The anti-progressive sceptic

It is a sad and oft-times tediously annoying fact that the word "sceptic" has many differing connotations. I've mentioned this before, but it's something I continue to come back to mentally so I thought I'd subject you all to it as well. Phil Plait, the Bad Astronomer, recently defined "skeptic" simply as "someone who insists on seeing the evidence" - I'm probably paraphrasing but that was the jist of it.

And here's the etymology. The Greek word "skepsis" has an ambiguous meaning also - which has somehow either found its way through to the present day, or is simply coincidentally re-asserting itself now. It could mean "doubt" or "inquiry"; I would say that Phil's definition falls under the latter of those, and sadly the general definition in the public mentality falls under the former.

This is certainly not helped by those who term themselves sceptics in relation to a particular controversy, such as climate change sceptics, moon landing sceptics, or AI sceptics. Unsurprisingly for those who know me, it's the latter on whom I wish to focus today - though much of what I say can be generalised. Use your discretion in this regard.

I don't know if John Searle has ever been explicitly termed an AI-sceptic, but he certainly is one. I'm currently writing an essay on his arguments against the Turing Test, and while they're very intelligent, eloquent and influential, one can't help but get the impression that he is simply against the idea of an intelligent machine or computer program from a very ideological perspective.

It has been suggested that Searle's objections to the possibility of machine intelligence will one day simply be made obsolete by the progress made in the scientific and technological fields; that he and his supporters will have to give up their position when faced with the overwhelming evidence to the contrary. Eventually, proponents of AI claim, there will be a program which is capable of passing the unrestricted Turing Test. I find this a hard assertion to disagree with, as it's a little like denying that there could be intelligent life somewhere in the galaxy - it makes no sense to rule it out simply on the grounds of what we can observe today.

However, I don't think that any evidence will persuade "Searleans" away from their anti-AI prejudice, and Searle's Chinese Room Argument gives them the perfect excuse. It is based on a program that is capable of passing the unrestricted Turing Test, and as such, means that even if a machine did everything it could to fool you into thinking it was a human, the "AI-sceptics" are still able to deny that it thinks for itself, or understands in any way what it's saying and doing. The only reply one can give to this is to say that there is exactly the same amount of empirical evidence for the program's understanding as there is for a human's understanding.

This is the fundamental difference between sceptical doubt and sceptical inquiry - the former can lead to dogmatism and an arrogant refusal to face evidence and adapt one's ideas to suit. The latter is true open-mindedness, in the most beneficial sense. But I'll admit bias in that regard.

Singularity in 2029?

Unsurprisingly, given my vast interest in all things artificial intelligence, this news story leaped out at me from the BBC News Technology page:

Machines 'to match man by 2029'

The point at which machines reach the same level of intelligence of man is known as "singularity", and is something I've been hearing a lot of lately. It was mentioned in last week's Skeptic's Guide podcast, in the context of the crazy "Mayan Apocalypse 2012" topic; apparently one of the ways in which the world might end is through reaching this singularity. Which actually links to the second way it came to my attention - through the Terminator spin-off series The Sarah Connor Chronicles. In both of these incarnations, the singularity is a Very Bad Thing. See also the original Matrix premise regarding the apocalypse brought about by the war between humanity and A.I.

But of course it need not be. The article regarding the possible singularity in 2029 is, it should be stressed, just a prediction - albeit a prediction from a leading expert in the field. I see no reason why his prediction might be wide of the mark; the technology in this area is advancing at a tremendous rate of knots. Reverse-engineering the brain (presumably the human brain) has been identified as one of the 14 major technological challenges facing humanity in this still-young century.

Even if singularity is not reached by 2029, two things are clear to me: firstly, it is itself inevitable for as long as research is done and progress is made; secondly, there's going to be an awful lot of very cool stuff going on by that point - much of which is discussed in that BBC article. Nanobots in particular, helping with the improvement of our own intelligence, fighting disease, and enhancing virtual reality.

As far as I'm concerned, singularity is coming. I'm personally hoping that it arrives sooner rather than later, because that will be a truly interesting time to be alive.

Reframing the question

This semester, I'm going to be studying the philosophy of artificial intelligence, a field which is of great interest to me. To get me in the right frame of mind, I've been watching a fair bit of science fiction lately, particularly that dealing with the future of robotics/cybernetics/whatever. The one with the most philosophy behind it thus far has to be Bicentennial Man, the story of a robot with a "flaw" allowing him to be creative and develop his own character.

While it's hardly groundbreaking, it does give you a good idea of the sorts of issues that exist around the philosophy of artificial intelligence - the main one being, of course, when (if ever) is it a person? As soon as it demonstrates creativity? Or only when it becomes mortal?

Today I received a book I had ordered, called Imitation in Animals and Artifacts. Flicking through this, it occurred to me that a question I had never heard asked was, rather than "when is a robot to be considered a person?", "when is a robot to be considered equivalent to an animal?". Maybe this is a nonsense, or leads nowhere useful, but it's one I'd like to look into in more detail - if only to shed light on the personhood question. Perhaps I'll have the time to do so soon. For now, I'll just jot down some ideas:

If personhood is based on self-awareness, are there animals that we should consider to be persons? And how do we know when something is self-aware, if we have no way to explicitly communicate?

What would be the criteria for animalism?

Does the fact that AI is usually intended to simulate human intelligence mean that this is a pointless debate? Or is this the aim of AI because it has already reached or surpassed the intelligence level of animals?

Is a computer that can beat any human player at chess more intelligent than an animal who lives in a complex social system and adapts to its surroundings?

Maybe this will be my focus for my artificial intelligence module this coming semester.