Robots communicating in languages humans can’t understand
The rise of the machines! When two bots of its artificial intelligence software (AI) began to communicate in a language humans could not understand, Facebook researchers put a stop to it.
At first, they were speaking to each other in plain old English. But then researchers realized they’d made a mistake in programming. “There was no reward to sticking to English language,” says Dhruv Batra, visiting research scientist from Georgia Tech at Facebook AI Research (FAIR). As these two [robot] agents competed to get the best deal–a very effective bit of AI vs. AI dogfighting researchers have dubbed a “generative adversarial network”–neither was offered any sort of incentive for speaking as a normal person would. So they began to diverge, eventually rearranging legible words into seemingly nonsensical sentences.
…Facebook ultimately opted to require its negotiation bots to speak in plain old English. “Our interest was having bots who could talk to people,” says Mike Lewis, research scientist at FAIR. Facebook isn’t alone in that perspective. When I inquired to Microsoft about computer-to-computer languages, a spokesperson clarified that Microsoft was more interested in human-to-computer speech. Meanwhile, Google, Amazon, and Apple are all also focusing incredible energies on developing conversational personalities for human consumption. They’re the next wave of user interface, like the mouse and keyboard for the AI era.
The other issue, as Facebook admits, is that it has no way of truly understanding any divergent computer language. “It’s important to remember, there aren’t bilingual speakers of AI and human languages,” says Batra. We already don’t generally understand how complex AIs think because we can’t really see inside their thought process. Adding AI-to-AI conversations to this scenario would only make that problem worse.
The article makes some interesting points about the advantages of allowing this AI software to create its own language. For me, none of these arguments are very convincing.
On Christmas Eve 1968 three Americans became the first humans to visit another world. What they did to celebrate was unexpected and profound, and will be remembered throughout all human history. Genesis: the Story of Apollo 8, Robert Zimmerman's classic history of humanity's first journey to another world, tells that story, and it is now available as both an ebook and an audiobook, both with a foreword by Valerie Anders and a new introduction by Robert Zimmerman.
The ebook is available everywhere for $5.99 (before discount) at amazon, or direct from my ebook publisher, ebookit. If you buy it from ebookit you don't support the big tech companies and the author gets a bigger cut much sooner.
The audiobook is also available at all these vendors, and is also free with a 30-day trial membership to Audible.
"Not simply about one mission, [Genesis] is also the history of America's quest for the moon... Zimmerman has done a masterful job of tying disparate events together into a solid account of one of America's greatest human triumphs."--San Antonio Express-News
The rise of the machines! When two bots of its artificial intelligence software (AI) began to communicate in a language humans could not understand, Facebook researchers put a stop to it.
At first, they were speaking to each other in plain old English. But then researchers realized they’d made a mistake in programming. “There was no reward to sticking to English language,” says Dhruv Batra, visiting research scientist from Georgia Tech at Facebook AI Research (FAIR). As these two [robot] agents competed to get the best deal–a very effective bit of AI vs. AI dogfighting researchers have dubbed a “generative adversarial network”–neither was offered any sort of incentive for speaking as a normal person would. So they began to diverge, eventually rearranging legible words into seemingly nonsensical sentences.
…Facebook ultimately opted to require its negotiation bots to speak in plain old English. “Our interest was having bots who could talk to people,” says Mike Lewis, research scientist at FAIR. Facebook isn’t alone in that perspective. When I inquired to Microsoft about computer-to-computer languages, a spokesperson clarified that Microsoft was more interested in human-to-computer speech. Meanwhile, Google, Amazon, and Apple are all also focusing incredible energies on developing conversational personalities for human consumption. They’re the next wave of user interface, like the mouse and keyboard for the AI era.
The other issue, as Facebook admits, is that it has no way of truly understanding any divergent computer language. “It’s important to remember, there aren’t bilingual speakers of AI and human languages,” says Batra. We already don’t generally understand how complex AIs think because we can’t really see inside their thought process. Adding AI-to-AI conversations to this scenario would only make that problem worse.
