Panic because chatbot talks to each other in their own language? This is the truth

On Monday, the excitement of the tech world culminating when Facebook spoke up would leave chatbot after they created their own language.

Some newspapers have called these programs 'scary'. Some journalists also hinted that Facebook had exhausted the idea before some super smart AI spoke up. The British Sun asked to know 'Is the machine taking over?'. The Australian Seven News TV channel also calls this an "emergency with AI". And the answer is: no.

These are merely two chatbot exchanging information with each other. Looking at the snippet code in their 'conversation', it is hard to see what language they are talking about, not to say that it develops the language itself.

The conversation exchange information of chatbot

The press exploded and feared AI after this incident could have originated from Facebook's project. Researchers of this social networking giant tried to teach two chatbot to use conversations to talk to each other. The goal is to teach bots how to plan and communicate more effectively.

When they say nonsense, no AI collapses or kills. Instead, we find software bugs and fix bugs so that bots can talk more like humans, researchers can produce their research results.

Zachary Lipton, an assistant professor of Machine Learning at Carnegie Mellon University, told The Register : 'This job is very interesting. But these are static models, like what Google used to play your board game or phone to predict what word you would write. They have no ability to perceive higher than a bowl of noodles or a pair of shoes. '

Can Chatbot really talk to each other in a new language?

Bot can talk not new or mysterious phenomenon. Researchers at OpenAI discovered that they could talk through something like Morse code when forced to communicate and work together.

Ryan Lowe, a PhD student at McGill University and a graduate student at OpenAI, said this was 'a very normal phenomenon' and 'nothing to worry about'.

'Whenever there is a lot of bot environment, it will be more effective when they talk to each other', he said, 'Their mission is very limited and the language they use reflects that, it is very simple and not ever like real language. Language has a certain syntactic and grammatical system. If it is not taught clearly, what it creates cannot be like language. '

I turned the elephant

Rumors and exaggerated information about AI appear a lot. Switching to AI and saying that your product uses Machine Learning can bring a lot of money. This is not only a problem for sensational journalists, technology giants are also very allergic and often use words like 'fantasy', 'intuition', 'reasoning' in blog posts. their.

'Others are responsible for this is the research labs of large companies, eager to implement groundbreaking projects that many scientists think cannot be realized for years', Lipton said.

Facebook recently posted AI editor positions to help 'develop and implement strategies and campaigns focused on their AI improvements'. Companies like Amazon, Microsoft, OpenAI, Google Brain and DeepMind have all written articles outlining the latest AI development steps.

'The pens can be smart and have good intentions,' Lipton said, 'but there is a problem here. What they are doing despite the feeling is journalism but not, it is not trying to bring about objectivity as well as without a doubtful look and analysis'.