介绍:
CG: It seems that there’s nothing that can really stop AI from being applied to a wider range of fields, but do you sense any limitations of what AI can do at this moment?
TPZ: Yeah, so the successful Alpha Go indeed states that AI can do certain things, for complete information games. “Complete information games” means that people can see what’s going on here, so it doesn’t really apply to incomplete information games, such as poker. So if you play Texas Hold’em poker, you have two cards hidden, right? So the robot and both of the players cannot see what’s entirely going on in the game, so for these kinds of games, AI has its limitations.
ZCG: Professor Zeleznikow, where will AI be in the next five or ten years’ time? If human-like artificial intelligence is a realistic possibility, how long will it take to emerge? What about “super-intelligent” AI?
JZ: Well that’s interesting, first of all I don’t think human-like artificial intelligence is a realistic possibility, nor do I think that’s where we’re wanting to go. Emotions, intuition: you can try and model those via computers using artificial intelligence, but I can’t see the sense in doing that. Where we, for example, are superb at using AI is in planning. If you’re driving a car and you go through a red light or you exceed the speed limit, then there are automated cameras or automated machines that in fact take a picture, record how fast you’re going, find out who owns the car and then produce a fine to the owner of the car. Now, all those sorts of planning are really quite simple and well-known, so we’re coming along, and in five years’ time we’ll be further advanced and we’ll have better robots. I’m not too sure that one can use the word ‘super-intelligent’, because our idea of what AI does and where it’s going really depends on the time. In ten years’ time we will have more services at home, we will have more automated services in public transport, on the roads, at airports and things like that, but it still won’t be super-intelligence that can read people’s emotions or minds. It will progress, but it will be evolutionary rather than revolutionary.
ZCG: Critics of AI do argue that if the technology becomes too sophisticated, it might one day challenge human pre-eminence, in a scenario some call an ‘intelligence explosion’. How likely is it that such a sequence of events may one day occur?
JZ: Well, we are moving more and more towards computer systems – artificial intelligence – modeling human decision-making and human understanding, but there are certain cognitive features, there are certain emotional features that in fact are very, very different and difficult to model. That’s why I don’t think we’re ever going to quite get to that stage.
