The question of whether machines will be capable of human intelligence is ultimately a matter for philosophers to take up and not something that scientists can answer, an inventor and a computer scientist agreed during a debate night at the Massachusetts Institute of Technology.
Inventor Ray Kurtzweil and Yale University professor David Gelernter spent much of the session debating the definition of consciousness as they addressed the question, "Are we limited to building super-intelligent, robotic 'zombies,' or will it be possible for us to build conscious, creative, even 'spiritual' machines?" Although they disagreed, even sharply, on various points, they did agree that the question is philosophical rather than scientific.
The debate and a lecture that followed were part of MIT's celebration of the 70th anniversary of Alan Turing's paper "On Computable Numbers," which is widely held to be the theoretical foundation for the development of computers.
In a 1950 paper, Turing suggested a test to determine "machine intelligence." In the Turing Test, a human judge has a conversation with another human and a machine, not knowing which responses come from the human or the machine. If it cannot be determined where the responses come from -- the human or the machine -- then the machine is said to "pass" the test and exhibit intelligence. Of course, this being at least in part a philosophical matter, the Turing Test itself is the source of ongoing dispute.
Kurtzweil and Gelernter weren't so much interested in that dispute, but more focused on how his test could be applied. Kurtzweil's position was that machines will, in fact, some day pass the Turing Test, that modeling of parts of the brain is already leading to the ability to replicate certain human functions in a machine.
"We'll have systems that have the suppleness of the human brain," Kurzweil said, adding that to contemplate how those machines will be developed, it's important to accept that current software and computing power aren't up to the task and that technological advances are necessary first. So, it's important to look out 20 or so years.
Humans will recognize the intelligence of such machines because "the machines will be very clever and they'll get mad at us if we don't," he joked.
Gelernter smiled at that, but he also shook his head. He's not buying it because logically any machine that is programmed to mimic human feelings, which are an aspect of consciousness, is programmed to lie because a machine cannot feel what a human feels. That's the case even if the machine seems to be able to "think" like a human.
"It's clear that you don't just think with your brain," he said. "You think with your body."
While Kurtzweil noted that recently a computer simulated protein folding, which is something that was believed to be impossible for a machine to do, suggesting that it's difficult to predict what machines will be capable of doing. Gelernter had an answer for that, too -- that's all that happened, just the simulation of the folding, the process stopped there.
"You can simulate a rainstorm and nobody gets wet," he said to use another example.