November 03, 2006

TLSX: reverse engineering the language module

I'm in Austin for TLSX ("Texas Linguistic Society, Ten"), an annual conference run by grad students at UT. This year's theme is "the application of techniques from computational linguistics to descriptive linguistics and the analysis of less-studied languages".

I'll blog from the conference site as time permits. I'm going to start with an entry that does have to do with linguistics and with blogging, but not with TLSX (though maybe there'll be some sort of connection, who knows). The subject is Ken Macleod's new novel, "Learning the world: A scientific romance", and I've had it on my to-blog list for a month or so.

From the cover blurb:

Humanity has spread to every star within five hundred light-years of its half-forgotten origin, coloring the sky with a haze of habitats. Societies rise and fall. Incautious experiments burn fast and fade. On the fringes, less modified humans get on with the job of settling a universe that has, so far, been empty of intelligent life.

Being at a conference run by grad students reminds me: does the appeal of space-colonization stories comes partly from their success as a metaphor for the experience of young people starting out to find a place in the world?

More from the cover blurb:

The ancient starship But the Sky, My Lady! The Sky! is entering orbit around a promising new system after a four-hundred-year journey. For its long-lived inhabitants, the centuries have been busy. Now a younger generation is eager to settle the system. The ship is a seed-pod ready to burst.

Graduate school does seem to last for centuries, for some people, and UT does have an unusually large linguistics department. And perhaps the UT department is entering intellectual orbit around some new topics. But I think this joke has been pushed far enough.

One of the Learning the World's narrative threads comes from a young girl's blog. The book opens with her first entries:

13 364:05:12 16:24

The world is four thousand years old. I was eight years old when I found that out for myself. My name is Atomic Discourse Gale and this is the first time I have written something that anyone in the world can read. It is strange and makes me feel a little self-conscious, but I reassure myself that not many people will read it anyway.

14 364:05:13 18:30

That was a joke. I see I have a few readers. J---- wants to know how I found out the age of the world. It was six years ago now but I remember it quite well. I was very young then and didn't understand everything that happened, but looking back I can see that it was a significant event in my life. That is why I mentioned it. So this is what happened.

OK, let's get to the point. The second planet of the new star system is inhabited by sentient bat-creatures, who have reached a roughly Victorian-era level of technology. Their economy is based on slave labor provided by trudges, members of a semi-sentient related species -- roughly as if humans had succeeded in domesticating and enslaving chimpanzees.

In their spread across the galaxy, humans have never encountered any life much above the level of slime moulds. But they've got ethical principles in place, all the same, which say that they should leave the bat-people alone, except for some discreet observation mediated by bioengineered beetles and the like.. But what does non-interference mean, in this case? Would colonizing the system's other planets, ithe asteroid belt, etc., be OK? Maybe not, since the bat people are on a track to invent space flight and do it themselves. Still, a dissident faction plans to force the issue by breaking off part of the ship (it's designed to work that way) and starting the colonization process anyhow.

And is non-interference really the right policy, given the bat people's warring societies and their cruel treatment of the trudges? Some crew members don't think so.

In this context, a bit of genetic hackery goes badly wrong -- or maybe exactly right, depending on which level of whose plans you attend to. From p. 281 of the TOR paperback

She reviewed what he had told her, replaying the words and sentences her anger had whited out and shouted down the first time.

The problem, the intellectual problem, was this. No Rosetta stone existed for the bat people's language. No amount of observation, no iteration of linguistic heuristics, could decode an unknown language from recordings alone. For mutual understanding, there had to be mutual interaction. One had to know directly what one side of the conversation was trying to say, and that meant one side of it had to be you. Faced with this impasse, the crew's scientists had, in all too characteristic a fashion, worked around it. Their solution had all the grubby fingerprints of a brute-force kludge.

The neural structure of the human brain's language-processing module, named in deep antiquity Chomsky's Conceit, had been known since the Caves. The genetic code of the Destiny II biosphere was known from aerial microorganisms returned to the stealth orbiter. The amount of information and genetic instruction that could be packed in a naonassembler was vaster by far than even the vast amount stored in natural genomes and machinery, cluttered as tehy were with redundancy and junk. The information-processing hardware capacity of the ship was beyond all human conception, and the amount of information its sceince software could extract from the slenderest and most fragile of evidence was limited only by the ingenuity of the human inquiry that initiated it.

So . . . they'd had the means to install Chomsky's Conceit on any big enough brain down below. They had the means to generate radio transmitters within host bodies, as they'd done with the dung-beetles. And faced with the crash-and-burn and banning of that project, they'd skipped blithely ahead to a bolder one. They couldn't install Chomsky's Conceit on the brain of bat people -- the aliens' brains already had a language module of their own. That would have given rise to wetwoare conflicts and deep grammar errors, and anyway, ethically, that would never have done. Oh no. That would have been wrong. That would have interfered. What they had done, bless their reckless little souls, was to set up the machinery to install the module on the brains of the slaves, who had (they'd figured) no language module (and who were, therefore, not slaves but beasts). And once thye'd received and filtered and processed and quantum-handwaved the information coming back from brains learning the bat people's languages, the translation protocols had been ---

Reverse-engineered from the language module!

Holy rocking shit.

From a narrow technical point of view, this is just clever intelligence work: getting the information needed to be able to understand what the bat people are saying to one another. But from another point of view:

"You realise what you've done?" she demanded. "Do you have the faintest conception of the harm this will cause?"

Constantine nodded. "The disruption will be immense. It'll destroy the entire slave economy."

"But they're not slaves!" Synchronic said. "If they had been, I could see why we might want to interfere. But you've taken waht are by your admission mute brutes, and given them language. Deep grammar. Self-awareness. Human consciousness. You've made them slaves."

Caliban writ large. (No, not Taliban -- that's a completely different problem.) For a later post: why Macleod's future humans have misread Chomsky. If you're really curious about this, you could check out these earlier posts:

"Homo hemingwayensis" (1/9/2005)
"Chomsky testifies in Kansas" (5/6/2005)
" JP versus FHC+CHF versus PJ versus HCF" (8/25/2005)

Posted by Mark Liberman at November 3, 2006 08:04 AM