Meta’s latest AI chatbot is now available online
Meta’s AI research laboratories have create a new high-tech chatbot and are allowing public members to interact with it to gather input on its capabilities.
BlenderBot 3 is a web-based bot that may be accessed. According to Meta, BlenderBot 3 can engage in ordinary conversation and answer questions. In reality, it can answer things like “discussing healthy food recipes to discovering child-friendly resources in the city.”
The bot is a prototype based on Meta’s previous work with large language models, or LLMS – powerful but problematic text-generation software, the most well-known example being OpenAI’s GPT-3. BlenderBot, like all LLMs, is trained on massive datasets of text, which it mines for statistical patterns to synthesize language. Such programs have shown to be incredibly adaptable, with applications ranging from producing code for programmers to helping writers in publish their next best-seller. These models, however, have severe flaws: they regurgitate biases in their training data and frequently create responses to user questions.
Meta is focusing on testing this latter issue with BlenderBot. A key element of the chatbot is its ability to search the internet for specific topics to discuss. To put it another. way, BlenderBot 3 may cite its sources.
Feedback for Meta
Meta hopes to collect feedback on the different issues that large language models face by making the chatbot available to a broader audience. Users that interact with BlenderBot can flag any questionable replies from the system. Meta claims it has worked hard to “minimize the bots’ use of foul language, insults, and culturally insensitive comments.” research community.
Historically, releasing prototype AI chatbots to the public has been risky for IT businesses. In fact, Microsoft released Tay, a Twitter chatbot that learned from its interactions with the public, in 2016. Twitter users quickly tutored Tay into repeating racist, anti-Semitic, and misogynistic remarks.
Thus, the ultimate prize is creating a system that can have free-ranging and natural talks like a human. Meta claims that the only way to do so is to allow bots to have free-ranging and natural dialogues.