According to The Verge, the bot is called BlenderBot 3 and can be viewed on the web. However, at the moment it seems that only residents of the United States can do this.
“We are committed to releasing all the data we collect in the demo in hopes that we can improve the conversational AI,” Kurt Shustera research engineer at Meta who helped create BlenderBot 3, reportedly said.
BlenderBot 3 is capable of engaging in general discussions, Meta said, but also answering the kind of questions you might ask a digital assistant, “from talking about healthy food recipes to finding suitable equipment to children in the city,” the report said.
The bot is a prototype and builds on Meta’s earlier work with what are known as large language models or LLMS – powerful but flawed text generation software of which OpenAI’s GPT-3 is the best known example.
Like all LLMs, BlenderBot is initially trained on large textual datasets, which it mines in search of statistical patterns to generate language.
Such systems have proven to be extremely flexible and have been used for a range of uses, from generating code for programmers to helping authors write their next bestseller.
However, these models also have serious flaws: they regurgitate biases in their training data and often invent answers to users’ questions (big deal if they are to be useful as digital assistants).