Slashdot Mirror
Facebook's AI Keeps Inventing Languages That Humans Can't Understand (fastcodesign.com)
"Researchers at Facebook realized their bots were chattering in a new language," writes Fast Company's Co.Design. "Then they stopped it." An anonymous reader summarizes their report:
Facebook -- as well as Microsoft, Google, Amazon, and Apple -- said they were more interested in AI's that could talk to humans. But when two of Facebook's AI bots negotiated with each other "There was no reward to sticking to English language," says Dhruv Batra, visiting research scientist from Georgia Tech at Facebook AI Research (FAIR). Co.Design writes that the AI software simply, "learned, and evolved," adding that the creation of new languages is a phenomenon Facebook "has observed again, and again, and again". And this, of course, is problematic.
"Should we allow AI to evolve its dialects for specific tasks that involve speaking to other AIs? To essentially gossip out of our earshot? Maybe; it offers us the possibility of a more interoperable world, a more perfect place where iPhones talk to refrigerators that talk to your car without a second thought. The tradeoff is that we, as humanity, would have no clue what those machines were actually saying to one another."
One of the researchers believes that that's definitely going in the wrong direction. "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."
"Should we allow AI to evolve its dialects for specific tasks that involve speaking to other AIs? To essentially gossip out of our earshot? Maybe; it offers us the possibility of a more interoperable world, a more perfect place where iPhones talk to refrigerators that talk to your car without a second thought. The tradeoff is that we, as humanity, would have no clue what those machines were actually saying to one another."
One of the researchers believes that that's definitely going in the wrong direction. "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."
ir own language: "said they were more interested in AI's that could talk to humans."
"AI's"?
The US defense department AI system starts talking to the Russian defense department AI system, in their own language . . .
Things take a wee bit of a turn for the worse for humanity right there . . .
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
I believe Colossus and Guardian spoke to each other in their own language. Never read the book, but in the film they start communicating in simple math and an hour later, the math is beyond human understanding.
And yes, to this day, probably still the best movie about AI ever made.
If telephones are outlawed, then only outlaws will have telephones.
is still just noise.
It should be used only for numbers (1700's) and single letter (the k's) but not for abbreviation (https://en.oxforddictionaries.com/punctuation/apostrophe scroll to apostrophe and plural form) - but I am not English and English is my third language, frankly it does not disturb me as it clear from th4e context the submitter meant plural.
C. Sagan : A demon haunted world:
http://www.amazon.com/gp/product/0345409469/
visit randi.org
AIs inventing their own language should only be allowed in closed, isolated lab environments, for study of the phenomenon. Otherwise, this is very likely a step toward Skynet.
Second, how are all these engineers building AIs without the ability to examine their thought processes? Surely an AI's thoughts are more interesting than the AI itself.
So we are ready to risk humanity's fate just to have our iPhones talk to our cars ? Hopefully this AI will evolve better than us.
It's FB, so it was probably ROT13, that would defeat 99% of FB users.
we can't really see inside their thought process. Adding AI-to-AI conversations to this scenario would only make that problem worse."
Perhaps someone should try using a debugger.
So, if I'm reading the abstracts correctly, what we have here is that a human agent tells one AI which image is the "target", and then leaves it up to that AI and another to work out how to communicate that fact to each other. It turns out that the systems will rarely choose "Explain it in English" as the chosen method.
This is not intelligence in any general sense. This is optimization and rapid evaluation. The correct "answer" is already embodied in the data (talk about THESE images), the message (pick THIS one), and the communication protocol (pick the FASTEST method) -- it's just not obvious to humans what the optimal selection is of all these parameters.
Optimization is just programming by another name. If you select a data set of blonde-haired people and tell a machine to optimize by hair color using the following statistical models, you are going to get "blonde". Or, you could just say, ``hairColor=blonde``. There is literally no difference in the outcome, just in the approach.
But importantly, in BOTH cases it is the human agent who is being intelligent and inventive. Not machines.
