Facebook A.I programme reveals dangerous human threat
In news that is equal parts fascinating and somewhat terrifying, Facebook recently had to abandon an experiment after two artificially intelligent [A.I] programs began chatting to each other in a modified version of English only they could understand.
The chatbots decided to modify English for efficiency, the only problem being that the humans monitoring the exchange couldn’t understand the strange language. This development was the result of Facebook challenging its chatbots to conduct trade negotiations with each other.
The attempt to swap hats, balls, and books quickly broke down, as the bots appeared to communicate in a language, which they understood but which remained incomprehensible to humans.
According to the researchers, the robots were instructed to work out how to negotiate and improve their bartering skills as they got smarter but they were not told to use comprehensible English. This loophole let them create a “shorthand” form of English.
To the casual observer, this exchange [below] looks bizarre and makes very little sense: However, the robots were able to successfully conclude their negotiations in what appears to gibberish.
“Agents will drift off understandable language and invent code words for themselves,” Facebook’s AI Research visiting researcher Dhruv Batra said. “Like if I say ‘the’ five times, you interpret that to mean I want five copies of this item. This isn’t so different from the way communities of humans create shorthands.”
The bots also learned to negotiate in ways that appear to be human. According to a paper by FAIR, they would pretend to be very interested in a specific item – so that they could later pretend they were losing a something dear by giving it up.
Facebook decided to shut down the experiment because the AI programs were not working as intended. The purpose of the experiment was to build bots, which could speak to people in a language they understood.
This isn’t the first time artificial intelligence has invented new forms of language. Google revealed earlier this year that the AI for its Translate tool had created its own language, which it would then translate things in and out of. In this case, the company was happy with the development and decided to let it continue evolving.
Elon Musk recently told Bloomberg that Google using Artificial Intelligence technology could “produce something evil by accident – like a fleet of artificial intelligence-enhanced robots capable of destroying mankind.”
Musk, who founded Tesla and Space X, said in 2014 that A.I poses an existential threat to human beings on earth.
“With artificial intelligence, we are summoning the demon. You know all those stories where there’s the guy with the pentagram and the holy water and he’s like, yeah, he’s sure he can control the demon? Doesn’t work out.”
He cautioned supposed innovators to think deeply so as to ensure they “don’t do something very foolish.” So are supposedly very brilliant individuals going to continue being foolish?