I discussed this subject a little in the book, in Chapter 10. I only said a little because I only know a little.
A recent issue of Scientific American had a fascinating story about brains, comparing the number of neurons and operations to a contemporary computer. Everyone knows the brain is an MPP processor, but the size of it is mind-boggling. As I recall it worked out to be something like 10^16 operations per second in the brain versus 10^8 for the fastest supercomputer. No wonder they are so far behind us. Perhaps we cannot expect any sign of real intelligence until they reach something like 6 orders of magnitude below us, plus of course it will require presently unimaginable improvements in software. I think it can be done, but I think it will take a long time.
Joseph Weizenbaum wrote a famous book, "Computer Power and Human Reason -- From Judgment the Calculation," (1976) which I read in English and Japanese. It might be summarized as being a diatribe against artificial intelligence. I think it is an excellent book and it makes many important points, but fundamentally I disagree with two points:
1. First, Weizenbaum thinks that artificial intelligence can have no practical use and it might even be unethical. He picked an example which turned out to be very unfortunate from my point of view. He said that as of 1976 a great deal of effort has gone into voice-recognition, but there has been no progress, and even if they could make it work it would only take jobs away from telephone operators and other people and it would serve no larger social purpose. I depend upon voice input every day in order to write. Thousands of seriously disabled people depend upon it for their livelihood and their everyday happiness, safety and sanity. The jobs now being handled by voice input such as telephone directory lookup were demeaning and terribly boring. So he sure missed the boat on that!
2. Second, Weizenbaum thinks that artificial intelligence will never lead to anything remotely like a human mind, with the human ability to make moral judgments and so on. I certainly agree, and even if something like a mind emerges I hope we never let it make moral judgments about human behavior, even if is inclined to do so. It's bad enough having Georgia school principals apply rules blindly & mechanically, punishing a high school kid because he talks to his mother on a cell phone at lunchtime -- because she is a soldier on duty in Iraq, for crying out loud! Anyway, putting aside that diatribe, and getting back to Weizenbaum's diatribe, I think everyone agrees that computer intelligence will never be remotely similar to our own. It will be alien, and restricted in many ways. It will be -- we hope -- totally passionless and without emotional content, and without goals (other than those we set for it), ulterior motives, paranoia, or anything of that nature. Nothing like HAL. But it will still be intelligence. It will be self-aware, self-directed, and capable of rigorous abstract analysis and many other activities that we now assume are vital aspects of human intelligence.
Late-model artificial intelligence programs have already independently reinvented many important early 20th century inventions, including some patents from around 1910 that were critical to the telephone industry. These computers were 90 years late, but it is only a matter of time before they catch up and begin inventing new things that we have not thought of. I think it is also only a matter of time before they learn to read x-rays, diagnose illness, cook meals, invent new physics, beat us in games other than chess, and do many other things better than people do. The sooner the better. I have read a lot of blather about how important it is to bring the "human touch" to medicine, yet most of the serious medical mistakes I have read about have been caused by simple, mechanical problems -- such as mixing of two different drugs, or prescribing a dose an order of magnitude too large -- that could easily have been prevented with proper data processing techniques. Even today would be better to leave most of this work to machinery, even if by doing so we cause doctors' skills to atrophy. I would much rather have a computer diagnose my illness correctly than have a human doctor make a mistake. I could not care less about the importance of the "human touch." All I want is to be cured.
Along the same lines, the recent railroad accident in Japan would never have happened with the latest ATS control technology. The railroad company did not get around to installing the updated controllers on that segment of track. (It isn't their fault; they cannot replace their entire infrastructure overnight, and they never expected the driver to go 100 kph in this slow, local section of track.) The Atlanta subway cars are made by the same company, but they appear to be fully automatic -- as far as I can tell. The driver does not touch the controls once the train starts to move, or when it pulls into the station and stops. It would be better if all trains were fully automatic, even though I guess that makes the train driver's job boring and low skilled. If computers eventually take away all skilled work and make us as dependent as children and unable to tie our own shoes or make toast, it might be a problem. But I would still prefer that state of affairs, because I trust machines more than I trust people, and I want to reach the destination on the train safely more than I want to enhance the driver's workday experience.
In Japan there has been a lot of talk lately about how things will be when the baby boomers retire to nursing homes. They may have to be treated by soulless computerized robots instead of nurses. This allegedly upsets people. NHK television has shown things like automatic Jacuzzi baths for people in wheelchairs. The people taking these baths have been interviewed (sometimes when they are submerged), and they seem perfectly content to me. I personally would WAY prefer to be scrubbed down and sprayed off by a robot than a person. It is much less embarrassing, and less of an imposition on the person. I myself would not want to spend time taking care of some old coot. The young women who do this sort of job should be coddling babies instead.
- Jed
