Sitting in the café this afternoon, a tiny little girl eating a muffin with her grandma saw my laptop and said “he’s got a small puter!”. I gave her a smile, and she declared “I want a small puter!”. I felt simultaneously pleased and somehow as if my masculinity had been insulted 
.
Mark tells an apocryphal story about early research into software translation between English and Russian: The first test sentence for translation from English to Russian was
The spirit is willing but the flesh is weak
which came back, in Russian, as
The vodka is strong but the meat is rotten
To the despisers of the body will I speak my word. I wish them neither to
learn afresh, nor teach anew, but only to bid farewell to their own
bodies,—and thus be dumb.
“Body am I, and soul”—so saith the child. And why should one not speak
like children?
But the awakened one, the knowing one, saith: “Body am I entirely, and
nothing more; and soul is only the name of something in the body.”
The body is a big sagacity, a plurality with one sense, a war and a peace,
a flock and a shepherd.
An instrument of thy body is also thy little sagacity, my brother, which
thou callest “spirit”—a little instrument and plaything of thy big
sagacity.
Friedrich Nietzsche, Thus Spake Zarathustra.
A conversation with a friend started me thinking again about long-term ways to avoid dying. There’s something… darkly heretical about the mere concept, I find. I don’t know if it’s some kind of Judeo-Christian baggage, or (as Iain Banks suggests in the excellent novel The Crow Road) if there is something inherently `uptight’ about wanting to hold onto life for much longer than average.
Having said that, I have a gut feeling that I’d rather not die – at least, not in the next 60 years which are my statistical expectation, and certainly not in the immediate future, which is always a possibility as things stand currently. In fact, when I really think about it, a large part of what motivates my research and interests is this vague, nagging desire to find a way of… escaping a dying body, maybe? I’m not sure how to describe that.
This desire to drive a wedge between `the spirit’ and `the flesh’ sounds quite religious, in a way, doesn’t it? Which should make me cautious: I think a lot of religious ideas are there because they feel comfortable.
Current artificial intelligence research sucks. Alan Turing, at the start of the digital computer age, thought that the AI problem was all but solved – it seemed intuitive to him that a computer capable of modifying its own instructions computationally could `think’. About 60 years later, we have some pretty sophisticated software, but I don’t think it thinks. Tests where judges chat with AI systems to differentiate them from humans still conclude in just a few sentences (usually with inadequate answers to questions like “where are you?” and “are you north or south of the equator?”).
Current AI systems are fairly high-level – the usual approach is to build a system capable of organising knowledge in some way. The area I’ve been studyingin, language technology, deals with a subset of the AI problem – that of understanding and speaking a natural language. It’s pretty common to equate language and thinking: I cannot imagine having a complex, abstract thought without the words to express it. I can think about doing something simple (walk to the fridge, pour some juice) without describing it in words, but I find it impossible to do mental arithmetic or something like that without thinking the names of the numbers I’m using. I think that this close relationship between language and thought has given people to the same mistake as Turing: If only the grammar describing our language/thinking-system could be worked out, we would have a description of how we think. In fact, we seem to have discovered that English grammar cannot be described as a set of rules. It clearly makes sense to humans – we can understand the speech of others and make more-or-less grammatically correct utterances from an early age. Maybe this is an indicator that attempts to describe a natural language using grammars and rules are incorrect.
I think we need toapproach AI at a much lower level. Rudy Rucker, in a recent Boing Boing post, said:
I’m estimating that you need an exaflop machine in order to have the hardware to simulate a human brain. But you would need a wattaflop machine in order to evolve the necessary neural net (?) software to simulate a human brain. You need so much power because finding some human-like brain software is a huge search problem, to be done by basically simulating three billion years of evolution.
Really the way to go is, of course, to turn to biology. But we’re already doing that when we fuck. Who needs more machines anyway?
So most AI researchers aren’t actually trying to create intelligence. They’re trying to systematise it, understand how it works so fully that they can simply write a set of rules describing it and have it work. I’m not so interested in that; like Rucker, I would be satisfied with a biochemical simulation of a brain, even if this means giving up trying to understand how it actually works, for the time being.
I’ve been playing with the BeOS a little recently. Unfortunately, it’s starting to get a little long in the tooth: It won’t support the i810(?) graphics in the Pentium2 on my desk at uni, and apparently it won’t boot on CPUs with the latest SSE-type extensions.
The plan instead is to boot it under Bochs and use it there. It seems like “BeOS 5 Personal Edition” (the free one) must be booted (initially at least) using a floppy disk and a system image on the hard disk, in the partition of a host operating system (either Gnu/Linux or Windows). So far, I’ve created a 700mb disk image, and a single partition on it formatted as ext2 (this tutorial on creating disk images using the loopback device is incredibly informative). Into this partition, I copied the ~500mb BeOS image (“image.be”). I now expect to be able to use Bochs to boot a combination of this disk image and the floppy image from BeOS5-PE. We shall see how that goes…