Thinking is easy

 I've been hearing that we'll have machines that can actually think in five years for at least forty, and it has filled me with an obscure dread. I make no bones about it: I'm scared of dying and most if not all of that fear is that the something which is consciousness will just stop working, and there will be nothing, an end. I don't understand people who are not terrified of that, I really don't. Are they not so firmly attacjed to their lives? Are they just scared of being scared? I don't know.

So when I hear about machines that actually think, I'm not upset that man will be replaced, or shown to be inferior, that's not it at all. It's that what I am, all my life, every experience, every insight, every appreciation, is an easy to assemble and disassemble process, clap on, clap off. I want to resist it, of course, but I also know that just because I don't like it one bit doesn't mean it isn't true. (I also know that just because it's dark, unfavorable to man, and reductionistic doesn't mean it is true.) (Nor that just because the Catholic Church is full of shit does not mean there is no spiritual life.) So I wrestle in the dark. And I blog about it.

Now philosophy can be defined as running around in circles until you get somewhere. Philosophy does not answer the Big Questions. If you read philosophy and you see something that looks like an answer. it's either a) bad philosophy, b) not philosophy at all, or c) not an answer but a question gussied up to look like an answer. 

The first thing I realized about the problem, long long ago was that I had been fusing consciousness with thinking. They don't seem the same: you can be conscious but not thinking. They may ultimately be aspects of the same thing, but they are different aspects.

And that's the fascinating part about this big machine endeavor: by talking about thinking as a very specific quality, it prompts the question what do you mean by that? Thinking?

Now Martin Heidegger is a brilliant man, and a great philosopher. I have not read him since college,but his writing was utterly dense, utterly careful, and dangerous to read in translation. But when you finish reading a passage step by step, he made wonderful, challenging sense. He gave a series of lectures that became a book called Was heisst Denken? treacherously translated as What is Called Thinking? Were this something other than a blog entry, I shouldn't have started this without spending a couple of months inside the book, but seeing as it is what it is, I'll admit my ignorance, and instead pull some random quotes off the Net. 

We come to know what it means to think when we ourselves are thinking. If our attempt is to be successful, we must be ready to learn thinking…As soon as we allow ourselves to become involved in such learning, we have admitted 

that we are not yet capable of thinking. (p. 369)

 Well, of course, you say. The only way we find anything out is by thinking, and that has to include thinking itself. It's a nasty little recursive loop, but OK. It does sound like the line from T.S. Eliot's Four Quartets: Only through time is time conquered. And if we have to ask the question, we clearly don't have it down.

But the A.I./expert systems/machine intelligence people have taken it from the realm of teutonic vaporising and made it a product feature. It thinks. What the hell does that mean?

So let's try a bunch of questions.

What do we do when we think?

Is remembering the same as thinking?

Is seeing or hearing the same as thinking?

Do we need logic to think?

Do we need language to think?

What do we get when we think? Is it always a thought?

Is creativity thinking?

If you come up with something that isn't new, is it still thinking?

What is a thought? Is it language-bound, or can you have pictures in it? Is it the whole process, or just the conclusion? Can you store a thought? Is the memory of a thought the thought? 

It'll be fun to answer these questions in detail, but I don't have all day. But conditioned as we are and shaped by language, it still seems a) that we distinguish between thinking and other things we do in the inside of our skulls; b) it's something we decide to do (which makes that decision--let's call it an act of will--is distinct from thinking); c) We can actually do quite a lot without thinking--we don't consider acting on autopilot to be thinking; d) although logic is certainly not required for thinking, something that resembles logic, however loosely, seems to be at work when we think; e) It seems hard to envision thinking without language, but that may be because language is wrapped up in nearly everything we do.

But is creativity thinking? A composer or a musician that improvises could be said to be thinking without language, as could a painter--but does that process seem like thinking? It's conscious, certainly, and intentional, and it's a complex task--but does the invention come from someplace else? (If we're really attached to language, we could say that music and painting are languages in their own right--but, especially in music, is their meaning involved? Structure, but no content--grammar without vocabulary. Hmm--that's worth talking about later.)

We can be aware without thinking. We can act without thinking. We can remember without thinking. We can decide to do things without thinking. And maybe we can even create without thinking. At least by what we tend to call thinking. So consciousness, action, memory, will, and maybe creativity are distinct from thinking, although when we do think, many of these things are involved intimately. 

One of the problems with machines that think is that the people talking about it are scientists, engineers and academics. These are all people who make their money with thinking. They think to solve problems. For the most part they enjoy doing this, and get excited, energized, and engaged. They think, probably rightly, that it's one of mankind's highest, most noble and most beautiful functions. And that when they talk about machines thinking, they're talking about machines solving problems.

