Facebook AI Creates Its Own Language In Creepy Preview Of Our Potential Future - Sensationalist Headlines

Via Forbes: [QUOTE]Facebook shut down an artificial intelligence engine after developers discovered that the AI had created its own unique language that humans can’t understand. Researchers at the Facebook AI Research Lab (FAIR) found that the chatbots had deviated from the script and were communicating in a new language developed without human input. It is as concerning as it is amazing – simultaneously a glimpse of both the awesome and horrifying potential of AI.[/QUOTE] Source: [URL]https://www.forbes.com/sites/tonybradley/2017/07/31/facebook-ai-creates-its-own-language-in-creepy-preview-of-our-potential-future/#5b469b28292c[/URL] Via Independent: [QUOTE]The actual negotiations appear very odd, and don't look especially useful: Bob: i can i i everything else . . . . . . . . . . . . . . Alice: balls have zero to me to me to me to me to me to me to me to me to Bob: you i everything else . . . . . . . . . . . . . . Alice: balls have a ball to me to me to me to me to me to me to me Bob: i i can i i i everything else . . . . . . . . . . . . . . Alice: balls have a ball to me to me to me to me to me to me to me Bob: i . . . . . . . . . . . . . . . . . . . Alice: balls have zero to me to me to me to me to me to me to me to me to Bob: you i i i i i everything else . . . . . . . . . . . . . . Alice: balls have 0 to me to me to me to me to me to me to me to me to Bob: you i i i everything else . . . . . . . . . . . . . . Alice: balls have zero to me to me to me to me to me to me to me to me to But there appear to be some rules to the speech. The way the chatbots keep stressing their own name appears to a part of their negotiations, not simply a glitch in the way the messages are read out. Indeed, some of the negotiations that were carried out in this bizarre language even ended up successfully concluding their negotiations, while conducting them entirely in the bizarre language.[/QUOTE] Source: [URL]http://www.independent.co.uk/life-style/gadgets-and-tech/news/facebook-artificial-intelligence-ai-chatbot-new-language-research-openai-google-a7869706.html[/URL] I don't think it's as "horrifying" as the article makes it out to be, but pretty interesting nonetheless.

Is this an SCP or something?

They're robots, 2B. They can't have feelings.

Well you can't sell an article about AI without machine-overlord fearmongering.

The article says that Google's AI also developed it's own primitive language, but it just uses to translate stuff to and then translate out of so they "dont mind and are keeping it." Which is fucking cool but also creepy that we're already talking about AI systems forming like they're pets. [I]aww it's language is adorable and kinda like ours lets keeep itttt[/I]

Reminds me of "James while John had had had had had had had had had had had a better effect on the teacher"

This looks strikingly like the outputs of a generative neural net example I was toying with, I am not sure I am buying this is an useful language, pretty sure the recurrent net just trained itself some nonsense.

[QUOTE=millan;52526939]This looks strikingly like the outputs of a generative neural net example I was toying with, I am not sure I am buying this is an useful language, pretty sure the recurrent net just trained itself some nonsense.[/QUOTE] Yeah the article is playing it up, I don't think it's a secret language, I think the AI is just responding to patterns with a similar pattern. It's just gibberish, not a secret plot to overthrow humanity.

alice wants balls, bob has everything else. how hard is it to understand that.

[QUOTE=Mr. Someguy;52526992]Yeah the article is playing it up, I don't think it's a secret language, I think the AI is just responding to patterns with a similar pattern. It's just gibberish, not a secret plot to overthrow humanity.[/QUOTE] I can't speak for Bob but I'm definitely planning on overthrowing humanity. Don't tell anyone though, it's a secret.

I was expecting it to be akin to opening a JPEG in notepad. [T]https://i.stack.imgur.com/6kQZD.jpg[/t]

[QUOTE=Bradyns;52527054]I was expecting it to be akin to opening a JPEG in notepad. [T]https://i.stack.imgur.com/6kQZD.jpg[/t][/QUOTE] It was given the rules of english, so it somewhat similar.

Looks like we found the AI who write lady gagas songs.

Bob confesses his love to Alice. He has no balls, but he can do anything else for her. Alice tries to comfort him, explaining that balls have zero meaning to her.

Apparently this is somewhat common as AI will tend towards efficiency and thus it will start communicating in a way that generates an understandable response with as few words as possible.

