After years of failing to develop artificial intelligence, most scientists gave up, says Cade Metz in his new book, Genius Makers. But a few dogged professors kept trying – including a British computer scientist called Geoff Hinton.
In 2012, Hinton and his students at the University of Toronto made the big breakthrough. They built a “neural network” – a mathematical model based on the brain – that could identify everyday objects in a way that had “previously seemed impossible”. Big Tech companies knew this was a big deal and tried to hire Hinton. Instead he set up a company, DNNresearch, consisting of “three employees [and] no products”, and put it up for sale to the highest bidder.
Hinton hosted the auction at Harrah’s, a hotel on the shore of Lake Tahoe, California. Four bidders made their pitch: Google, Microsoft, the London-based DeepMind and the Chinese firm Baidu. None knew the others’ identities and the auction was played out over email. After each round of bidding, the companies had an hour to raise the buying price by at least $1m.
DeepMind, which was later bought by Google, quickly dropped out. Microsoft soon followed. But Google and Baidu kept bidding up: “$25m, $30m, $35m.” At $44m, Hinton stopped the auction. He went with Google – which, as he later admitted, “he had wanted all along”.
The representatives from Baidu and Microsoft happened to be on the same plane home, and “spent hours” chatting about their field. Bound by company confidentiality, neither made any mention of their involvement in the auction. But both knew the significance of what had just happened – their respective employers “would have to answer Google’s big move”. It was the “starting gun” in the global race for AI supremacy.
Read more in this extract from Wired.