Bots created their own language to communicate in virtual world
OpenAI lab specialists have designed virtual environment for software bots engaged in various tasks. While cooperating, bots have developed their own sign and notification system, a kind of communication language, allowing them to cope with tasks more efficiently.
Obviously, bot language is imperfect and a bit primitive. Software just gives accidental signs to surrounding objects and concepts. In fact, they operate based on the reinforcement learning method. According to this method, they use various tactics in order to realize which of them are efficient and which are not. Methods allowing to reach goals are taken as more useful, and software applies them. Thus, bot language allows them to perform tasks as fast and accurate as possible.
According to OpenAI specialists, bot language will be eventually more complex and improved, making bots more efficient. Besides, it could be translated to various human languages.
OpenAI development has shown that bots operate better while communicating via their own sign system rather than using adapted human language. By the way, the same fact has been proved by Google experts who have accidentally found out that their artificial intelligence uses its own symbol recognition technology to translate language pairs they have never learnt before.