contador web Saltar al contenido

Blender: The new Facebook chatbot has emotional intelligence and even shows empathy

While it is possible to speak to an intelligent personal assistant or chatbot, conversations are far from appearing "normal". Facebook's Artificial Intelligence research department has been trying to solve the problem for some time now and is introducing Blender, an open-source chatbot that promises to make conversations between humans and machines a little more natural.

According to Facebook AI Research, Blender is the fruit of several years of research in the field of natural language processing technologies. The company claims that Blender is the first chatbot to be able to present a wide range of competencies at the conversational level, being able to talk about any topic, assume a certain personality and even demonstrate empathy.

The AI ??department of the company led by Mark Zuckerberg also claims that Blender is the largest open-source chatbot created to date. Blender has been trained using large-scale neuronal models with about 9.4 billion parameters. Facebook also trained him with about 1.5 billion examples of dialogue lines from Reddit discussion groups.

The experts in charge of the development of Blender claim that the chatbot has enough emotional intelligence to, for example, congratulate you when the user tells you that you were promoted on the job, or to give the psames when they tell you that someone died. Blender is able to go further in the information it makes available to users, unlike some smart personal assistants who are programmed to read the first few lines of a Wikipedia article.

A group of users after Blender was proof and, according to Facebook, most preferred it over other open-source chatbots. About 67% of respondents said that Blender sounds more human and 75% revealed that they preferred to chat longer with Facebook chatbot than with others.

Facebook AI Research experts indicate that Blender still has a long way to go. Now, the team of experts is making important adjustments to the chatbot, such as helping you more easily detect offensive language or dealing with racist or sexist content.