Can We Learn Anything from Brain Simulation?: A Hermeneutic Case Against Strong AI

Résumé En

If you figure out how machines learn, then you will figure out how the brain works, and what the brain’s functions are. Such an idea is widespread among philosophers and computer scientists who agree with a functionalist reductionist point of view of consciousness. This theory leads to hold that the more accurate the simulation of cognitive behavior is, the more the math behind it must be true – when true means what really happens in our brain. In this article, we aim to show that, on one hand, brain simulation is nothing more than just another simulation, and it offers very little help to understand – nor to produce – the vivid experiences (qualia) of cognitive functions. On the other hand, we would like to emphasize that when it succeeds at predicting a mechanism with less ambiguity and more accuracy than without a simulation nor direct observation, it really develops the knowledge of our brain. As long as brain simulation follows scientific principles, it should be regarded as valuable, even though the knowledge it brings to science must not be confused for the real phenomenon. Brain simulation, like all simulation, cannot fill any reality or epistemic gap. It is a consolation prize.