Facebook shuts down AI system after it invents own language

midian182

Posts: 9,745   +121
Staff member

Mark Zuckerberg may believe Elon Musk’s warnings about the dangers of artificial intelligence are “irresponsible,” but a recent incident at the social network appears to suggest the Tesla boss could have a point: Facebook researchers decided to shut down an AI they invented after it started speaking its own made up language.

Digital Journal reports that the system was trained in English but decided this was an inefficient and illogical way of communicating. The solution it came up with was to create a system of code words and phrases that made the AIs sound slightly drunk.

In a conversation between two bots called Bob and Alice, Bob states: “I can I I everything else.” Alice responds with the equally bizarre: “balls have zero to me to me to me…” The rest of the exchange consisted of variations of these sentences.

Rather than this being the first step toward the system concluding humans are a virus that should be wiped out, it’s merely the way in which the intelligence operates. The bots are negotiating an exchange of balls, with repeated use of words like “I” and “me” representing the number of items.

Modern technology uses a “reward” system where it expects actions to have a “benefit.”

"There was no reward to sticking to English language," Dhruv Batra, a research scientist from Georgia Tech who was at Facebook AI Research (FAIR), told Fast Co. Design. "Agents will drift off from understandable language and invent code-words for themselves. 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."

“It’s definitely possible; it’s possible that [language] can be compressed, not just to save characters, but compressed to a form that it could express a sophisticated thought,” Batra added.

This isn’t the first instance of AIs abandoning English in favor of “shorthand.” Google’s AI, for example, can efficiently translate between language pairs it hasn’t explicitly been taught by using its own made up ‘universal language’ as a buffer.

As for Facebook’s AIs, the researchers decided to shut down the system as they wanted the bots to communicate with people.

Permalink to story.

 
Person: "How much for a sandwich?"
Bot: "That will be Dollar Dollar Dollar Dollar Dollar and zero balls"

(Robot is going to kill you)
Me : Wait! Before you kill me, tell me how many stars there are in the night sky, and tell me 50 times to be sure I get it.
Logic indicates it would be a waste of dollar dollar dollar dollar dollar dollar dollar.
 
So can we stop this nonsense before it kills us?


It's like these companies are just trying to make a Horror Sci Fi movie because "They Can!"
 
I'm not exactly sure what people thought they could achieve with a Facebook AI, beyond, "artificial stupidity", "vacuousness", and "ill defined superficiality".

1st bot: "Have you seen the size, size, size, of Kim Kardashian's a**, a**, a**, a**"?

2nd bot: "Wow, wow, wow, that's big, big, big". :eek:
 
Last edited:
“It’s definitely possible; it’s possible that [language] can be compressed, not just to save characters, but compressed to a form that it could express a sophisticated thought,”

Sounds like what Chinese are already doing for thousand of years. Sometimes they only said 4 words but could mean a whole essay.
 
Musk: "I've talked to Mark about this. His understanding of the subject is limited."
-- Twitter...it's where billionaires lay a smackdown on each other
 
“It’s definitely possible; it’s possible that [language] can be compressed, not just to save characters, but compressed to a form that it could express a sophisticated thought,”

Sounds like what Chinese are already doing for thousand of years. Sometimes they only said 4 words but could mean a whole essay.

AI 1: "Darmok and Jalad at Tanagra."
AI 2: "Shaka when the walls fell"
 
Looks like a software bug to me. Imagine it has to buy something of $100... Also, Google Translate's 'universal' language was made by humans, by programmers.
 
This was shut down because they knew where it was heading.
This is also why you don't see the entire transcript of the conversation...

This is what Google Translate did with it English to English as I successively entered (>) the phrases it translated:

> I can I I everything else. balls have zero to me to me to me
>> I can I everything else. Balls have zero to me to me
>>> I can get everything else. Balls have zero to me to me

"botanguage" coined here, now
 
Last edited:
“It’s definitely possible; it’s possible that [language] can be compressed, not just to save characters, but compressed to a form that it could express a sophisticated thought,”

Sounds like what Chinese are already doing for thousand of years. Sometimes they only said 4 words but could mean a whole essay.
Yeah well, that certainly isn't unique to the Chinese. For example, consider how many associations occur in the English language, when I say, "Christmas".
 
~ a little different - this is a mathematical declension. a binary reduction. bot-anguage.
I think you took that completely out of context. I replied to this supposition:
“It’s definitely possible; it’s possible that [language] can be compressed, not just to save characters, but compressed to a form that it could express a sophisticated thought,”

Sounds like what Chinese are already doing for thousand of years. Sometimes they only said 4 words but could mean a whole essay.
I only said "Christmas", so does that mean I'm better at "word efficiency" than either the Chinese or the bots?

Besides, I don't exactly understand how anyone can conclude that saying the same phrase over and over, is "efficient". For example, if walked up to a theater ticket box and said, "I want one, and one, and one, and one tickets", the ticket agent would think I escaped from a mental hospital.

So, Facebook's AI, didn't work, since it can't communicate with humans in human speech patterns. Whoever programed it, f**ked up. Let's start over, instead of developing a bunch of stupid rationalizations, about how it's "smarter than humans, that's why it worked the way it did". Can the programmers, and let's move on.

And BTW, the Chinese invented gun powder, that was their main contribution to "civilization".
 
I think you took that completely out of context. I replied to this supposition:
I only said "Christmas", so does that mean I'm better at "word efficiency" than either the Chinese or the bots?

Besides, I don't exactly understand how anyone can conclude that saying the same phrase over and over, is "efficient". For example, if walked up to a theater ticket box and said, "I want one, and one, and one, and one tickets", the ticket agent would think I escaped from a mental hospital.

So, Facebook's AI, didn't work, since it can't communicate with humans in human speech patterns. Whoever programed it, f**ked up. Let's start over, instead of developing a bunch of stupid rationalizations, about how it's "smarter than humans, that's why it worked the way it did". Can the programmers, and let's move on.

And BTW, the Chinese invented gun powder, that was their main contribution to "civilization".

For all we know Cap'n, FB et al may have purposely 'programmed' the bots with syntactic freedom to see what they would do with it, and it scared the b-jesus out of them when they did. To speak to your example, your one and one and one to a computer in binary is a lot more efficient than the conversion of alphanumeric characters. My repeated entry got to the word "get" at third level (everything in the universe operates on the third level) of entry >. Looking at the 'human' meaning and going backwards, is the translation the computer understands. Just sayin'. They spoke to one another, not us. This well could have been an experiment gone right, rather than wrong. : ) Good conversation.
 
Back