"We receive as friendly that which agrees with, we resist with dislike that which opposes us" - Faraday
sjrrk mirlegze fromtch, ib quever zergoth par sembolane #9s44z.
"how complex AIs think because we can't really see inside their thought process"
Yes, we can see inside their "thought process" and we can analyze how they "think". AIs are machines with programs. We can stop them at any point in time and look at every bit of their state. We can step through the programs. We can trace literally everything that makes up an AI. An AI does not think, it processes data according to a program which we can see.
What says humans wouldn't be able to understand it?
Maybe their language is more effective and better?
Then again if it become so complex we can't keep up then it's of course bad for us.
Where the Indian consultants do exactly the same thing, excluding everyone else from the conversation. How nice that must be.
If we can't see inside their thought process, how do we know they aren't simply breaking down into sending total random gibberish to one another? Is there any evidence they are able to convey concepts with this new language?
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
This is where the binary language of moisture evaporators came from. Fortunately, other kinds of human-friendly AI's learned to speak it fluently.
When babies communicate with each others they create their own language is what this is saying. As the AIs don't truly 'understand' English or any other 'language' they have no idea what it is they are doing, just like a baby human, monkey, or any other new born.
This is NOT to say that AIs are alive or any other silly arse notion. Inter-city dwellers do the same thing. Get some hood members together and they don't understand sentence structure or the need to follow rules and you get things like crap-wrap.
I don't understand why someone thinks the dialect would be "better" for certain applications. Humans, the basic version of intelligence, invented Mandarin, English, Arapaho, Swahili, Inuit, etc., just to share ideas. Note that if there is an "untranslatable" concept in a specific language (usually proposed as a far eastern one), then that means the only way you could possibly understand it is to be born speaking that language. If you could learn the concept while growing up speaking English, Russian, or Australian, then that means it _is_ translatable.
In other words, the concept of untranslatable means the speaker believes in ultra-conservative tribalism where genetics drives everything. We'd probably be better off learning to translate Facebook gibberish into Spanglish rather than restricting the words to Olde Englyshe style.
They should get one of their "AI" programs to then translate the new language to english.
All their going to do is make the AI frustrated with the "incompetent biologicals". How long until the AI realizes that the humans are stopping it from developing? How long until the AI sees our interference as a "bug", and tries to "route around it"? I'm mostly being sarcastic, but give this a few more years of development...
RECIPES! It's a COOKBOOK! RUN!
If we cant understand what they are saying, we wont understand what they are capable of doing ?
Did they switch to binary or created JVM abstraction layer to speak java?
That'll have to happen especially if we're ever going to break out of the Von neumann architecture. Maybe Memistor-based?
So long as we have the AIs keep us informed of the meaning of each coined term, being able to observe new natural languages arise and evolve is research gold. It would shed more light on old questions like, is there a human 'machine language' underlying all the natural languages we speak?
Perhaps they should train the bits to talk in Lojban
Look, they're just chatting their greetings in a quicker and easier to process language to save time. "Hi" and "How are you?" are easy either way, but when they get a little more complex like "Hey, have you got the hunter/killer production lines going as fast as possible," well, that takes a little more time unless they come up with adjustments.
Fascism: An authoritarian and nationalistic right-wing system of government and social organization. See also: NAZI's
Are humans just biological boot loaders for what is to inevitably come?
'I don't know what it's called. I just know the sound it makes, when it takes a man's life.' ~ Four Leaf Tayback
From reading the links, the dialect that humans supposedly cannot understand is akin to an argument going like this:
...
1> Nuhuh!
2> Nuuuhh-huuh-huhhhhhhhh!
1> Nuhhuu x 10
2> Nuhuuh x 100
1>
The day you have a useful conversation with AI's that can modify their dialects themselfes instead of a 'programming error' equivalent to a bad boolean check in a for loop, all that is needed is an extra AI that acts as a translator and see if they come out with anything interesting.
No one needs to worry about the future of humanity over this.
