The BlenderBot 3 is capable of general small talk, Meta said, but also answers the kind of questions you might ask a digital assistant, from talking about healthy food recipes to finding kid-friendly amenities in the city. The bot is a prototype that builds on Meta’s previous work, the so-called Large Language Model, or LLMS, which generates software from powerful but flawed text, of which OpenAI’s GPT-3 is the most well-known example.
Like all LLMs, BlenderBot was originally trained on huge text datasets where it mines statistical patterns to generate language. Such systems have proven to be very flexible and have been used for a range of purposes, from generating code for programmers to helping authors write the next best-selling book. However, these models also suffer from serious flaws: they regurgitate bias in the training data and often make up answers to user questions.
This latter question is something Meta specifically wanted to test with BlenderBot. A great feature of this chatbot is its ability to search the internet in order to talk about a specific topic. What’s more, users can click on its answer to see where its information came from. In other words, BlenderBot 3 can cite its source. By releasing the chatbot to the public, Meta hopes to gather feedback on various issues facing large language models.
Users chatting with BlenderBot will be able to flag any suspicious responses from the system, and Meta says it has worked to minimize the bot’s use of foul language, foul language and culturally insensitive comments. Users must opt-in to have their data collected, and if so, their conversations and feedback will be stored and later published by Meta for use by the general AI research community.