Slashdot Mirror
AI in Sci-Fi
An anonymous submitter writes: "Stumbled upon a pretty interesting article considering the idea, 'What would machines do if they did achieve sentience?' It's by a sci-fi author I haven't heard of but worked with Kubrick on AI, he takes the whole AI or sentient machine idea a little further than we normally see in film."
write A.I. and not Al as in Al-qaeda or Al Capone!
Ian Watson's Novels
An Interview with Ian Watson
Ian Watson's Bibliography
Science Fiction Weekly Interview
How to Download YouTube Videos
The idea goes as follows: If a self-aware "real AI" ever existed, one capable of self-understanding and self-modification (called the seed AI), it would be in a much better position to create AI than its original creators. So would begin a chain of self-refinement and the creation of progressively smarter intelligences with decreasing time gaps between stages. Eventually a point is reached, called the singularity: nothing about the future past the singularity can be predicted by humans who live in the pre-singularity world. A common interpretation is that the chain of AIs would become more intelligent without bound, leading to a verticality.
The singulaity was first popularized by Vernor Vinge.
I've been doing a lot of reading on the singularity lately, and I've become more and more convinced that it is certain to happen.
More singularity links:
The singularity institute - A nonprofit working to hasten the singularity
Extensive writings by Eliezer Yudkowsky.
I've myself written a bit on singularity and AI related topics.
Yes, not everyone knows what an actuary does. An actuary is a statistician who computes insurance risks and premiums (usually they advise management on other issues too - for instance how an increasing life expectancy will affect how much the company pays out in pensions). It wouldn't be very difficult to write a computer program to answer an actuarial exam correctly as maths is the one thing computers are very good at. However you would end up with the computer getting 100% in a nanosecond - then twiddling its thumbs for the next two hours - waiting for the humans to catch up with it. ;o)
Video Game cheats, hints a
unless you're a Buddha seeking to negate the self
It's a common misconception that Buddhism is just about "negating the self". In fact, the purpose of it is precisely to be able to do what you want better. A buddhist also has a self and has desires, needs, etc, just like any other human being. The difference is just that he's aware that those are desires and needs and he has more control over them. He also has the discipline to listen to his intuition to decide whether a particular desire is worth pursuing or not. But he's not some empty zombie that doesn't desire anything.
Daniel
Carpe Diem
Several authors/books related to this subject that might be of interest are:
1. Stanislaw Lem's "Golem XIV" (it appeared as part of the "Imaginary Magnitude" collection (which also contains other stories about machine intelligence, for instance about machine literature), as well as apparently as a separate book). It is a story told as a series of lectures by a superintelligent computer (the Golem of the title). While some of it is pretty hokey (and some of it pretty funny), it contains some interesting speculations as to what superintelligence could consist of and how the physical and evolutionary contraints on human intelligence may make machine intelligence (which would presumably not be similarly encumbered) very different.
2. Daniel Keyes' "Flowers for Algernon". It is a story of a mentally retarded man who is given surgery that not only corrects his retardation, but makes him superintelligent. The story is told from a first-person perspective, so the level of the narration reflects his changing intellect. It has been 10+ years since I read it -- I would be interested in seeing how his superintelligent-phase writing held up.
3. Stephen Wolfram's "A New Kind of Science". Last year's geek-must-read book about how the entire universe is a cellular automatum (of course, I am compressing). It speculates -- and I am sure that I am getting this wrong (experts, please correct me) -- that the level of complexity of relatively simple CA rule sets is the maximum possible level of complexity, which would seem to have implications for limits on superintelligence.
A few additional thoughts:
4. One of the themes that seems to come up in SciFi treatments of AI is that a AI would have amazing predictive powers. I would think, however, that principles from chaos theory, the uncertainty principle, etc. would place real limits on that area of intelligence for most real world purposes.
5. I would be interested in hearing how cognitive psychologists and computer scientists even define intelligence, particularly at the high end of the (human) scale.
No. Common sci-fi misunderstanding.
Sentience is the ability to sense. Some plants are sentient. Sapience is the ability to reason. Most mammals have limited sapience.
Self-awareness is a specialized skill in the scale of sapience.
Defining self-awareness is a circular and fuzzy propostion. My CPU knows how warm it is and can change its operating speed to protect itself, but does it really know? Converseley, many humans don't have any understanding of how they behaving.
This makes it good for skiffy writers. They don't have to worry that someone will call them on their central conceit. It's ineffable.
According to dictionary.com, you are partially right. The first definition is actually The quality or state of being sentient; consciousness, which supports my definition, but the second is Feeling as distinguished from perception or thought (which supports your definition).
But being partially right makes you wrong on the idea that the first definition of sentience is a "sci-fi misunderstanding". It's the primary dictionary definition of "sentience", so it's certainly not a misunderstanding.
Daniel
Carpe Diem