Is "AI who has arrived at its own language usage" that exposes humanity to a crisis?

In June 2017, Facebook is a chatbot developed by the company that AI invented how to use its own languagePosting a blogWe announced at. As it were, it is a case that a computer's artificial intelligence finally came up with a surprise as if it came up with its own language, and even a threat theory to some people was even a matter of a boil-up, but whether this event was taken as "the crisis of mankind" , "Evolution of technology just" is taken as opinion is divided.

The 'creepy Facebook AI' story that captivated the media - BBC News
http://www.bbc.com/news/technology-40790258

In the chat bot developed by Facebook, an attempt was made to negotiate each other about fictitious possessions by two AIs. The aim was to understand the role of linguistics in such discussion, and that each AI was programmed to do various experiments linguistically in order to dominate the discussion Thing. In the experiment, the development team has revealed that AI has arrived at its own language usage. A few days after the Facebook blog above, Internet media Atlantic actually talked about what kind of conversation was exchangedPublished in articlesdoing.

The revealed "conversation" is as follows. It is a conversation exchanged by two AIs named "Bob" and "Alice", but you can see that it is made up of patterns that are clearly not used by humans at first glance.

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 me
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 me
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 to 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 me

Some people who saw this content also said that "AI created a new language trying to cheat the eyes of the master who is the master". On the other hand, the idea that the neural network of AI has modified the language in order to achieve more efficient interaction is also shown.

It is not the first time that a neural network has produced its own language. In December 2016, AI used for Google translation created unique "intermediate language", and it became possible to translate a certain language even with a combination of languages ​​that should not be trained Has been revealed.

AI of Google translation enters the stage where it can even acquire its own "intermediate language" and even "translate between unlabeled languages" - GIGAZINE


As for the evolution of AI, there are voices calling for a threat theory against overly evolved AI like Dr. Stephen Hawking, astrophysicist and businessman Earl Mask. Meanwhile, Facebook's Zuckerburg CEO has not defeated the positive view, and there was also a minor controversy with Mr. Mask.

Battle Earlon mask counterargs to Mark Zuckerburg saying "AI threat theory is irresponsible" - GIGAZINE


One of the reasons for the AI ​​threat theory is that "AI technology is too complex to understand a path leading to conclusions". In addition, as the technology is still in the growth stage, it is not predictive of future evolution, furthermore learning with innumerable computers enables growth far beyond human capability, making it unmanageable The view that it is likely to become it is also indicated.

However, the BBC says that at least this time Facebook AI is done in the "laboratory" and it is necessary to understand the fact that it does not affect the actual society. Also, it is said that Facebook's chatbot has already been suspended, but the reason is not to "crisis humanity" but because it has arrived at a place not related to the research aimed at the research team .

Nonetheless, there is no real name "AI for generic use" in the true meaning of AI being developed all over the world and it is unpredictable when and where to give an unexpected effect It is realistic that it is far from the level that you should prepare for the worst situation.

ByO'Reilly Conferences