Demis Hassabis, the founder and CEO of DeepMind, announced at the Neural Information Processing Systems conference (NIPS 2017) last week that DeepMind's new AlphaZero program achieved a superhuman level of play in chess within 24 hours.
The program started from random play, given no domain knowledge except the game rules, according to an arXiv paper by DeepMind researchers published Dec. 5.
“It doesn't play like a human, and it doesn't play like a program,” said Hassabis, an expert chess player himself. “It plays in a third, almost alien, way. It's like chess from another dimension.”
I started programming IBM machines in the late 60s, and at the time there was talk about the possibility of a computer someday beating a competent human at chess. Though the first programs stumbled along like children learning to walk, slowly, over the years, they improved, thanks in part to Moore’s Law and the genius of certain computer scientists. In February 1977 Chess 4.6, the only computer entry, won the 84th Minnesota Open against competitors just under Master level; it later defeated the US chess champion. [source] In 1988, Deep Thought became the first computer to defeat a grandmaster in a tournament. IBM bought Deep Thought, pumped it up and renamed it Deep Blue, and beat World Chess Champion Garry Kasparov in 1997.
Today, the chess prowess of Deep Blue is available on our laptops, or even in our pockets, on handhelds. The seven foot tall mainframe towers that housed Deep Blue’s “mind” are gone, and strong computer chess is a commonplace . . . [source]
These programs were essentially “taught” chess by human experts. They were one-trick ponies: great at chess but nothing else. The next step was to develop an algorithm that could learn from first principles (rules), enabling it to play chess and other challenging games at a high level.
The London-based DeepMind researchers pursued this goal and developed AlphaZero:
Instead of looking at games like Chess and Go as search problems, [the creators of AlphaZero] treated them as reinforcement learning problems. Reinforcement learning may sound vaguely familiar if you took an Intro to Psychology class in college; it’s precisely the way humans learn. . . .
The mathematical basis of how we apply reinforcement learning as humans has been painstakingly worked out over the last 30 years. That brings us to AlphaZero. By simply playing against itself for a mere 4 hours, the equivalent of over 22 million training games, AlphaZero learned the relevant associations with the various chess moves and their outcomes. . . .
Deep reinforcement learning is nothing less than a watershed for AI, and by extension humanity. With the advent of such über-algorithms capable of learning new skills within a matter of hours, and with no human intervention or assistance, we may be looking at the first instance of superintelligence on the planet. [emphasis added]
In a paper presenting the AlphaZero algorithm, the developers claimed that “Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) as well as Go, and convincingly defeated a world-champion program in each case.”
What will AlphaZero be doing in three years? Five? Will we be carrying AlphaZero around in our pockets? Our brains? Will some other AI be the new king of the hill? Will AlphaZero be regarded as quaintly primitive by then? Will Ray Kurzweil's 2029 prediction (and bet, with Mitch Kapor) of a computer passing as human in a Turing test arrive earlier than expected?
And what will humans be like in 2029? Here's a guy working from the other end:
Bryan Johnson isn’t short of ambition. The founder and CEO of neuroscience company Kernel wants “to expand the bounds of human intelligence”. He is planning to do this with neuroprosthetics; brain augmentations that can improve mental function and treat disorders. Put simply, Kernel hopes to place a chip in your brain. . . .
It may sound far-fetched, but Johnson has a track record of getting things done. Within his first semester at university, he’d set up a profitable business selling mobile phones to fellow students. By age 30, he’d founded online payment company Braintree, which he sold six years later to PayPal for $800m. He used $100m of the proceeds to create Kernel in 2016 – it now employs more than 30 people.
But Johnson, 40, says he is about more than money. He was raised as a Mormon in Utah and it was while carrying out two years of missionary work in Ecuador that he was struck by what he describes as an “overwhelming desire to improve the lives of others.”
Are politicians out to “improve the lives of others”? Their report card for the last 120 years tells us they’ve been heaping misery on those they didn’t murder. Today they’re still at it, working anxiously to obliterate the planet in a nuclear firestorm. The political class absolutely, totally flunks the humanity test.
When will that sink in?
The next time you feel nauseated after ingesting the latest political sewage, remember Kernel and DeepMind. Not everyone is corrupt. Not everyone acts like an idiot. If you had to bet on who would take us to a better place, I would recommend putting your money on the researchers and entrepreneurs.