• Question: Do you think that robots could ever be more clever then humans

    Asked by 09gurdcar to Alan on 19 Mar 2011 in Categories: .
    • Photo: Alan Winfield

      Alan Winfield answered on 19 Mar 2011:

      Yes I do, but not for a very long time – much much longer that some people predict. The problem you see, is that we don’t understand how the human brain works. We understand some things about it, of course, but not about how intelligence really works. We don’t know for instance how it is that you know that you’re you. We don’t know how human emotions work, and the relationship between emotional intelligence and rational (logical) intelligence. We don’t understand how you make decisions, or learn to ride a bike, or how you can recall facts or what the brain mechanisms are that make you laugh at jokes. One very fundamental thing we don’t understand are the brain mechanisms that allow human babies to learn language so easily and quickly.

      Some experts in Artificial Intelligence think that human level AI is only a few decades away from being developed. I think they are completely wrong – in my view it will be hundreds of years into the future (too far to be able to predict). One of the reasons they give is that computers are getting faster and faster every year. It’s true that computers are getting very fast – but just having fast computers isn’t enough to solve the problem of AI. Computer power is like raw material, but having lots of raw material isn’t enough. Imagine you had a huge pile of Italian marble – does that mean you can make a Cathedral? No, it doesn’t – you need the design, as well as the raw materials. And the fact is, we don’t have the design for human-level AI. Or even AI equivalent to much simpler animals.

      That doesn’t mean we can’t work out the design eventually. It’s just a very very hard problem that will take a very long time.

      Sorry if I’ve rambled on but this is a question I get asked alot (and I’m sorry if my answer might be a disappointment to you).