Thursday, February 17, 2011

Human vs. Artificial Intelligence

Last night MIT Enterprise Forum in Chicago hosted “Debate Night: Human vs. Artificial Intelligence.”

Unfortunately, Artificial Intelligence failed to show up and instead, was represented by a very intelligent bit of carbon and water named Kristian Hammond, a Professor of Electrical Engineering and Computer Science at Northwestern University who fancies himself as “a machine, or else I don’t know what to think.” Representing Human Intelligence was the delightfully non-machine-like Valerie Pajak, a manager of Business Intelligence and Analytics at APP.

The television show Jeopardy was kind enough to schedule their Watson competition to wrap around the MITEF event, and it was a topic of conversation for the group. The breakthrough for Watson, was that rather than rely on a set of programming steps, the driver of Watson is ‘search’.

Whether it is search, or programming steps around an expert system, or even a set of genetic algorithms, what passes for artificial intelligence to date has been limited to retrieving relevant existing knowledge. This is an important part of human intelligence and, as such, has been improving for hundreds of years. The index at the back of a text book is a form of this sort of intelligence. And while it is an important part of human intelligence, it is only important because of what it enables. Humans dominate the earth not because we can retrieve knowledge, but because we can create new knowledge. And since no new knowledge is ‘whole cloth’ it always depends upon some level of prior knowledge. Existing knowledge is the foundation of new knowledge.

But to create new knowledge, the useful existing knowledge must be in the brain of the human who creates the new knowledge. We create new knowledge in three ways:

  1. Discovery - We stumble across something that provides a solution to a problem we have.
  2. Experimentation - We are pretty sure we can find the right solution if we just try a whole bunch of alternatives.
  3. Synthesis - We have a problem stored in memory, and we’re probably not actively thinking about it when something magical happens, and, voila! a solution pops out of (or is it ‘into’) the brain.

What will truly make Artificial Intelligence beat human intelligence is when the computer can mimic whatever it is that happens in our brains when we’re taking a shower, or out for a long walk, or we’re in bed at 3am dreaming, and that problem solving insight explodes and we create new knowledge. When a computer does that, maybe someone should pull the plug so we can figure out how bad it could get.

Post a Comment