Meta’s AI research labs have created a new state-of-the-art chatbot and make available to public to talk to the system in order to collect feedback on its capabilities.
The bot is called BlenderBot 3 and can be accessed on the web. BlenderBot 3 is able to engage in general chitchat, says Meta, but also answer the sort of queries you might ask a digital assistant, “from talking about healthy food recipes to finding child-friendly amenities in the city.”
The first notable improvement in BlenderBot 3 is its size. The new release features an astonishing 175 billio parameter architecture based on the OPT-175B language model. This represents a 58x size increase compared to BlenderBot 2.
One of the main promises of BlenderBot 3 is its ability to improve itself organically based on the feedback collected in active conversations. The model obviously builds on the capabilities of its two previous released which include ternet search, long-term memory, personality, and empathy. Additionally, BlenderBot 3 inherited over 1000 pretrained skills. Despite those inherited capabilities, building an architecture that improves over time is far from an easy endeavor. BlenderBot 3 achieves this with an architecture called Director. The architecture generates responses using two main mechanisms:
1) Language Modeling: Provides the model with the most fluent and relevant responses.
2) Classification: Informs the model about correct and incorrect answers based on human feedback.
The Director architecture allows BlenderBot 3 to collect human feedback in real conversations while also penalizing toxis, biased or low-quality responses. In that area, the new release also incorporate new learning algorithms that helps to distinguish helpful from harmful responses.
By releasing the chatbot to the general public, Meta wants to collect feedback on the various problems facing large language models. Users who chat with BlenderBot will be able to flag any suspect responses from the system, and Meta says it’s worked hard to “minimize the bots’ use of vulgar language, slurs, and culturally insensitive comments.” Users will have to opt in to have their data collected, and if so, their conversations and feedback will be stored and later published by Meta to be used by the general AI research community.
Releasing prototype AI chatbots to the public has, historically, been a risky move for tech companies. In 2016, Microsoft released a chatbot named Tay on Twitter that learned from its interactions with the public. Somewhat predictably, Twitter’s users soon coached Tay into regurgitating a range of racist, antisemitic, and misogynistic statements. In response, Microsoft pulled the bot offline less than 24 hours later.
