Here’s an interesting story for you. It’s about the role games have played in the development of artificial intelligence (AI), which today is nearly ubiquitous.
In 1954 England, a 41-year-old man was found dead; supposed suicide. Though it was never tested, a partially eaten apple next to his body was suspected of having been laced with cyanide.
Two years earlier, following conviction for “gross indecency” under Britain’s (then) law against homosexuality, the man had been given the choice of prison or chemical castration. He chose castration.
Approximately 12 years previous, that same man, mathematical genius Alan Mathison Turing, was busy developing the “Bombe,” a device for deciphering coded messages transmitted by German Nazis on machines called Enigma. The Enigma code had been broken by Polish mathematicians, but that knowledge required a means to implement it. That was the Bombe’s contribution, and its development consumed the bulk of Turing’s attention; but not all of it.
His “side interests” were computer science and artificial intelligence, both of which Turing is now widely credited as having founded. Through his writing and lectures, he described how computers could learn to think like humans. He explained how code could enable computers to learn from their own experience, then generalize their new knowledge to the solution of novel problems, as do humans.
Turing predicted computers with AI would learn chess. They did. But the first AI program (1951) taught a computer to play checkers, and win. Forty-six years later (1997), IBM’s Deep Blue computer beat reigning world champion Garry Kasparov in a six-game chess match. Thereafter, AI development shifted into high gear. A scant 14 years thence (2011), you may have watched IBM’s Watson beat Ken Jennings and Brad Rutter in a “Jeopardy” playoff.
Particular AI could now outsmart humans at checkers, chess, and “Jeopardy.” It accomplished that by, for example, storing in the computer’s memory, then indexing for use, virtually every move every chess grandmaster ever made. That approach emphasized rote intelligence. That was soon to change.
In 2014, Google acquired the British company DeepMind. DeepMind was co-founded by Demis Hassabis, who had been a child chess prodigy. Hassabis had developed an AI program called AlphaGo, which, in 2017, defeated world champion Ke Jie at a game called Go.
Go, if you’re not familiar, was invented by the Chinese 2,500 years ago. Unlike most games, Go leans heavily on a player’s intuition and instinct. So Hassabis recognized that traditional rote approaches would never defeat champion Go players. What was needed for Hassabis’ AlphaGo was AI with neural networks similar to neurons in the brain.
To equip himself, Hassabis took a Ph.D. in neuroscience. AlphaGo uses two sets of deep neural networks with millions of connections. Says Hassabis: “I think what we’ve done with AlphaGo is introduce with the neural networks this aspect of intuition, if you want to call it that. We’ve imbued AlphaGo with the ability to learn and then it’s learnt it through practice and study, which is much more human-like.” And that is exactly what Alan Turing predicted.
Of course, AI for winning games was never the endgame. Hassabis, e.g., seems about done with game programs but plans to use variations of the technology for things like helping paralyzed people function again. Elon Musk is developing an invasive brain-computer interface. Google is using AI to help folks stay between the ditches with fully autonomous cars. And so on.
Back in 1950, Alan Turing proposed the “Turing Test” to determine whether a machine demonstrates human intelligence, i.e., “true AI.” Basically, a “judge” submits written questions to an intermediary who submits the questions to a computer, which writes its responses; and to a person who writes responses. The computer and person are hidden from view. The judge must determine which written responses came from the human respondent.
If the computer is erroneously selected as the human more than 30 percent of the time within a five-minute interval, the computer passes the test. None passed until June 9, 2014 — possibly — that “success” is now questioned by many computer experts. That might be a good thing.
Before his death in 2018, world-renowned physicist Stephen Hawking foresaw a point where warp speed AI evolution would outstrip humans’ ability to keep pace. “Humans, who are limited by slow biological evolution, couldn’t compete, and would be superseded. The development of full artificial intelligence could spell the end of the human race.”
One thing seems certain. The AI genie that Alan Mathison Turing released in the early 1940s ain’t going back in the bottle.
Belatedly, in Sackville Park, Manchester, England, there is now a statue of Turing that reads, “Father of computer science, mathematician, logician, wartime codebreaker, victim of prejudice.”
Queen Elizabeth II pardoned Turing in 2013 for his “gross indecency” crime.
Michael Smith is a Southern Pines writer.
The post Column: Artificial Intelligence Makes Big Leaps, With Biggest to Come – Southern Pines Pilot appeared first on abangtech.
source https://abangtech.com/column-artificial-intelligence-makes-big-leaps-with-biggest-to-come-southern-pines-pilot/
No comments:
Post a Comment