Since the early 1950s MIT professor Marvin Minsky has been researching ways to infuse computers and machines with human intelligence and emotion. In 1951 he built the world's first neural network computer. In 1959 Minsky co-founded the MIT Artificial Intelligence Laboratory with John McCarthy. He was also an advisor to director Stanley Kubrick for the film 2001: A Space Odyssey. In short, the entire field of artificial-intelligence research has been shaped by Minsky's research, writings, and innovative ideas.
On tonight's episode of "The Screen Savers" Minsky discusses present and future developments in the field of AI. Our PC Prophecy special simply wouldn't be complete without Minsky's views of the future.
Interview with Marvin Minsky
TechTV: Back in 1951 you built SNARC, the first neural network simulator --
Marvin Minsky: Yes, it was a
learning machine that simulated a few neurons interconnected by about 40 devices that imitated what some synapses do. It started by doing things at random. Then, by "rewarding" it for doing things that you liked, it could learn certain kinds of behavior. However there were other things that it couldn't learn, and it took almost 20 years to find out the causes of some of its limitations.
TechTV: The
MIT Artificial Intelligence Lab is almost 50 years old. How different is AI research now from when the lab was established?
Minsky: Research in the first few years progressed at an amazing pace. But around 1980 it greatly slowed down. I think that this was because most researchers kept trying to find some single, wonderful way to make a machine solve all kinds of hard problems. The trouble is that good thinking requires many different methods. So
my present research is focused on how to design a larger machine that has and makes use of a great deal of knowledge about which methods to use on each part of each problem.
TechTV We're not getting much smarter (brain synapses) and not living longer. Are we a "failed species"?
Minsky: Generally, except when diseased, our brain cells last much longer than almost all other cells in our bodies. I expect that in the next century we should be able to double our life spans, and this will help us solve many more problems. But it also will raise many other problems -- and we'll need new,
highly intelligent machines to solve these.
TechTV: Which problems do you have in mind?
Minsky: Well, I would include all the usual ones: epidemics, education, pollution, energy sources, wealth distribution, etc.
But the most serious problems will be aging and population, and that's where intelligent robots will come in.
TechTV: Do you predict we will soon run out of "real workers"?
Minsky: Exactly. There will be one young 100-year-old for each aging 200-year-old. So we'll need intelligent robots to do most labor that was formerly done by younger people. But we'll also use them to solve other problems that our old-fashioned brains have not overcome.
TechTV: Is that because of how slowly we think in comparison to a computer?
Minsky: I would say that's an open question! I suspect that even a present-day computer could think as fast and as well as a person could -- except that we haven't been programming them right.