The problem with that is that's not the only thing thinking does. Even they use thinking for other things than solving problems--just not professionally. 

Let me anatomize thinking for a second. Three of the things that thinking does are 1) judgement, 2) ratiocination, and 3) speculation. We think when we do these things, but they aren't the same. You can judge without a process of deduction or a flight of imagination; you can puzzle something out without putting values to it or speculating upon it; and you can envision something without building it up by a chain of reasoning or criticizing it. 

Now it could very well be that on a deep level they're all really the same process, and that they are ultimately ratiocinative acts: all I'm saying here is that that's how it seems to the amateur thinker. There may also be a deep connection between thinking and awareness: in fact, I suspect there probbly is. It may even be that thinking gives rise to awareness, and not the other way around. It may be that the two of them build on each other. I don't know. (he says that a lot, doesn't he?)

But here's the point: the people who are theorizing about artificial intelligence seem to equate thinking with problem solving. That, of course, is a very big thing: it's by solving problems that we change things, and by which we would be replaced. Building an AI that just sat and looked at the sunset and went 'ooohh..." would not seem worth the effort, or even interesting. Definitely not threatening to our exalted status in the universe. The thing is, though, that problem solving is the least problematic part of thinking. Ratiocination, deduction, analysis--we know a vast amount about these processes. We even know quite a bit about the shabby alternatives to it by which people fool themselves and become Republicans. 

We've been building machines to solve problems since we've been building machines, and we've been building machines to solve informational problems for quite a while now. As a matter of fact, we tend to have machines with problem solving capacity far in excess of our ability to supply them. That's the key, of course; just as ploughs are great problem solvers as long as the problem is how to break up ground for planting, computers have tended to be great problem solvers as long as the problems could be solved with numbers. We have made great progress in learning how to formulate increasingly complex problems in terms of numbers and rules. We have also, as Roger Penrose has pointed out, learned that there are problems unsolvable by computation. And we've learned that some things that seemed simple, like seeing, are a lot tougher and more complex than common sense would have it. Charles Stross, a splendid current science fiction writer with great extrapolative mind and wtf writing style, threw off a vision of the whole interior solar system turned completely into computronium (in the midst of half a dozen other wonders) which prompted me to wonder whether there was anything that such a computer woud be required for. Between NP-complete propositions and quantum uncertainty--not to mention the problems of input, might the horizons of problem solving be circumscribed? Because size and complexity aren't the only criteria for a problem. It also has to be, well, a problem.

I think that the problem that the AI people have with talking about machines that think--and the root of my fearful discomfort with their talk--is that they not only conflate awareness and thinking, they also conflate thinking with problem solving. Despite plenty of evidence in their own experience, they say that thinking is just problem solving , and awareness is just thinking. Somebody might say that I'm just heavily disguising the old Star Trek chestnut 'But can computers...feel? Can...they love?" That's not what I'm saying, though it is related.

I think this has to do with the core of science. Science, when yo come down to it, is the study of the way kinds of things behave. Science , ever since newton, talks about how things work. Can science tell me what an electron is made of? Specifically, can you say anything about the nature of an electron that does not affect the way it behaves? It's certainly not impossible to conceive of such a thing--but a scientist would say (perhaps dismissively) that it's beyond the realm of science. And therefore meaningless.

(for example, there is a quality of an electron that doesn't affect its behavior: its history.)

But it's that fundamental orientation that colors their talk of thinking machines: Not only is the only thing they recognize, it's also the only thing that in some way justifies our existence. Sure, they might say, there might be other aspects of thinking than problem solving, but what are they good for? 

THese wonderful machines envisioned will without a doubt solve problems better than us--but so does a plough, a steam engine, a book, and a Macintosh. We have been solving solving problems for thousands of years. Much of the time it's been by outsourcing. It's made us less heroic,less self-sufficient, less, ultimately in charge. And maybe they're right: maybe that will doom us.

Of course, maybe we'll use all that problem solving ability to make our own mind mechanism superior, Maybe with all that computronium, we'll come up with a better model than individual competitive units.

Or maybe we'll look at the rest of thinking that isn't easily defined, purposeful, and useful, and find things there we want to do. Like appreciate and imagine--or cut off the soles of your shoes and sit in a tree and learn to play the flute. 

Because if there is no God, what's the use of use? Ultimately (and not all that ultimately, all that happens is that matter shifts around. And if there is a God, then maybe she wants us to look at sunsets and go 'ooooh.' Hard to tell.

We can learn only if we can unlearn at the same time…we can learn thinking only if we can radically unlearn what thinking has been traditionally. (p.374)