Thinking about the unthinkable

This morning’s Observer column.

Then Google launched its autonomous vehicle (aka self-driving car) project. By loading a perfectly ordinary Toyota Prius with $250,000-worth of sensors and computing equipment, the company created a vehicle that can safely navigate even the more congested road conditions. So far, these cars have logged something like half-a-million accident-free miles, which implies that robotic cars are actually far safer than ones driven by humans.

For me, the implication of the Google car is not necessarily that Kurzweil’s “singularity” is near, but that our assumptions about the potential of computers – and, therefore, artificial intelligence – urgently need revising. We need to think seriously about this stuff, along the lines demonstrated by the philosopher David Chalmers in a terrific paper, or by Erik Brynjolfsson and Andrew McAfee in their book, Race Against the Machine.