Gigaom AI Minute – October 11

Transcript

Researchers disagree about many of the most basic issues surrounding an artificial general intelligence. There is the question of what approach should be used. Should we build an AGI by modeling the human brain? Brains and computers are just so very different from each other. They are each great at what they do and pretty terrible at what the other one does. A computer simulating a human brain process is as inefficient as a human thinking in machine code. So perhaps it is a better approach to ignore human biology entirely. Heavier than air flight was delayed for a long time because we tried to model machines after birds. Even Leonardo da Vinci fell into this trap.

Next is the question of whether intelligence is, at its root, a set of relatively simple principles that build up into complex intelligence through iteration. Or is intelligence a huge conglomeration of a range of different skills that come together to form intelligence? Watson, the IBM super computer that defeated Ken Jennings at Jeopardy was built using this latter approach. According to IBM, "more than 100 different techniques are used to analyze natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses." In short, is our brain just a giant hack?