Slashdot Mirror
Review - Bicentennial Man
In the last few years, Robin Williams has gone from being one of the funniest movie stars in Hollywood to one of the sappiest, so there is reason to be suspicious of "Bicentennial Man," which was previewed in some theaters across the country this weekend.
The movie is plenty syrupy, but also surprisingly faithful to the Isaac Asimov story that it's based on, and to many of the issues it raised. Like the evolution of artificial-intelligence (AI) machines, questions about whether they can possibly have emotional lives, and their relationship with human beings.
Williams plays a household-appliance robot named Andrew who develops human-like characteristics - friendship, loyalty, humor, creativity, and faces some tough questions as a result? Is he a human or a robot? Are his feelings real, or simply the pre-programmed responses of neural pathways? Is he an appliance or some new form of life? Does he have rights and can he in any way be called human?
Probably the central question, and one Asimov often raised in his wirtings, was how exactly, creations like "The Bicentennial Man" are supposed to live in a culture with enough gee-whiz technology to create them, but that typically hasn't given a though as to how they'll get along in the world.
Andrew is programmed to live forever, at least theoretically, and this puts him in conflict with the life he wants to lead, especially the fact that he feels much more human than robotic, and that everyone he loves grows old, then dies.
Most humans see him as a machine, but spurred by a sympathetic and ethical owner and his daughter, Andrew sets out to hone his evolving skills and - one can see this coming from the first scene - ends up having a lot of heart and wanting a real one (in one hilarious and self-knowing scene, a fellow robot taunts him by singing the "Tin Man" song from the "Wizard of Oz)." Andrew decides that he needs to be free to figure all of this out, and he sets off on a quest to find his place in the world. It's at this point - three-quarters of the way through, that the plot begins to unravel, and stops being even remotely plausible.
The movie is at times gorgeously-shot, and makes innovative use of computer graphics to render cities, hospitals and offices of the future. It also deals sensitively and intelligently with a lot of the issues many suspect are coming, if even only a fraction of the predictions about AI machines come to pass.
Williams can't help but lapsing into the most wide-eyed, saccharine dialogue and character-development.
But this doesn't keep the movie from being surprisingly thoughtful and touching. And prescient, raising issues about technology and the future that hardly anyone in the United States really wants to talk about.
I adhere in some ways to the Behaviorist notion that what matters about intelligence is a) what goes into the machine and b) what comes out. There is nothing else. If you feel that there is more going on inside you than what can be summarized by your external stimuli and your external reactions, then you are mistaken. You are only observing an internalized output to external stimuli. The feedback you would normally express in the outside world is instead being piped directly to your brain's input valve.
There are serious problems with behaviourism.
First of all, if behaviourism were true, we could teach pigs to sing. We can't. There are built-in functions in the brain that make a difference in intelligence.
Second, human children learn language in a fashion which behaviourism can't account for. Children will learn whatever language they are exposed to. They learn the rules of their language without ever understanding them as rules. They do not make the type of mistakes a trial-and-error behaviour renforcement model would require of them. They always group words into structures, even in highly inflected languages and even when they get the fine points of syntax wrong. Furthermore (and most damningly) a human child can become fully functional in a foreign language in under a year, while few adults can do so under any circumstances. The most parsimonious theory that includes these facts is that humans, like other animals, have biological mechanisms in the brain that enable them to do these things.
Thirdly, there are growing bodies of evidence that large areas of human behaviour are biologically influenced. Several forms of psychiatric disease can be clearly traced to biochemical roots. Human sexual behaviour has clear biological roots (I doubt anyone would much bother with sex if their brains didn't force them too.) Even areas like anti-social behaviour are increasingly believed to have partially biological origins, possibly hereditary ones.
That means that humans are not tabula rasa as the behaviourists believed. What goes on in our brains is not a simple function of external stimula.
Now, that does not mean it isn't possible to understand these parts of the brain and program computers to emulate them effectively, but if we do so, we are emulating a human, not creating a truly new machine intelligence.
I can easily imagine a machine pretending to be human wanting to become fully human. Such a machine would likely have emotional states, since we are unlikely to be able to separate these genuinely human conditions from an abstract intelligence. We don't even have a good definition of intelligence, and even if we fully understood the biology and functioning of the brain, we are unlikely to be able to discuss intelligence apart from it's structural framework.
An AI that thinks like a human is likely to want the range of experience and the level of autonomy that humans enjoy. It's not implausible that it would want to be seen and treated as an equal to humans. It is conceivable that it would view itself as superior, but I find it hard to believe that any probable AI would wish to reject the ensemble of human experience in the way you suggest.