Although there are two other interesting pieces I’d like to talk about, there’s one or two more pieces of Sherry Turkle’s work that I’d like to talk about.

Turkle describes a bizarre interaction she had with a Scientific American writer. Apparently they had a quarrel in which the reporter likened her discrimination of human and robot to discrimination between homosexuals and heterosexuals. I find that shocking because, first, it’s a completely irrational comparison, and second, Scientific American has always been to me an excellent representative of clear thinking. She also describes a book by prize-winning computer scientist David Levy who seriously argues that by the middle of this century, it will be commonplace for humans to have earnest romantic relationships with robots. That’s stupid and I agree with Turkle’s criticisms.

But there’s something I disagree with. She implies that it’s impossible for an artificial intelligence to ever be sentient. I used to hold the same view, but I’ve come to change my mind about that. I think that although we’ll probably never see a sentient artificial intelligence in our lifetimes, such a thing is not theoretically impossible. The question opens up a host of problems in philosophy and theology.

This is how I’ve changed my mind.

I used to think that machine sentience was impossible because no matter how we programmed an AI, it would only be executing a set of human commands. Even the most convincing set of responses would be only a set of what Turkle calls “as-if” actions: For example, responding to an insult “as if” the machine really is insulted, when in fact there’s “nobody home.”

But consider this line of thought: If a non-sentient animal can eventually gain sentience, why can’t a non-sentient machine do the same?

It seems clear that species change over time, just as languages do. Nothing is static, and given a sufficiently long stretch of time, amazing changes will occur. American Sign Language originated from French Sign Language in the 1800s and “speciated” into its own language over the generations. A similar thing happened with all the languages emerging from older languages, like the Romantic languages stemming from Latin. All the world’s ~6,000 languages share a “family tree” analogous to the family tree of biological species. Dogs originated from older species of wolves. The new species of bacteria evolved in a laboratory at Michigan State University originated from the original E. Coli species that biologist Richard Lenski started with in 1988. Homo Sapiens originated from older primate species. A series of innumerable small steps over the amazingly huge history of life (3,500,000,000 years) has produced amazingly huge changes.

Given that, it seems clear that we evolved. We’re part of the tree of life. This means that our great-great-great-etc. grandparents were substantially different. Take it back far enough, and one ancestor was a cave-man. Take it back further, and their ancestors were proto-humans. Further back still, their ancestors didn’t even resemble humans. We are sentient beings whose great-great-great-etc. grandparents were not sentient.

This has profound implications for the way we think about ourselves. I had a friend in high school who repeatedly insisted, “I didn’t come from no monkey” when evolution came up. We’d hate to think of ourselves as a monkey’s relative, but the conclusion is inescapable.

And so, if our lineage gained sentience from a state of non-sentience, why couldn’t another?

A new sentient intelligence could be created by Man in the following ways:

  1. A breeding program selecting for intelligence.  In Siberia, a team of biologists has for decades been conducting a breeding experiment with silver foxes.  One group is selected for aggression.  Another group is selected for docility.  In just a few generations, substantial changes have resulted in both groups.  It would be possible to do a similar experiment selecting for intelligence with chimpanzees.  Chimps are already on the cusp of sentience, recognizing themselves in a mirror and communicating with a limited sign language.
  2. Direct cellular or genetic manipulation of an existing organism.  Researchers at the University of Rochester in New York recently announced their findings that implanting human glial brain cells into mouse brains resulted in significantly increased intelligence.  Maybe we’re facing a Planet Of The Mice scenario.
  3. The “classic” approach to artificial intelligence, using a digital format to program and develop an intelligence.  This approach has so far not yielded human-like intelligence, but it may be only a matter of time.  It took the Earth 3.5 billion years to produce sentience by accident.  It would take us a significantly shorter time.

Fortunately, none of these are likely to happen any time soon.