[ by Charles Cameron — on one current comparative advantage of being human, and calling for the design of a ReSearch Engine ]
.
You may not like hymns — or thrash metal. Facebook, whose market value topped $100 billion about a year ago, “thought” that if I liked this video:
I might also like this “related video”:
State of the art! Big Data! Analogical thinking!
Seems like the algorithm didn’t listen to the music, it just decided “King of Heaven” and “in Heaven, King” were pretty similar as word-groups go.
Actually, their reasoning is not that bad, once you think about it in DoubleQuotes terms — they’ve stumbled on an “opposite” rather than a “similar” — but as we’ve seen with such examples as Oxford and Cambridge, or the Army / Navy game, opposites and similars aren’t so dissimilar after all.
Sadly, when it comes to musical tastes, opposites don’t necessarily work too well, and similars would in this case have been preferable.
**
But the issue is human cognition and the attempts of computer scientists to match it — and specifically, to match and even surpass our analogical powers.
As I wrote in WikiLeaks: Critical Foreign Dependencies, in an online chat session with David Gelernter years ago, I said:
My own hunch is that an aesthetic sense is *the great sorting principle*, that it has to do with pattern recognition, and specifically the recognition of isomorphisms, parallelisms in deep structure. So an AI that recognized deep isomorphisms across wide topic distances would be the ideal web navigator, as an I that recognizes deep isomorphisms across wide topic distances is a creative mind. It would also be playing Hesse’s Bead Game, no?
to which he responded:
Hipbone, I think basically, that’s exactly right. I wrote a book about this issue of what you call recognizing isomorphisms in widely different domains, a tremendously important issue in how the human mind works.
From my POV, the human mind recognizing a rich correspondence between two rich insights, perhaps even from widely separate domains, is the very essence of creativity — isn’t that what the Taniyama-Shimura conjecture – and thus the eventual proof of Fermat’s last theorem – was all about?
My brief chat with Gelernter dates to 1998, his book The Muse in the Machine: Computerizing the Poetry of Human Thought, to 1994. On pp. 2-3, he writes:
Reasoning is one big part of human thought, and thought science has reasoning decently under control. Philosophers and psychologists understand it and computers, up to a point, can fake it. But there is one other big piece of the picture, which goes by many names: creativity, intuition, insight, metaphoric thinking, “holistic thinking”; all these tricks boil down at base to drawing analogies. Inventing a new analogy — hitching two thoughts together, sometimes two superficially unrelated thoughts — brings about a new metaphor and, it is generally agreed, drives creativity as well. Studies (and intuition) suggest that creativity hinges on seeing an old problem in a new way, and this so-called “restructuring” process boils down at base to the discovery of new analogies. How analogical thinking works is the great unsolved problem, the unknowable longitude, of thought science. “It is striking that,” as the philosopher Jerry Fodor remarks, “while everybody thinks analogical reasoning is an important ingredient in all sorts of cognitive achievements that we prize, nobody knows anything about how it works” — not even, Fodor adds (twisting the knife) in an “in the glass darkly sort of way” (1983, 107)
**
For a comparable, consider this NYT evaluation of another tricky issue for AI — Brainy, Yes, but Far From Handy:
The correlation between highly evolved artificial intelligence and physical ineptness even has a name: Moravec’s paradox, after the robotics pioneer Hans Moravec, who wrote in 1988, “It is comparatively easy to make computers exhibit adult-level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a 1-year-old when it comes to perception and mobility.”
Brainy the current AI’s may be, and even beginning to manage physical agility — but mentally agile?
If they still can’t tell that a taste for classic hymns does not correlate closely with a taste for German thrash, they’re not agile enough for the HipBone / Sembl style of games..
**
Derek Robinson wrote a piece about my HipBone Games and AI back in the 1990s. It’s succinct, it’s relevant.
Here’s how I see these matters: I am calling for the development of a ReSearch Engine, with the HipBone Games, Sembl and DoubleQuotes as devices to be used in its construction.
The ReSearch Engine’s purpose would be to learn from humanly identified analogies — gleaned from repeated playings of the HipBone, Sembl and DoubleQuotes games — to recognize deep and richly textured analogies across the breadth of human cultures, following the principle laid out above:
deep isomorphisms across wide topic distances
Such an Engine could hopefully provide us with the links of associative links that at the moment are glimpsed in moments of genius (think: Taniyama‘s conjecture of 1956 connecting the mathematical realm of elliptic curves and that of modular forms), which then take years to be ironed out and brought to fruition (think: Wiles‘ proof of the Taniyama–Shimura–Weil conjecture, along the way to his proof of Fermat’s Last Theorem, 1993).
The successful design of such an Engine would be a — hmmm– singular event.
