Artificial Intelligence

What nonsense. Supposedly we are to cower in fear because AI robots will first of all take all our jobs and then rule us. While I don’t want to take away from the skill and expertise of the developers, the sophisticated software that is driving cars around is no more intelligent than this toaster. It can recognize properly cooked toast.

The novel technology that has enabled so-called AI in recent years is neural network based pattern recognition. Often called “deep learning,” it allows, for example, vision systems to recognize objects and audio systems to recognize speech. Powerful and useful stuff indeed, and a true breakthrough, but the resemblance to human intelligence stops there. Then the programmer must take over.

The reality of this is well explained in this article, which describes the approaches taken to win Amazon’s Alexa Prize, a $1 million competition to build a chatbot capable of carrying on a 20 minute conversation. The point is not the failure, but that the approaches taken were basically text manipulation, not understanding. Intelligence requires semantics, not just syntax.

Autopilots and flight management systems (FMS) have been safely navigating aircraft from takeoff to roll-out for years. Yes, ATC keeps them clear of traffic although they do warn of potential conflicts. The real enhancement that enables this kind of technology to be extended to the highway is the ability to recognize objects in the car’s environment. The car’s response to identified conflicts, though, has to be programmed case-by-case in the old-fashioned way.

If your work requires intelligence, you have nothing to worry about. Perhaps more breakthroughs will be made. However, Sir Roger Penrose has speculated that the brain is actually a quantum computer. In which case you probably don’t have to worry for quite a long time.

Both comments and trackbacks are currently closed.
%d bloggers like this: