce that, whereas you have no problem with the idea of the correct behavior being produced but have no reason to think that the simulation would be anything other than a kind of "zombie" -- an imitation of a mind. This is different to the common view of AI enthusiasts who would say that it is impossible to produce a zombie like this: they would say that if you can make something that acts conscious then it is conscious.