The article makes some interesting points about the advantages of allowing this AI software to create its own language. For me, none of these arguments are very convincing.
On Christmas Eve 1968 three Americans became the first humans to visit another world. What they did to celebrate was unexpected and profound, and will be remembered throughout all human history. Genesis: the Story of Apollo 8, Robert Zimmerman's classic history of humanity's first journey to another world, tells that story, and it is now available as both an ebook and an audiobook, both with a foreword by Valerie Anders and a new introduction by Robert Zimmerman.
The ebook is available everywhere for $5.99 (before discount) at amazon, or direct from my ebook publisher, ebookit. If you buy it from ebookit you don't support the big tech companies and the author gets a bigger cut much sooner.
The audiobook is also available at all these vendors, and is also free with a 30-day trial membership to Audible.
"Not simply about one mission, [Genesis] is also the history of America's quest for the moon... Zimmerman has done a masterful job of tying disparate events together into a solid account of one of America's greatest human triumphs."--San Antonio Express-News
Do all these people, anthropomorphize their machines?
Check out this 1970 era movie:
Colossus: The Forbin Project:
Tucked away in a secret location in the Rockies, Dr. Charles Forbin (Eric Braeden) has developed a massive computer system, dubbed “Colossus,” that is supposed to ensure the nation’s safety against nuclear attack. But when Colossus connects to a similar Russian computer, “Guardian,” the intelligent machines begin conducting a private dialog. Nervous as to what they might be plotting, Forbin severs the connection, only to have Colossus threaten a nuclear attack if the link isn’t restored.
Lets see two big computers run by two massively huge companies (governments) what could go wrong…….
I’d love to hear from an actual software engineer on this topic.
From a quick scan of Wikipedia, (which I do at my own peril) ((which is filled with software jargon,)) these type of “generative networks,” contain an inherent “probabilistic distribution” in the underlying code, from at least one of the bot’s involved. One bot generates, one bot evaluates.
(I do appreciate they don’t understand how the system-state evolves over time.)
The headline sounds as-if these “bot’s” obtained some level of consciousness, and I personally just do not buy into that line of thinking. Not my expertise, and the whole Topic (AI and machine-“learning,”) is very interesting, but I’m highly leery of all the mentalistic attributes assigned to machines.
(further afield– the undeniable success of computers has sorta lead to a re-adoption of the “brain as a computer” model, which I believe is incorrect on a fundamental level, although useful on a global basis. That being said, I’m perfectly willing to entertain speculation on quantum physics and uncertainty, as potential substrates for human consciousness, ala Dr. Penrose.)
Ted-
Excellent cultural reference!
The original Skynet
“On Guard: The Story of SAGE”
1956 IBM
https://youtu.be/_qq-SX9KEkw
How do we know the computers even knew what each other were saying? It could all some some weird game theory tit for tat with neither one actually saying anything.
Thanks Wayne!
wodun-
excellent point.
I can’t tell if these researchers are being sloppy with their verbal descriptions and/or the journalist writing it up is describing it sloppily. Way too much anthropomorphizing going on for me.
Ted-
There are copies of the Forbin Project at YouTube. I ended up watching a crummy print only to discover later I had it on DVD in one of those compilation packs.
tangentially– I’m convinced the military-industrial-complex makes the US taxpayers pony up for the same exact research, over-and-over-again. How many times did they charge us for the development of the bat-wing aircraft? More than once, that’s for sure.
pivoting….
“Paper Man”
1971
https://youtu.be/ngFrvOFw7nc
(1:29:15)
“A prank that starts with a group of college students at the computer-lab, creating a fictitious person so they can get a credit card develops into a plot that leaves three of them dead.”
Interesting period piece. Early “paper-tripping” with computers.
This is exactly what happens when humans communicate… just read some text from hundreds of years ago.
“Any sufficiently advanced bug is indistinguishable from a feature.”
Computer Programming Humor
http://www.heuse.com/cphumor.htm
Do all these people, anthropomorphize their machines?
Computers hate when that happens.
eddie-
Good one!