Tuesday, February 03, 2009

Robot Love

I read an interesting passage today in Ger Gigerenzer's Gut Feelings: The Intelligence of the Unconscious. The focus of his book is the idea that effective intuitive processes are the product of discarding information -- not retaining it. He illustrates his point with a short story, The Fable of Robot Love. The point is interesting, so bear with it:
In the year 2525, engineers finally managed to build robots that looked like humans, acted like humans, and were ready to reproduce. Ten thousand robots of various types had been built, all of them female. A research team set out to design a male robot who would be able to find a good mate, found a family, and take care of the little robots until they were able to take care of themselves. They called their first model Maximizer-1, M-1 for short.

Programmed to find the best mate, M-1 proceeded to identify a thousand female robots that fit his goal of not marrying a model older than himself. He detected five hundred features on which individual female robots varied, such as energy consumption, computing speed, and frame elasticity. Regrettably, the females did not have their individual feature values printed on their foreheads; some even hid them, trying to fool M-1. He had to infer these values from samples of behavior.

After three months had passed, he had succeeded in getting reliable measures on the first feature he tested, memory size, from each female robot. The research team made a quick calculation of when M-1 would be ready to pick the best and discovered that no one in the team would still be alive at that point -- nor would the best female robot.

The thousand females were upset that M-1 could not make up his mind, and as he began recording the second feature, the serial number, they pulled out his batteries and dumped him in a scrap yard. The team went back to the drawing board.

M-2 was designed to focus on the most important features and to stop looking for more when the costs of collecting further information exceeded its benefits. After three months, M-2 was exactly where M-1 was, and in addition was busy measuring the benefits and costs of each feature so that he could know what to ignore. The impatient females ripped out his wires and got rid of him, too.

The team now adopted the proverb that the best is the enemy of the good, and designed G-1, a robot who looked for a mate that was good enough. G-1 had an aspiration level build in. When he encountered the first female who met his aspiration level, he would propose to her -- and ignore the rest. To make sure that he found a mate if his aspirations were too high, he was equipped with a feedback loop that lowered the aspiration level if none of the females were good enough for him over too long a period. G-1 showed no interest in the first six females he met, but then proposed to number seven. Short of alternatives, she accepted.

Three months later, to everyone's pleasure, G-1 was married and had two small kids. While writing the final report, however the team learned that G-1 had left his wife for another robot. Nothing in his brain had prevented him from running off to what looked to him like a better deal. One team member pointed out that M-1 would never have left his wife, because he would only have accepted the best in the first place. That's true, responded the others, but at least G-1 found one.

The team discussed this for a while and then came up with GE-1. He was happy with a good enough female, just like G-1, but was additionally equipped with an emotional glue that was released when he met a good enough robot and adhered more strongly with physical contact. Just to be sure, they inserted a second form of emotional glue into his brain that discharged when a baby was born and tightened after each physical contact with the baby.

GE-1 proposed to a female as quickly as G-1 did, married, and fathered three babies. He was still with them when the team finished their report. He was somewhat clingy, but dependable. Ever since, GE-1 robots have conquered the earth.
In the fable, M-1 failed because he tried to find the best, as did M-2. Both ran out of time. G-1 was fast by going for a good enough choice, but was also fast in dropping it. However, the capacity for love, the glue, provided a powerful stopping rule that ended GE-1's search for a partner and strengthened his commitment to his loved ones.

Similarly, feelings of parental love, triggered by an infant's presence or smile, free parents from having to decide every morning whether they should invest their resources in their children or in some other business. The questionof whether it is worthwhile to endure all the sleepless nights and other frustrations of having a child simply does not arise, and our memory ensures that we forget these tribulations soon.

The evolved brain keeps us from looking too long and thinking too much. That is what love is. ;-) The culture it is embedded in influence what the objects of love or trust can be, or what makes us upset or feel hurt.

Gigerenzer goes on to discuss this problem in humans -- the deliberate search for the best can conflict with emotions (pride, honor) in humans. Johannes Kepler, for instance, was short, sickly, and born into a poor family. However, his fame and intellect ranked him a good catch. In 1611, he left his unhappy and arranged marriage, and embarked on a systematic search for a second wife. Unlike Barbara Bush, who "married the first man she kissed," Kepler investigated eleven (!) possible replacement wives in under two years. His friends urged him to pick bachelorette number four, a refined lady of high status and dowry, but he insisted on being thorough. This insulted the fine lady, and she rejected him in the end.


Post a Comment

Subscribe to Post Comments [Atom]

<< Home