Late last year, Robert J. Marks and Michael Medved continued an earlier discussion on computers. The earlier discussion had focused on the way that computers magnify what we do, as opposed to having independent brand new abilities. This time, they focused on natural intelligence (both human and animal) vs artificial intelligence.
“Great Minds: Medved and Marks on Artificial Intelligence (December 13, 2018)
Michael Medved: How is animal intelligence different from machine intelligence?
Robert J. Marks: One of the areas I did research in was swarm intelligence. One of the things that engineers do is, they go into the field and they learn about the way nature works and they extract algorithms of ways of doing things from nature. And, for example, you never think of insects, social insects, as being very smart. But incredibly, even though individually they are stupid bugs, they can do incredible things because of their emergent behavior.
Marks referenced the Monarch butterfly migration as an example of the swarm following a plan not devised by any individual member.
Michael Medved: Are there similar mysteries to the way computers work?
Robert J. Marks: There are mysteries to the way computers work. One of the examples is that you cannot look at a computer program and have another computer program analyse what it will do. In fact, there is no way you can know what a computer program will do until you run it. And this is a remarkable result that was pioneered by Alan Turing in his so-called Turing halting problem…
Michael Medved: Do we have any evidence of computers wondering about things?
Robert J. Marks: No. They might say “I need more information, something like that, but… as far as any sort of query that is based on creativity, I would have to say no.”
Medved commented that he had heard that a computer at NASA had shown creativity, using a process called “evolutionary computing” to design an antenna that they used.
Robert J. Marks: The idea was that they applied the theory of evolution to the design of this antenna, they came up with a design that looked like a bent paper clip but it ran very well and was actually deployed to outer space.
However, the reality is that, first of all, there is an end result that these engineers wanted. And it was called the specification for the fitness of the design, if you will. And they placed the fitness and they used the powerful electromagnetic software which placed the fields to make sure that the antennae received it correctly. So the basic idea — and this happens alot in artificial intelligence in terms of search, is that the program will go in and they will put out a field of billions and trillions of different solutions and they’re not sure what the consequences of those solutions are going to be and the question is, how do you wade through these solutions in a smart way? Because you don’t have the time to do them all. Sometimes those solutions are surprising. But they are not creative because it has all been placed out there by the computer program.
Michael Medved: What can human beings do that even the most- well-designed and sophisticated computers, artificial intelligence systems, can’t do.
Robert J. Marks: I would say that the thing that really stands out is creativity, the ability to actually do something that is against the accumulated knowledge that is, say, in computer memory, to do something external to that. In fact, in order to do something creative, you have to abandon that and do something which is brand new. I think that creativity is something that computers will never do. They get surprising results, unexpected results, but you’ll never get creativity.
Previous podcast: Robert J. Marks Talks Computers with Michael Medved Computers can magnify what we do, he says, and that’s the real threat
Further reading: Things exist that are unknowable: A tutorial on Chaitin’s number (Robert Marks)
Human intelligence as a halting oracle (Eric Holloway)