AI has always intrigued me. The idea of recreating human or at least intelligent behavior with software is an amazing idea. Unfortunately modern AI is nothing like the dream of creating human behavior. Sure the current solutions can solve some pretty cool problems, but its not the types of problems that people think of when they dream of AI. At h+ Magazine an article was posted discussing what the experts of the AI field predict to happen over the next century. Most seem to think we are a long way off from even being able to pass a 3rd grade exam and possibly 100 years off from making smarter than human AI.

The thing about most AI style problems is that we can solve them in a much easier ways without AI. Passing a 3rd grade exam could be done with advances in natural language processing and some clever Internet searching. It turns out to be very difficult to find ways to test AI. Even the famous Turing Test may not be as satisfactory as one might think. Its hard for a human to tell even if they are talking to another human even if they are, in fact, talking to another human.

The reason why it is so hard to identify another human during the Turing Test is because we don’t know what we expect from another human. It comes down to the fact that we have no idea what makes us human. Its nearly impossible to describe a humans intelligence and identify the intelligent parts. So it would be very difficult to quantify how an AI system is better than a human, let alone equal.

I think before we are able to make any true AI system we first need to figure out how the human mind works. Which is turning out to be much easier said than done. Every time we figure something out about the brain, it creates at least ten more questions about the brain. To put this problem in some sort of perspective: We had a theory of computation decades before the first computer hardware and we had predictive models of space centuries before we had the ability to test them. However we have the hardware to run our AI systems on, but no theory as to how to do it. So I think we are at least half a century away from having anything near what we dream AI to become.