人工智能的,不知道你要不要,不过这篇文章质量蛮高的。至于时间的话看你语速咯,鄙人大约用7分钟搞定……
computers still a long way from truly aping human mind
Artificial intelligence'will equal and surpass human mental abilities;If not in 20 years then surely in 50.'So claimed US computer scientist Nils Nilsson back in 1984.Some people may still agree with him.
However, many more must still feel that computers are not going to catch up with the human mind quite so quickly.It has long been tempting for artificial intelligence (AI)researchers to underestimate the gap between their latest lavoratory toys and the human mind.
The enthusiasm of the early 1980s was for expert systems.But despite ambitions to matching the reasoningpowers of human,expert systems ended up being little more than glorified spreadsheet packages which could crank outanswers from built-in rules and 'facts'.
Today,enthusiasm is for neural computing.The idea is to wire up a network of logic cells in a way that appears to match the connection of neurons in the human brain.A web of feedback relationships allows a neural computer to 'learn from experience'.Instead of having to write programs from scratch ,a human programmer can keep restarting the computer until it adjusts it internal logic.
But neural computers,like expert systems ,are still a long way from the visionary's dream of an intelligent and aware computer.The problem is that to mimic the workings of the human mind,computer scientists must make something of a marriage of these two AI approaches--and do so on a vastly great scale.
The human mind is formed of two parts : the animal brainthat equates to the bottom-up processing of a neural computer, and the top-down patterns of thought built out of language,that are captured by expert systems.The hardest part to replicateis the functioning of the basic animal brain.Humans have a brain about four times the size of a chimpanzee's or gorilla's,but it is really little different from an ape's ,as the neurons are wired up in the same way.
Awareness, too, is created in the same way.Millions of detector cells in the sense organs channel a torrent of information across the grey,wrinkled surface of the brain.
The difference between humans and other animals is in the richness of detail that can be painted on the larger processing area humans process.With 30,000 in a pinhead-size lump of human grey matter,the number of possible connections runs into billions.
No computer has yet got near lhis level of connectedness.A leading massively parallal computer, such as Conection Machine built by Thinking Machines, has a mere 65,000 processing elements.Nor are brain cells the simple on/off logic gates of a digital computer.The functioning of nerves depends more on their chemistry than electrical impulses,which are just an efficient way of getting messages from juction to junction.
the real computational work is done at the junctions,where dozens of neuro-transmitters are involved in the complex committee decisions.Compared with even a cetipede's brain, today's hardware is rigid in its responses and build, deaf and dumb to the outside world.But the good news is that if researchers can copy the brain's working,the next step--making this neural computer self-conscious and human-would not be so hard.