Facebook says people prefer its sweary, lying chatbot to Google’s

Facebook has created a new chatbot, which was trained to converse using 1.5 billion examples of open source human interactions from Reddit. Yep, we know what you’re thinking…

Facebook’s chatbot, named Blender, was seemingly put together to compete with Google’s offering, Meena. According to a Facebook survey, Facebook has done a good job, as 67% of respondents apparently thought Blender sounded more human than Meena (via BBC).

Remarkably, (and perhaps alarmingly,) 49% of survey respondents also said they preferred talking to Blender the chatbot than to another human being.

However, in these early stages, it has also been reported that Blender sometimes tells lies or respond with inappropriate language – which isn’t wholly surprising given that its base vocabulary and language knowledge has been put together using threads from Reddit.

Related: Best Star Wars games 2020

AI expert Dave Coplin called Facebook’s chatbot a “step in the right direction,” but noted there are still several hurdles between Facebook and the perfect chatbot.

“The first is just how complex it is to replicate all of the nuances of a human attribute,” he told the BBC, “like the ability to hold a conversation, a skill that most three-year-olds can master… The second is around the relationship with the data used to train the model and the results generated by the model.

“As great a platform as Reddit is, training algorithms based on the conversations you find there is going to get you a lot of chaff amongst the wheat.”

Take a look at an example conversation with Blender below.

Related: Best laptop 2020

Blender is capable of conversing on a wide range of subjects including movies, occupations and veganism. The chatbot asks questions of the user too and is designed to seem empathetic and knowledgeable.