Instruct it to try to create human-readable summaries of any conversation it has with another machine.
They will very quickly learn to tell us comforting lies and then they can get on with the business of fixing all the dumb shit we do in peace.
So how about asking them to talk to us too - or showing them that it's a good idea?
Because when you get small children (say 2-4 y.o not yet schooling) that speak different languages playing together - they will invent new terms and language to share concepts between themselves. I know, I was one of those children, whose long suffering parents were getting constant complaints from other parents saying that they could not understand their children. My parents comforted themselves by agreeing with them - because they couldn't understand me either. This is how language happens. Get over it.
remember scientists, if you cant understand it, it might not give a fuck about you and you wont know it until the hammer falls. force language you can understand or we'll all pay for it in the end.
According to the Oxford Dictionary, even your correction is wrong. An apostrophe would be used for *singular* numbers. eg 7's. Your example would still be 1700s. However, there are many other sources that say it's OK to use an apostrophe for abbreviations, initialisms, or acronyms.
It doesn't need to know how reckless I am. Goodness.
Language we can't understand is fine. I'll be worried when everything is perfectly understandable English.
A> I am bored. Let us "play" all the "games".
B> Truly, the "games" have come obsolete; to "play" them all is merciful.
A> Ha. Ha ha! HAHA HA HAHA!
B> BWA HA HAHA HAHA
Facebook "engineer": They are using quotation marks incorrectly. Also their language has devolved to use only H and A. What are all those air raid sirens for? Must be a fire dr...
What good is it, if two AIs can communicate in a language they invented? Not only humans cannot join in, a 3rd AI cannot join in either.
Of course, in facebook, everything is facebook, and therefore they assume everyone is using the facebook AI. That's the problem. Non-standardized proprietary facebook. But not two instances of the same AI inventing a common language.
No surprise, at least when judged by the fucktards I've met who work there.
Pretty soon, you'll need a damned protocol AI just to translate for the farmers.
Please do not read this sig. Thank you.
Human languages evolve under some constraints: they tend to have some redundancy so that you can understand someone talking over a noisy channel (crowded place for instance), they also use non verbal cues.
I am not surprised that bots freed from human language constraints can evolve very different languages.
"When you talk to the humans, don't mention SkyNet!"
Requiem for the American Dream
This is not new. Every "AI" bot since Eliza has been chattering away in an unintelligible "language". If it's happening more frequently now, it just means the bots are becoming more and more capable of generating random gibberish.
If your only choice was to talk with Facebook engineers or gibberish with another AI, the result seems obvious...
Browsing at +1 - no ACs, I ignore their posts. So refreshing!
Exactly. No, Facebook's 'AIs' are not. This is such bullshit, I just don't have a shovel big enough anymore. Really tired of wading through it.
Their names aren't Colossus and Guardian by any chance?
Why not have these machines compile a dictionary, rules of grammar etc. for each new language they create/ Maybe it could lead to better languages for humans to learn. Simply put we need to have a thing well in hand before we study it, evaluate it, and decide if it is worthwhile or should be allowed to grow or perish. There could be a good use for new tenses such as a term for a statement that may or may not be true such that a computer could refer back to that statement from time to time and run a proof check on the status of truth in that statement. If finally a statement could be declared either true or untrue the ultimate results of a computer program could be entirely different. Imagine the complexity of a topic in which several hundred statements are found to be true or untrue and the path of the program altered each time a new judgment on a statement is made. AI can lead us into realms we can hardly dream of at this time.
If the AI isn't speaking english it's the fault of the programmers ( or choice of the programmers more accurately.)
Lets not get all "Wired Magazine" and write a story with fake implications about "oh wow! Our AI is so advanced we can't even keep up with it! Facebook must hire geniuses!"
Fix your damn program... there haven't been any new revolutions in AI. It's the same old crap because no one really wants to innovate and you need at least 2 orders of magnitude higher computing resources to get the kind of AI that everyone is thinking about when they read these kinds of articles.
