Site icon Enrich of Tech Updates Across the World

Meta is asking the public for help with its chatbot – that’s never gone wrong before


Meta has just unveiled the third version of its AI chatbot, and you can talk to it now.

Taking a break from the usual updates about the Oculus Quest 2 or the ways it will destroy Instagram next, Meta announced BlenderBot 3 – a follow-up to the previous BlenderBot and BlenderBot 2 robot prototypes.

Heading towards Website blenderbot.ai (opens in new tab) you can ask BlenderBot questions as if it were a search engine, or you can chat with it like you would a friend. If you use its search capabilities, BlenderBot will not only give you the answer you wanted, but will share where it got its information, allowing you to click through to learn more.

Unfortunately, there is one restriction on who can chat with Meta’s new bot: you have to be in the US. Some of our writers have tried using Express VPN to spoof their location from the UK, but were unsuccessful – however, you may find that a different VPN can bypass Meta’s location barrier.

Risky business

AI chatbots have a less than stellar track record when it comes to interacting with the general public

Microsoft famously released a Twitter chatbot in 2016 called Tay, which was quickly taken over by malicious users. Tay was designed to learn from the users who chatted with it. However, a concerted effort to spam Tay with hateful messages skewed the bot’s responses.

Just 16 hours after activation, Tay went offline. The vast majority of tweets on her Twitter account were deleted because they were filled with racist, homophobic and misogynist filth.

Don’t destroy BlenderBot 3, it looks so cute (Image: Meta)

Fortunately, Meta’s chatbot will likely be less prone to hijacking, as it won’t redirect user comments to it. Instead, it will rely on data previously fed to it by researchers. This so-called large language model (LLM) approach to artificial intelligence can help create some capable bots – like the hugely popular DALL-E Mini – but they have other flaws that Meta hopes to solve in its quest for a reliable digital assistant .

If an LLM AI doesn’t have an answer to your question in its database, it will often do what any parent does when their child starts quizzing them on a topic they know nothing about – they’ll make something up. Meta hopes that its new bot will be able to overcome this problem by using the Internet to find answers to certain questions, although Meta also wants to make sure that its bot is providing the right information from reliable sources.

That’s where you come in. Users testing BlenderBot 3 are invited to provide feedback if they choose to opt-in for their data collection.

We’ll have to wait and see how Meta’s experiment goes, but if you’re a developer who wants to dig deeper into the mechanics behind BlenderBot 3, you can check out the code and training dataset that Meta has shared. ParlAI (opens in new tab). You can even request access to the largest available copy model of BlenderBot 3 via a Google Form (opens in new tab).

If you want to see some other fun AI, here are the seven best masterpieces created by DALL-E Mini.

(via The lip (opens in new tab))



Source link

Exit mobile version