At one time, the AI that beat humans at chess calculated strategies by studying the outcomes of human moves. Then, it turned out, there was a faster way:
In October 2017, the DeepMind team published details of a new Go-playing system, AlphaGo Zero, that studied no human games at all. Instead, it started with the game’s rules and played against itself. The first moves it made were completely random. After each game, it folded in new knowledge of what led to a win and what didn’t. At the end of these scrimmages, AlphaGo Zero went head to head with the already superhuman version of AlphaGo that had beaten Lee Sedol. It won 100 games to zero.Joshua Sokol, “Why Artificial Intelligence Like AlphaZero Has Trouble With the Real World” at Quanta Magazine (February 21, 2018)
But does that mean AI can take over and run everything? The University of Washington’s Pedro Domingos offers a different take: “Games are a very, very unusual thing”:
One characteristic shared by many games, chess and Go included, is that players can see all the pieces on both sides at all times. Each player always has what’s termed “perfect information” about the state of the game. However devilishly complex the game gets, all you need to do is think forward from the current situation.
Plenty of real situations aren’t like that. Imagine asking a computer to diagnose an illness or conduct a business negotiation. “Most real-world strategic interactions involve hidden information,” said Noam Brown, a doctoral student in computer science at Carnegie Mellon University. “I feel like that’s been neglected by the majority of the AI community.”Joshua Sokol, “Why Artificial Intelligence Like AlphaZero Has Trouble With the Real World” at Quanta Magazine (February 21, 2018)
Self-driving cars, for example, have a hard time dealing with bad weather, or cyclists. Or they might not capture bizarre possibilities that turn up in real data, like a bird that happens to fly directly toward the car’s camera. For robot arms, Finn said, initial simulations provide basic physics, allowing the arm to at least learn how to learn. But they fail to capture the details involved in touching surfaces, which means that tasks like screwing on a bottle cap — or conducting an intricate surgical procedure — require real-world experience, too.Joshua Sokol, “Why Artificial Intelligence Like AlphaZero Has Trouble With the Real World” at Quanta Magazine (February 21, 2018)
IBM Watson triumphed at Jeopardy where correct answers were available online but then flopped in the medical environment where “correct” answers don’t even exist.
For example, there is no “correct” way to tell a patient that further efforts against a disease may result only in a less pleasant life, not greater life expectancy. We don’t use machines for that. We just walk with the person for a while.
And, as Jeffrey Funk and Gary Smith remind us, failed prophecies of an AI takeover come at a cost: We don’t improve what we could improve in human-based services like health care if we are waiting for the phantom AI takeover.
You may also wish to read: How to flummox an AI neural network. Kids can figure out the same-different distinction. So can ducklings and bees. But top AI can’t. Can every form of thought be a computation? If not, same-different may be a fundamental limit on computer-based AI.