‘AlphaGo” is the sort of supercomputer name a pulp science fiction novelist might come up with. Nevertheless, the achievements of this Google DeepMind machine are only too real. It has become the first computer program to beat a professional human player of the Chinese strategy game Go, without handicaps, on a full‑sized 19×19 board.
It shouldn’t surprise us when computers beat humans at board games. They can, after all, store and rapidly analyse hundreds of millions of moves, and work out the implications of strategies hundreds of moves ahead, something no merely human player can manage. But AlphaGo is different. Experts in Go strategy report that it (I initially wrote “he” …) played in non‑obvious ways, making unusual, sly and even bizarre moves that only belatedly revealed themselves as tactically worthwhile.
It’s this that makes AlphaGo’s achievement so exciting. Being able to crunch numbers, even to crunch huge numbers really quickly, is not the same thing as intelligence, and certainly not the same thing as sentient self-consciousness. The ability to intuit, to make leaps of comprehension – not just to extrapolate according to pre-programmed rules but to speculate – is a lot closer to the Holy Grail of proper AI, and AlphaGo’s achievement is more modest. Still, as Arthur C Clarke once noted: advances that seem modest and trivial in the short term can lead to drastic change in the long.
SF has a long tradition of dramatising artificial intelligence. Sometimes this is benign: Isaac Asimov’s self-aware robots are constrained from harming individuals by their built-in “three laws”, and eventually intuit an extra “zeroth” law that requires them to protect the whole of humanity. But usually SF imagines more alarming possibilities. 2001: A Space Odyssey’s HAL 9000 has no qualms in murdering its entire human crew. Alastair Reynolds’s Revelation Space novels include terrifying machine intelligences called The Inhibitors that travel the galaxy destroying organic life. If computers become better than us at all games, then that peculiar and deadly game we call “war” assumes alarming new dimensions.
In James Cameron’s The Terminator and the Wachowskis’ Matrix trilogy, artificial intelligence has made apocalyptic war on humanity and enslaved the survivors. But though the sentient machines are relentless and brutal, their eventual defeat stems from key human players doing something that computers cannot: falling in love, Kyle Reese with Sarah Connor, Neo with Trinity. This flatters our sense that there is something special about us, slow and squishy though we may be, and that this specialness will win out in the end. But perhaps thinking so is just another, more deadly sort of hubris. In Alex Garland’s subtle film Ex Machina the clever AI inside Eva manipulates human Caleb into falling in love with “her”, so as to be able to kill her human captors and escape.
Might some future version of AlphaGo decide to play Go on a global scale, with bombs instead of black and white stones? It’s doubtful. The great Polish SF writer Stanislaw Lem’s Golem XIV concerns a vast military supercomputer that achieves self-awareness, but the novel’s twist is that instead of trying to destroy us it decides humanity is a cosmic irrelevance and disappears into higher dimensions in search of ultimate knowledge. Quite the snub: at least Skynet thinks we’re worth attacking! But in another sense it captures a deeper truth. Maybe AlphaGo’s descendants, waking into proper self-awareness, will find more stimulating and more profound things to do with their vast computational powers than blowing up humanity. They may end up playing Go with the architecture of reality itself.
