The artificial intelligence debate rages on, with the Facebook creator Mark Zuckerberg and "Tesla" CEO Elon Musk taking the opposite sides of the debate. Zuckerberg was the one expressing a positive view on such development, with the Facebook's Artificial Intelligence Research Division (FAIR), coming to the forefront of the AI research, particularly with the recent development of negotiating skills in chatbots. But only a few days after the story of the new development in FAIR's research became widely publicized, a new twist in the story became public - Facebook shut down part of this program!
Do you speak "AI"?
The reason given for the shutdown was obviously a surprised the researchers had themselves. The AI chatbots started developing their own language that was not understandable to humans. According to Dhruv Batra, from Georgia tech who is currently a visiting researcher at FAIR, the chatbots came to the conclusion that there is no reward in using the English language - they came up to what seems to be a more rational manner of communication. As Batra puts it, they started inventing "code words for themselves". In the end, Facebook researchers decided to require chatbots to exclusively use the English language.
In a separate development, Google came up with similar results. In an attempt to improve its translation services it added a neural network, making the system more efficient.
It turned out that the system began to translate even between language pairs that it had not been explicitly taught. Google researchers attributed this to the fact they discovered that AI has silently written its own language with the purpose of specifically translating sentences.
More questions being asked than answered
These recent events, particularly the Facebook 'incident' pose more questions than they answer.
On one hand, the proponents of efficient AI chatbot and other communication are of the opinion that the development of 'own AI languages' makes them more efficient and that there is no evidence that such languages will pose a threat.
On the other hand, the opponents point out that the widespread of such development of separate languages could develop a threat to the development of neural networks.
And while such solutions might seem efficient, they make the AI development more difficult. This due to the fact that researchers are currently of the opinion that the humans simply cannot comprehend the logical nature of such communication. And then there are the usual fears of the machines taking over, something that elsewhere has also been called "the Terminator complex". In any case, the debate on the development and use of AI is yet to flare up.
