Facebook’s Chatbots Invented Their Own Language

Image credits: Shutterstock

It is no secret that artificial intelligence (A.I.) has a number of interesting applications for the future. But sometimes very curious things emerge out of this new field. While Facebook was using one Chatbot to train other Chatbot something amazing happened. The Chatbots started to develop a language not comprehensible by human and started to communicate in this language.

HOW A.I. WORKS

In order to understand how this was possible, it is important to know some basic concepts. AI is a field of study that tries to emulate human decision process, learning through an algorithm. In general, the idea is to try to identify a pattern in the input data and after recognition, make predictions accordingly. This could mean, analyzing medical data and finding the disease, responding to catastrophes or identifying language patterns in a chat and responding to it.

In terms of Chatbots, the idea is to identify patterns of questions and generate answers thus setting up a dialogue. This is a very popular use in the field of AI and is already been applied by all major players such as Apple (SIRI), Google (NOW), Amazon (Alexa), Microsoft (Cortana) and IBM (Watson). Facebook joined this race by releasing in 2016 its bots that are able to interact with its users.

BUT FACEBOOK’S CHATBOTS ARE DIFFERENT

Unlike others, Facebook was training its bots to develop a negotiating skill. They were very successful in it on it but at some point, while negotiating with another bot, they started to speak in another language. This was possible, because bot 1(say) modified the way of communicating and bot 2 adapted it. In this process of modification and adaptation, the language changed completely.

While this was the most curious aspect of this particular research, it was not the only interesting one. The chatbots developed good negotiating skills such as showing a false interest to something in order to exchange it for a better deal in something that it really wants. Anyway, this research highlights that even though we design the algorithms, we do not fully know how it works.

As A.I. analyzes our pattern of “liking” and “sharing”, it learns what we like and based on that; what will be shown on our timeline. The same data is used to determinate advertising data also. To know exactly how the machine learns and not only its algorithm without any doubt will be a very important field of study in A.I. Knowing this better would give us control over the process and help to achieve even more significant results related to the applications of A.I. in our daily life.

Check out Mark Zuckerberg discussing facebook’s chatbots.

Video credits: IDG.Tv
Source Facebook The Atlantic

You might also like More from author

Comments are closed, but trackbacks and pingbacks are open.