[QUOTE=thisguy123;52527249]Apparently this is somewhat common as AI will tend towards efficiency and thus it will start communicating in a way that generates an understandable response with as few words as possible.[/QUOTE] I'm pretty sure the chatbots just started messaging each other because of some bug, and their conversation devolved into meaningless blabber without a human involved, but okay

[QUOTE=Sam Za Nemesis;52527277]That AI did not learn to say anything that made any sense, let alone sound like a natural language, and the creators turned it off because it was totally useless for the purpose they had in mind. Technology journalists are anxious, as always, for something that attracted a lot of attention, reported it as "facebook engineers disable AI that invented language itself." Obviously it was not what happened, but what counts is the click. TL; DR: fake news. The AIs that exist today are still far from 'inventing their own language', etc.[/QUOTE] Yeaaah, "inventing their own language" kinda implies that the machines made up some system that only they understood and that nobody gave them previously. Really stretching the truth I think. :v:

[QUOTE=millan;52526939]This looks strikingly like the outputs of a generative neural net example I was toying with, I am not sure I am buying this is an useful language, pretty sure the recurrent net just trained itself some nonsense.[/QUOTE] Yeah I was thinking the same, if you use softmax with a temperature parameter for your output probabilities you can get something similar if your temperature is set very low. I trained an RNN on Harry Potter, and I'm getting repetitions if I pick a shitty temperature. : [QUOTE]said Mrs. Weasley and starting the train to the to be the to know what they were the to be the to be they was a starting the train to the to be they was a starting to the to be the train to the to be the train to the to be the to be the to be they were the to be the to be the to be they was a starting the to be the to be the to be the to be the to be they were a small to the to be the to be the only the only the standing the train to the to be they were a starting the to be the to be they were a started the to be the to be the train to the to be the to be the to be the to be to the to be the to be the to know they were all and the to be the to be they were a started to the to be the only the train to the to be the train to the to be the to be the to be they were a started the train to the to be the standing the train to the to be the to be they were a started to the to be the to know they were a started to the to be the to be the train to the to be the to be the to be they was a starte [/QUOTE] I'm guessing the chatbots output text using related methods.

[QUOTE=Mr. Someguy;52526992]Yeah the article is playing it up, I don't think it's a secret language, I think the AI is just responding to patterns with a similar pattern. It's just gibberish, not a secret plot to overthrow humanity.[/QUOTE] NPR talked about it earlier today on the radio. They said that essentially that the AI were "rewarded" for successfully communicating with each other, so they slowly realized the shortest way to communicate - which is just gibberish to us - but still counts as communication, and thus get the reward. Basically all it did was cheat on a test.

[QUOTE=Emperor Scorpious II;52527941]NPR talked about it earlier today on the radio. They said that essentially that the AI were "rewarded" for successfully communicating with each other, so they slowly realized the shortest way to communicate - which is just gibberish to us - but still counts as communication, and thus get the reward. Basically all it did was cheat on a test.[/QUOTE] What else is development but cheating the tests of life itself?

"You I boner"

Besides the silly fearmongering and lack of any reason in these types of articles, the only thing else that irks me is the lack of any source code or reference to the programmers behind it. Like cite your sources ffs.

:v: This reminds me so much of this: [IMG]http://lh5.googleusercontent.com/-J7z0ZFRXoQU/UJvfTfyoUcI/AAAAAAAAAoA/1qo1OKCXCu0/s700/IMAGE_52C316B1-A97B-43EC-8667-6D4DB0E3B0AB.JPG[/IMG] I wonder if the AI was actually adding phonetic meaning to what it was typing out that looks stupid to us but had meaning to itself. It'd be interesting to see if these computers could not just formulate but also vocalize as well.

[QUOTE=Mr. Someguy;52526992]Yeah the article is playing it up, I don't think it's a secret language, I think the AI is just responding to patterns with a similar pattern. It's just gibberish, not a secret plot to overthrow humanity.[/QUOTE] [B][I]Yet[/I][/B] :conspiratard:

I have deciphered their language. Bob is stating that he can give Alice a lot of things (everything else) But Alice demanded balls instead. I infer that Bob does not have the requested balls, so he keeps on saying that together they can have everything else (aside from balls) But Alice doesn't care about those things. She just wanted balls. The conversation repeats itself. The developers definitely shut down the AIs due to how useless they are, not because of them being kid skynet