Well, not me per se, but my AI and computational linguistics training seems to have been overtaken by the march of progress. I refer to Androidal Systems, Inc., whose breakthrough processor “understands the meaning of language.” Yup, the meaning of language — these folk aren’t into understatement. Universal translation, that’s what they’ve cracked. And they mean universal; the world’s 6500 languages, all existing programming languages and hardware descriptions, they even have a patent (granted 2002) mentioning “arbitrary sensory/motor receptions” for human/machine interaction. Impressive! Below the fold, more excerpts from this significant work, and my noises of awed appreciation.

So what are they doing? Quite simple really: they’re translating every language according their own proprietory Universal Grammar. It’s a common idea: to translate from language A to language B, you extract the UG form of an expression in A, then build up the expression in B expressing the same UG form. In the crazy world of academia we usually refer to this “UG” as “logical form”, sometimes even going so far as to suggest particular formalisms, like predicate logic, as candidates. It’s becoming clear where we went wrong.

Because what we (the academics) are missing is the link between language and logical form. It’s the up-and-down translation from/to our logical form (“UG” for ASI) that we’ve been struggling with for as long as we’ve been aware of the problem. It turns out we’ve been using the wrong logical form; ASI has discovered “a universal grammar that explains the basic structure of knowledge and the nature of our expression of language”. And it’s this that makes the whole translation problem basically trivial:

The knowledge processor’s operation is transparent to the user and eliminates the need for designing either software or hardware through conventional means, since the KP designs the integrated machine platforms by its own intelligence specifically to implement the user’s language. (ASI, product applications)

This apparently solves the industry’s long-standing problem: “to construct a ‘universal computer’”). This just goes to show how out of touch we academics are, I was under the impression that we already had. Shows how much I know…

Aha! But reading further, I find that they’re letting out their trade secrets on the company website! They point out that a fundamental problem of AI is that human life “does not abide exclusively by a theory of the physical universe”, and that their androids (yes, they call them ‘androids’) are “guided by an endless synthetic ‘consciousness’”. Not only that, but the machine is “controlled by the manipulation of language-words from Roget’s thesaurus”. It seems we academics have been using a substandard thesaurus all these years. Now we know better, I suppose.

As I said, this technology is patented. I’m still working through the text, trying to find if there are any loopholes left to allow me some remnant of an academic career. It seems to be pretty comprehensive though. The technology, for instance, “comprise[s] at least one of discrete automations, continuous time systems, electronic devices, electromechnical devices, biological devices, and chemical devices.” (All quotes are from the USPTO database entry linked to above, which has no section anchors, I’m sorry.) Does this cover quantum computers? Perhaps there’s hope for me yet…

(As a matter purely of interest, does anyone know what “Rg” standardly abbreviates? I mention it because the patent refers to the ‘second aspect of existential form’ as an “Rg Module,” but doesn’t say why. In case you’re interested, the first aspect of existential form is “the embodiment of human being.” It’s unclear whether this ‘aspect’ is included in the patent itself; if so, it would seem that someone was a bit lax in searching for prior art.)

At this point I run into rather a stumbling block: I can’t see the 169 illustrations (browser/plugin problem, I should get around to fixing it…). And it seems that they would be decidedly helpful for a full understanding of this text. Between this and my lack of patience for legal patent-speak, I’m left not really able to understand exactly what is being patented here, except that (a) it seems pretty comprehensive, and (b) it probably makes my entire field of study, in fact most of human science and reasonable chunks of established religion and philosophy, redundant. There is good news, and bad news. Good news: there is a book explaining it all. Bad news: it costs $59 (US). Presumably it includes those tantalising illustrations, but I might have to leave this one until my finances improve.

At least one thing is reassuring: at that price, I guess not so many people on my future grant application committees will have read it. Perhaps, if I’m lucky, I can make a reasonable living as an academic in the ‘uptake period,’ before this revolutionary and truly transformative technology becomes generally accepted.

Otherwise, well, I for one welcome our new android overlords.