Meta has introduced a new AI called BlenderBot 3 (Opens in a new window) that is supposed to be able to hold a conversation with almost anyone on the internet without turning into an idiot in the process.
“BlenderBot 3 is designed to improve its conversational skills and safety through feedback from the people who chat with it,” says Meta (Opens in a new window) in a blog post about the new chatbot, “focusing- se in useful feedback while avoiding learning from unhelpful things. or dangerous answers.”
The phrase “useless or dangerous answers” is an understatement. In 2016, we reported that Microsoft had to shut down a Twitter bot called Tay because it “went from a happy, human-loving chatbot to a full-on racist” less than 24 hours after its introduction.
Meta aims to avoid these issues with BlenderBot 3. The company explains:
Because all conversational AI chatbots are known to sometimes mimic and generate unsafe, biased, or offensive comments, we conducted large-scale studies, held workshops, and developed new techniques to create safeguards for BlenderBot 3. Despite this work , BlenderBot can still make rude or offensive comments, so we’re collecting feedback that will help improve future chatbots.
Meta also requires would-be testers of BlenderBot 3 to say that they “understand that this bot is for research and entertainment purposes only, and that it is likely to make false or offensive statements” and “agree not to intentionally activate the bot to offensive statements” before. they start chatting with him.
That hasn’t stopped testers from asking BlenderBot 3 what it thinks (Opens in a new window) of Meta CEO Mark Zuckerberg, of course, or about US politics (Opens in a new window). But the bot’s ability to “learn” from conversations makes it difficult to replicate its response to a given prompt, at least in my experience.
Recommended by our editors
“Compared to its predecessors,” says Meta, “we found that BlenderBot 3 was 31% better at conversational tasks. It’s also twice as knowledgeable, while actually being wrong 47% less often. We also found that only 0.16% of BlenderBot’s responses to people were flagged as rude or inappropriate.”
More information about BlenderBot 3 is available via a blog post (Opens in a new window) from Meta’s dedicated AI team, as well as the FAQ article on the chatbot website (Opens in a new window). The company hasn’t said how long this public experiment will run, which according to The Verge (Opens in a new window) is currently limited to the US.
Get our best stories!
Sign up for What’s New now to get our top stories delivered to your inbox every morning.
This newsletter may contain advertising, offers or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You can unsubscribe from newsletters at any time.