AI in AI





  Python Book



AI in "AI"

The recently deceased film director Stanley Kubrick is the father of the best-known example of artificial intelligence in the movies: the HAL-9000 computer, which controls a spaceship in the movie "2001: A Space Odyssey" (1968). As an artificially intelligent computer HAL is convincing: it can converse intelligently and has a sense of humour, but it shows virtually no emotions and the question whether or not it's self-conscious remains open. HAL's power over the spaceship, combined with its lack of feelings and the goals it has been assigned, lead to a crucial scene in the movie in which HAL makes a decision, which is correct for the mission, but fatal for the astronauts. Here Kubrick raises moral and ethical questions, which are currently not an issue, but which the research into artificial intelligence in the future won't be able to ignore.

After finishing "Full Metal Jacket" (1987), Kubrick worked for a long time on a movie of which only was known that the title would be "AI", and that the subject would be robots in a flooded New York. In the end Kubrick didn't make this movie, supposedly because he saw no way to make it work on the big screen. Instead he spent his time on "Eyes Wide Shut" (1999), which was released six weeks after his death. His work on "AI" was inherited by Steven Spielberg, who made the movie and released it in the United States in the summer of 2001.

While Kubrick and Spielberg are both extremely skilled visual artists, their styles of storytelling are antagonistic. Kubrick's movies are known for the minor role that "humanity" and "feelings" play in them. His characters are either emotionless beings or extravagant caricatures. For example, in "2001" HAL probably has the most engaging personality. Kubrick's movies are concerned with harsh realities, and since there are no happy endings in the real world, they end without exception tragically or indifferently. Contrariwise, in Spielberg's movies emotions play a major role. He is a master of touching people. His movies are filled with sentimentality, love, valour, sacrifice and childish wonder. His mastery of the art shows in the fact that his movies never degrade to being tearjerkers.

The movie "AI" plays in the near future. The polar icecaps have melted and natural resources have become scarce. The Earth can only support a relatively small number of people, and therefore human-like robots have become an economically viable resource. They are used as worker, as baby-sit, as play-pal or as sexual partner. Robots differ from people in that their intelligence is of a human level, but devoid of emotions. At the start of the movie, a university professor in robotics decides to build a robot that can love. Less than two years later he has succeeded in creating a robot who looks and acts like an eleven-year-old boy. A grown-up can make a so-called "imprint" on its software, which makes the artificial boy love this person forever.

When this robot, named David, is first introduced in the movie, he visibly distinguishes himself from human beings in his behaviour. He walks and moves rigidly, he imitates a lot, and his laugh is obviously artificial. This is all plausible from the point of view of artificial intelligence, namely as examples of flawed replication and learning by imitation. This changes when David's "mother" makes the imprint (for totally unclear reasons): David magically transforms into a real boy, who not only loves his mother dearly, but who also behaves naturally.

There are a few things I wish to say about this process. Firstly, two years seems an impossibly short period to get from an emotionless intelligence to a loving one. On the other hand, when reviewing the simplicity of the whole imprinting process (the mother follows a simple procedure which, as it were, "burns" her identity in David's software), one must inevitably conclude that David's love for his mother is no more than the execution of a series of preprogrammed actions which only need an object. David does not learn to love; he simply starts doing it at one particular moment.

But how, then, is it possible that David at that same moment not only activates his "love program" (which consists of calling the object of his love "mommy", embracing her, and panicking when he isn't near her), but also starts to act naturally? Ah, this is typical Spielberg manipulation. For Spielberg the ability to love is the one thing that distinguishes human beings from perfect machines. When a machine has received the ability to love, it becomes the equal of a human being, and so, when his love is activated, David becomes a real elevenyear- old boy and acts accordingly. David's unnatural behaviour before his metamorphosis has obviously been added to stress that point. Spielberg tries to subconsciously convince us that David's love is as real as human love (though he seems to think it's also necessary to make this point explicit by stating it as a motto on the movie poster).

There is no doubt in my mind (based on twenty years of experience with Kubrick's films) that Kubrick would have treated this part of the movie differently. He would have made it clear that emotionless intelligence plus simplistically coded emotions wouldn't amount to more than that, and that it's questionable whether it is at all possible to create artificial emotions. The deep disappointment this would mean for the mother could be the reason why David gets separated from her (for which Spielberg has a less plausible reason). The remainder of the movie is concerned with David's search for his mother, which for Spielberg is a tale of the inseparable bond between a child and his mother, and which for Kubrick would have been the execution of a useless piece of programming by an advanced Tamagotchi. The story would remain virtually the same, but the meaning would be different.

While professionally made, as may be expected from Spielberg, the ending of "AI" as shown in the movie theatres is too infantile for words. If it says anything about artificial intelligence, it is that David's love for his mother is even deeper than human love. This leaves the average viewer with the impression that one day, evil programmers will be able to write a couple of lines of code to create digital emotions, which are so intense that they will exist for eternity. I don't think there is one AI specialist in this world who believes that even for an instant. If artificial intelligence will at one time have emotions, these will slowly emerge while the intelligence is developing. It's ridiculous to suppose that you can simply program them. Kubrick would presumably have given a more realistic impression of coded emotions, but his vision has regrettably not made it to the white screen.

© 2001 by Pieter Spronck

Published in BNVKI/AIABN Newsletter, August 2001, Vol. 18, No. 4, pp. 89-91.