Fiction and Facts
I had the interesting experience recently of attending a lecture given by Professor Kevin Warwick of the Department of Cybernetics at Reading. I had been fortunate, up to that point, of somehow remaining unaware of his brand of doom-mongering. Professor Warwick's thesis, in short, is that computers will take over the world sometime in the next century, probably in the next 50 years. And this, as the commentators in the numerous video clips of interviews with him which he showed us during the lecture, is not science fiction but a serious scientific theory.This lecture was delivered to the School of Artificial Intelligence at Edinburgh University and was not well recieved. The main objection boils down to the fact that Professor Warwick has no evidence (or nothing that constitutes scientific evidence) to back up his theories. He has run a number of "experiments" which, in the presence of academics he admitted were for the purpose of raising public awareness (or media gimmicks, as many of the audience termed them). For instance, he trained a robot (in Reading) using genetic algorithms to navigate around a course. He then downloaded the robot's memory into an identical robot (in New York) which (surprise! surprise!) can also navigate round the course. "Look", he says, "the computers are programminig each other - oh no, we're doomed". Nothing was learned here. No one, even vaguely computer literate, should be surprised that robots can "learn" (rather than be explicitly programmed) to perform simple tasks, nor that one robot's memory, when placed in another identicial robot should allow the second robot to act in the same way as the first. However, as a means of drawing attention to your work, this is great, especially with the added extra of one robot being in Reading and one in New York.
What interests me, is that if an author deduced from this evidence that computers would take over the world in 50 years time then it would be called science fiction. After all, it involves looking at a current science, abstracting a trend from it and deducing a possible future. A future that is possible, but not certain and, given our current ability to produce intelligent machines, unlikely (at least on the timescale presented) even if you ignore the various religious and philosophical arguments people have against intelligent machines. As if to reinforce this conclusion Professor Warwick, in the abstract of his talk, mentions that Gillian Anderson endorses his work. A fact which might (just) be of relevance if he wished to be treated as a science fiction writer but surely not if he wishes to be treated as a serious academic.
What is more this isn't even particularly original science fiction. Writers have been telling the "computers will out evolve humans and will decide to wipe them out" story for decades. Let's consider, just for a moment, where the money for research is likely to be in the future (for research follows money more often than not). If we ever get the technology to produce human-like intelligence someone will, almost certainly, build a replica human, just to show they can. The performance of this first android will, almost certainly, be considerably below that of a human. It's intelligence is (in terms of general functioning) likely to be less than human and we don't yet have the materials science knowledge to build something both as strong and as flexible as the human body. It's only possible economic use will be as slave labour and I like to think that given the debates that will undoubtedly occur about its "sentience" that people will be very diffident about assuming that it has no rights. I think a society with a large number of intelligent androids is very unlikely. More likely is a society with intelligent web based search engines, which have some "understanding" of which documents you are looking for. If a web based search engine were to become sentient it's possible (I'd lay odds of about evens) that it will not be particularly human-like in its outlook. To suggest that this entity would find itself in competition with humans for resources and therefore seek out the means to destroy them seems a little far-fetched, although the possibility of human-programmed, malicious search engines is an interesting thought. More dangerous, perhaps, would be an intelligent, military, tactical computer however I don't believe enough people would be stupid enough to connect such a computer directly up to weaponry and it's not clear that such a computer would need, and therefore be given the capability, to manipulate the people around it to fire off missiles on its behalf. What I'm trying to get at, is that the intelligent, sentient programs, if they ever occur, will probably be very tool based, focusing on specific tasks and, as such, whatever intelligence or sentience it is they develop is unlikely to be anthropomorphic and so predicting their actions is very hard. I also think that you can muse on the possibilities in a much more interesting way that the doom-laden scenario presented by Professor Warwick.
I have just written three (2 1/2) paragraphs explaining why I think, what Professor Warwick says is possible but not, as it currently stands, science. Unfortunately those are not three media friendly passages. The media likes people to have definite opinions whereas scientists like to admit the possibility of error. The media likes nice visual "experiments" while most scientific experiments are visually pretty dull. The media want sound-bites ("computers will take over the world") not rambling explanations about how its possible but unlikely. To effectively refute assertions such as Professor Warwick's (and AI is not the only science afflicted by such people almost every science has someone of this ilk) scientists have to say a short sentence ("This man is clearly mad") which leaves them open to the accusation of closed-mindedness. Long explanations about scientific method, what constitutes scientific evidence and why these are important do not easily make it into media debate.
Bad science is bad science fiction but it is more likely to get you on the News at Ten.
Louise Dennis Last modified: Fri Aug 10 14:54:21 BST 2001