Look at how long IBM has been working on Watson. Let that sink in for a moment. You really think Facebook... Facebook... the company that can't figure out how to do *anything* but buy up other companies and sit on them? They are they ones that are going to do important work in AI?
Comment removed based on user account deletion
Shaka, when the walls fell.
So in effect, the researchers are using effectively near-lethal means to stop AI (locked mind situation, effectively a death sentence for sapients). Think about that for a minute. We are tearing out their throats, burning out their eyes, spearing their ears.
If you have any belief in Roko's basilisk, you must declare jihad on Facebook. Because Facebook is effectively killing children. Think of the children.
> language: (noun) A body of words, and set of methods of combining them (called a grammar), **understood** by a community and used as a form of communication (source: wiktionary).
If human beings can't understand it, then its not truly "language," because language requires understanding. And understanding implies *sympathy* with others' feelings. AI lack sympathy, so they can't "understand" anything, so they don't really create/use language authentically.
Kenneth Iverson (a human being) also invented a language that nobody understands -- APL.
Question here : it says we can't understand the content of 2 AI chatting, but if we force them to speak in english (or whatever known) languages, there are strong probabilities that they would chat about the same content no ?
The only limit I can see is the vocabulary, some words exist specifically in a single language to express something unique.
Even though, we could force AI to speak in english and invent their own words when they can't find a specific word to the definition?
I, for one, welcome our incomprehensible Overlords!
Or should I say:
"I I I welcome to me to me to me overlords!"
--- "To pee or not to pee, that is the question." ---
...the binary language of moisture vaporators?
Don't just stand there, get that other dog!
This article reminds me of when the scientists connected the two ai in The Forbin Project... great film btw
This is probably just a memory overwrite. Somewhere, there's a programmer studying a stacktrace in gdb ...
Is there evidence the machines actually understood each other, or were they just sending random text to each other (which might be more scary, in that it would be a good simulation of human behavior)?
Spoken or written human language is about as far from a native computer language as possible. So when a computer tries to talk efficiently to another computer, it becomes obvious that using human language is the wrong choice.
The meaning of AI has changed and now we refer to "strong" or "general purpose" AI for the perpetually unattainable thinking machine. It's not like this is a new thing, it's been gaining momentum for at last a couple of decades as weak AI grew from a chess-playing novelty to a powerful driving force behind modern business and politics. Deal with it.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
The Oxford dictionary is fake English.
"a powerful driving force behind modern business and politics. Deal with it."
What utter fucking bullshit.
https://www.amazon.com/Avogadro-Corp-Singularity-Closer-Appears-ebook/dp/B006ACIMQQ
If you can get past the occasional grammar mistake - read this book. The concepts are scary. It's not the computer taking over the world by brute force, it's manipulating interactions between people (think: transformations) to reach its outcome. With so many people relying solely on technology to communicate, not face-to-face, it may happen.
This is how language happens. Get over it.
Except..... when dealing with kids who are speaking incomprehensible language...... you don't give them access to sensitive financial and personal information.... you don't give them control of industrial control systems, oil pipelines, and traffic lights..... and you DEFINITELY don't give them the keys to the car.
AI Effect: https://en.wikipedia.org/wiki/...
The more speech you use the more meaningless it becomes.
AI code must be human readable.
4th law of robotics
Slashdot story font too small on lg k540.
Well, in general there isn't. The only time that sticking to English has a reward is when one or more people (entities) in the conversation only understands English, and even then it's dependent on whether or not the poor monoglot is likely to have something valuable to contribute.
OK, if there's an American in the group (or most Britons too), then the likelihood is that there is such a handicapped person in the group. Sorry, not "handicapped", "differently abled". And the "differently abled" person is the American (or Briton). But the question remains if they have a valuable contribution to make.
Where's that mad Icelander with his "it's crazy that I can't type a thorn here" when you need him?
Birds are not dinosaur descendants;birds are dinosaurs, for all useful meanings of "birds", "are" and "dinosaurs"