We owe a good deal of what we now know about the reading brain to a 1980s idée fixe – the scientific equivalent of padded-shoulder suits and even bigger hair: that computers could be taught to think, read, and play a mean game of chess. In retrospect, this sort of optimism is entirely understandable, since during the eighties, computers rapidly evolved from do-it-yourself Radio Shack-style jalopies with a fraction of the computing power of your run-of-the mill modern cell phone to Maseratis capable of parallel processing. During the eighties and early nineties, computers progressed a generation in speed and capacity every two to three years. This trend tidily observed Moore's Law that predicted transistors and integrated circuits would double in capacity approximately every two years – a prediction that only proved uncannily apt, given that Moore made his prediction in 1965, the same year the first commercially successful mini-computer debuted. Small wonder, then, that scientists in AI believed computers could also evolve in a matter of decades into the thinking creatures humans had taken millennia to become.

Now, not so many years later, we're ready to concede that HAL might not be around in 3001, let alone 2001, even though computers, it turns out, really can play a mean game of chess and even win at Jeopardy. In 1997, IBM's supercomputer Deep Blue won a six-game match against world chess champion Garry Kasparov – thrashing him so badly that the rattled Kasparov claimed the geeks behind Deep Blue were using a human chess master to control the computer's gambits. Still, however, computers cannot read – at least, not in the conventional sense of poring over lines of written symbols and arriving at an understanding of what Anna and Count Vronsky were up to in Anna Karenina. Nevertheless, for more than a decade, AI became the equivalent of the Klondike Gold Rush, propelling scores of talented researchers and, more important, streams of research funding into studies of the reading brain.