Philosophy of AI

Names

Philosophy of AI is a history of "big names". The debates are great fun to watch. Here are some big names and my take on them. You don't have to agree with me of course (The great thing about philosophy is it's not falsifiable!):

Turing
- His paper Computing Machinery And Intelligence - The Turing Test. Stop debating what intelligence means. If you can't tell the difference, it is intelligent.
- I say: Has a point, but The Turing Test is easy to pass. And the question of trickery and fakery still remains, and is a legitimate question. And it's all very linguistic-oriented. See Discussion.
Dreyfus
- His books, What Computers Can't Do and What Computers Still Can't Do - AI machines only look intelligent because they are programmed to output their meaningless tokens as English words. They have no idea what they are saying. He seems at times to say AI is impossible.
- I say: Fair criticism of much of AI. Doesn't apply to the new stuff. Dreyfus leads us to the Symbol-grounding problem.
Searle
- His thought experiment The Chinese Room (also here) - AI is impossible. Instantiate your algorithm as a roomful of 1 billion people passing meaningless tokens around. You're telling me you can have a conversation in English with China, yet not one of the Chinese understands English?
- I say: Yes.
Godel, Lucas, Penrose
- AI is impossible because of Godel's theorem (or here) on the limits of logical systems. There are propositions that we can see are true that the AI logical system can't.
- I say: Any working AI would not be a logical (truth-preserving) system. It would be stochastic, statistical. See My comments on Penrose.
Penrose again
- His books The Emperor's New Mind and Shadows of the Mind - The brain works by exploiting quantum physical effects. Machines that exploit quantum physical effects are what you need to solve AI.
- I say: These look interesting but how do they solve AI? See My comments on Penrose.
Edelman
- His books Neural Darwinism, The Remembered Present and Bright Air, Brilliant Fire - AI is the wrong way to do it. It should be done this way.
- I say: Edelman is doing AI. (And not very well.) See My comments on Edelman.
Rosen (see here)
- His book Life Itself - The machine metaphor is incorrect. Here are types of self-referential system that cannot be implemented as a machine.
- I say: They can be implemented as a machine.
- See the Church-Turing thesis.
Brooks
- Many papers. e.g. Intelligence without Reason - Traditional AI is the wrong way to do it. We should do this New type of AI.
- I say: I pretty much agree. Brooks' work is not the final answer of course, but his analysis is excellent. Other people disagree. See Today the earwig, tomorrow man?.
Lots of people (*)
- The soul or spirit exists. There is a spiritual world. You are more than just the matter of your brain and your body. You are not just a biological machine operating according to the laws of physics. You have free will.
- I say:
  1. Well what does the brain do then? Why is it the most complex object in the universe?
  2. Also, this spirit clearly must have causal effects on your brain/body. So at what point does it interact? In the brain, should we expect to see uncaused causes? Neurons firing for clearly no physical cause? The idea is outlandish.
  3. Also, how did this "spirit" evolve from creatures that were just matter? (Most of the world still believes in a "first human", over a hundred years after science abandoned the concept.) Evolutionary theory as good as proves you wrong (as Darwin saw, but Wallace refused to see).
  (*) Who are "Lots of people"?
  Mainly, lots of civilians. - Unable to explain how the brain works, just about every human throughout history has believed in souls or spirits. Even today it is still part of the official doctrine of most churches, and is claimed by almost every religious thinker and theologian. Perhaps 95 percent of the world's civilian population accepts this view.
  Strangely, almost no one who actually studies the mind ever argues this view. With a complete disregard for public opinion, most cognitive scientists are materialists (perhaps Eccles (see here) is the only exception). Yet if spirits were really needed to explain the mind, then surely it should not be too hard to construct a scientific argument that the brain is not enough. It seems to me that the soul or spirit is merely a remaining example of the "God of the gaps" argument (also here) that theists should have left behind in the Middle Ages.
  Q. Could we ever prove this? Could AI and Cognitive Science prove that we are material?
  A. Very hard to prove it. If we invent AIs, they clearly have no soul. But that doesn't mean we don't.
Moravec
- His book Mind Children - AI is coming, and humans will go extinct. And that won't be necessarily bad. AIs will be our inheritors.
- I say: I like a lot of Moravec, but I have doubts about this. First, Who's to say we won't become AIs ourselves? Second, who's going to "mop up" the humans who don't co-operate with this "evolutionary inevitability"? Evolution is not in charge now. We are. And the only way humans will go extinct is by genocide.
Warwick
- Lots of sensational media statements - AI is coming, and it's dangerous. And it's going to happen soon (10/50 years).
- I say: AI is harder than you think. Your simple evolutionary robots hardly qualify as "breaking the back" of the AI problem.
Weizenbaum
- His book Computer Power and Human Reason - AI is dangerous.
- I say: We've hundreds of years to adjust. AI is hard, and even if machines do get intentions, life will be much, much harder for them than us. What we are really doing is not creating another species but understanding how we ourselves work, so we can mess around with it. Long term, this is about immortality, not extinction.
My favourite debaters with the AI critics:

Philosophy of AI

Names

Recommended Reading