快猫短视频

HOW intelligent are you? It鈥檚 a question psychologists often ask but now two
mathematicians argue that it鈥檚 meaningless. They say their experiments with
computer 鈥渕inibrains鈥� prove that intelligence depends on the environment and
cannot exist independently of it. The work has reignited a fierce debate on the
nature of intelligence.

Per Bak and Joe Wakeling from Imperial College in London pitted simple neural
network agents known as minibrains against each other in a game in which they
had to choose one of two groups to join. To win, they had to join the least
popular group.

A memory of previous rounds turned out to be an important factor in the game.
When the minibrains all remembered the results of an equal number of previous
rounds, they tended to make the same choice鈥攁nd so did extremely badly.
None of the agents managed even a 50 per cent success rate鈥攚hich would be
expected through chance alone. 鈥淚t was a big shock the first time we saw it,鈥�
says Wakeling.

But if an agent with a memory of three previous rounds was pitted against
lots of agents able to remember two, it achieved an astonishingly high success
rate of over 98 per cent. In some combinations, however, minibrains with
supposedly better memories did worse than simpler ones.

Bak and Wakeling say these findings show that it鈥檚 impossible to predict how
well a minibrain will do without knowing which individuals it will be competing
against, and argue that the same is true of any intelligent being. 鈥淭he mind and
its surrounding environment are inseparable.鈥�

Artificial-intelligence expert Dave Cliff of Hewlett-Packard Laboratories in
Bristol says he agrees with their conclusion but adds that this is not a new
idea. 鈥淭o judge intelligence in a meaningful way you have to do it relative to
some population.鈥�

Linda Gottfredson, a cognitive psychologist at the University of Delaware, is
equally convinced the argument has been won鈥攂ut by the other side. Bak and
Wakeling 鈥渉ave not become acquainted with the century of evidence that says
people carry around with them differences in their capabilities to behave
intelligently regardless of the circumstances,鈥� she says. She points out that
people who do well on one type of mental test tend to do well on them all.

Gottfredson argues that 鈥渋ntelligence鈥� is an inherent ability to 鈥渢ake in,
comprehend and process information鈥� in a way that promotes success no matter
what the task. That ability differs among individuals and can be measured, she
insists. But Bak and Wakeling go further and say that any ability to survive can
be considered intelligence. An immune system that survives would be more
intelligent than one that does not, for example.

Even superior information-processing capabilities aren鈥檛 necessarily better.
Cliff says one could argue that a thermostat is more intelligent than a PC
because it鈥檚 more efficient for the job it does.