In contemplating whether the brain is a good model for machine intelligence (Nature 482, 462–463; 2012), I believe that Alan Turing's principle that the brain performs computations will continue to hold true. But it seems clear from the state of machine intelligence today that we're missing some basic insight into the language of brain computation.

Advances in genetics, electronics and optics are now enabling us to look into simple brains while they're living and behaving, and to observe the activity of every neuron in real time. This influx of information should eventually help us to make intelligence comprehensible and replicable.

A good approach to interpreting these data is the re-creation of neural processing in simulation, from sensation all the way to behaviour. But once we understand what's going on, it should be possible to create machine intelligence that doesn't rely on reproducing low-level chemical dynamics.

Over the coming decades we are likely to take high-level cues from biology on how to organize our silicon. Mind is merely a function of the brain, however, so we should be able to capture that function in any Turing-complete system once we know how the brain is organized to perform it.