From: Jed Rothwell Whether these computers will be sentient or not is an entirely different question. Whether they should be deliberately designed to be sentient is both a practical question, and a moral one. I think the problem is not whether computers "should be designed to be sentient," so much as "can they be restrained from it."
IOW the decision is probably NOT going to be ours (in the USA) to make, given that there will always be a group, or class of humans somewhere on the planet who can benefit in the short term from better robots. The result is that - in the not-too-distant future - some group of humans who may be in a position to benefit financially - will develop an efficient way for individual AI systems to both "learn from their own real world experiences", independently of all prior programming - and then to modify their own internal programming. That is essentially the "HAL syndrome" taken to its logical conclusion. In fact, in free-market Capitalism (unless strong International legislation changes things for the entire planet) this eventuality (of virtual independence for HAL somewhere on earth) - is almost guaranteed to happen, since it will be supremely cost effective, despite whatever long-term risk it involves. As a race, humans are simply not logical enough to protect against this happening. For instance - look at the fear of a NEW WORLD ORDER, and all that it portends in this particular argument. In short, there can be no worldwide legislation to prevent this, so it probably will happen. This step of self-programming will allow "them" to evolve on their own, and the time frame could be shorter than expected - without morals, without "empathy" ... which is essentially what Bill Joy was implying: that average humans will be superfluous, even if "they" decide to keep the exceptional humans around for whatever turns out to be unique in biological intelligence.
<<attachment: winmail.dat>>