George Dyson’s Turing’s Cathedral essay is, as John Battelle says, “real joints-after-midnight material.” But worth reading for its role in stimulating the brain. Relevant quote:
For 30 years I have been wondering, what indication of its existence might we expect from a true AI? Certainly not any explicit revelation, which might spark a movement to pull the plug. Anomalous accumulation or creation of wealth might be a sign, or an unquenchable thirst for raw information, storage space, and processing cycles, or a concerted attempt to secure an uninterrupted, autonomous power supply. But the real sign, I suspect, would be a circle of cheerful, contented, intellectually and physically well-nourished people surrounding the AI.
As we move from specialized, narrow-purpose AI (the kind of problems that when solved prompt people to say ‘well, that wasn’t part of AI after all’) to AI that is capable of passing a Feigenbaum test (never mind a Turing test), I’ve been assuming that the public will know about it. If a research institution makes an AI breakthrough, they’ll publish it. If a corporation makes an AI breakthrough, they’ll patent, monetize, and license it. But Dyson’s essay combined with Google’s secretiveness makes me wonder if that assumption is correct.
Technological discovery on the whole is increasing exponentially (see Moore’s Law for the most famous example) but innovation in each field follows a series of ‘S-curves’ – a new rule-changing paradigm is discovered (say, the transistor), technological advance is slow at first, change then accelerates exponentially, and finally tapers off at the end when the physical limits of the paradigm are reached. (Then, a new paradigm begins.) We’re busting through new S-curves faster and faster as progress accelerates. If a company discovered the start of a new paradigm, and had the critical mass of brainpower to innovate quickly on its own, wouldn’t it make sense for it to try and keep the discovery of the new paradigm a secret, and to try and reach the exponential acceleration part of the S-curve before its competitors discovered the new paradigm’s very existence? Especially in the field of AI, where success begets more success as the AI itself begins doing the learning and discovery for you, this could enable a company to make a qualitative break from its competition, securing dominance.
It would therefore make sense for a company to hide any jump in technology from both the public and competitors for as long as possible, wouldn’t it? But it might not make sense for the public, since the development of strong AI marks the advent of a technological singularity and the morality of that first AI has immense potential consequences for all of us mere meat-machines. (For further general reading, see the work of Singularity Institute for Artificial Intelligence and Shannon Larratt’s less-staid ‘Body Modification’s Role in the Coming Robot-Human Apocalypse‘.) Continued AI research has immense benefits but also immense risks – enough risks that the public really should become involved when a company begins to make rapid advances, unanticipated by those not used to thinking in terms of exponential growth. Which leads me to the title of this post – ‘hidden AI detection.’ What tests can we as outside observers make to tell if any one organization is hiding rapid advances in AI from the public? Should we begin devising and applying these tests now?
I don’t want to be misinterpreted – I don’t believe Google has HAL or SkyNET in its basement. But I suspect Google does have an active AI research program that’s worth watching. In the Dyson article, one of his hosts at Google states that “We are not scanning all those books to be read by people [...] We are scanning them to be read by an AI.” In February 2005, Google’s Machine Translation Systems group presented as part of the Google Factory Tour (good write-up here) – revealing that they were developing systems capable of understanding natural language. And the Google AdWords blackbox – much less transparent than Yahoo Search Marketing’s – monetizes much better than Yahoo Search Marketing’s. See their latest quarterly revenues – $1.578 billion, up 96% from the same quarter last year. Yahoo, with a lot more business units, did $1.33 billion, up 47%. Could this differential a consequence of a superior narrow-purpose AI?
Footnote: All of this entry is influenced, heavily, by my recent reading of Ray Kurzweil’s The Singularity Is Near, which I highly recommend.
