Experiments conducted by the artificial intelligence research group OpenAI have revealed that when made to cooperate with each other on an appointed task, robots are increasing their efficiency in solving it; through refining the language they communicate each other with. The robots learned to collaborate and communicate through trial and error, taking note of signals that helped achieve the goal and storing that information in a private recurrent neural network.
“We’ve just released initial results in which we teach AI agents to create language by dropping them into a set of simple worlds, giving them the ability to communicate, and then giving them goals that can be best achieved by communicating with other agents,” according to a blogpost from the OpenAI team. Adding “If one agent realizes that it could have performed a task better if a second agent had sent different information, the first agent can tell the second exactly how to modify its messages to make them as useful as possible,”
“In other words, agents ask the question: ‘how should I modify my communication output to get the most communal reward in the future?'”
As the tasks in the experiment got tougher, the language the robots developed also got increasingly complex accordingly, although not always as predictable to the researchers. Even though the experimental setup influences how the language evolves, the artificial intelligence would not always use the compositional traits the researchers wanted, often times preferring to use single words to encode entire sentences even when given increased parameters so that the robots could use larger vocabularies. The problem this presents is that it runs counter to the researcher’s goals of shaping an artificial intelligence language that is interpretable to humans.
“We hope that this research into growing a language will let us develop machines that have their own language tied to their own lived experience. We think that if we slowly increase the complexity of their environment, and the range of actions the agents themselves are allowed to take, it’s possible they’ll create an expressive language which contains concepts beyond the basic verbs and nouns that evolved here.”
