Google’s new Gmail tool dramatizes non-existent emails

You only have one job!

Bachelor’s Assistant (R) Dr

It looks like Google’s new Bard extension will summarize your emails, plan your trips, and — oh, Yes – Fabricating emails you didn’t actually send.

Last week, Google plugged a major language model-powered chatbot called Bard In a group of Google products Including Gmail, Google Drive, Google Docs, Google Maps, and Google-owned YouTube, among other apps and services. While it’s understandable that Google would want to integrate its latest AI efforts into its already established suite of products, it appears that Google may have moved too quickly.

according to The New York Times columnist Kevin Roose, Bard isn’t the helpful inbox assistant that Google seems to want it to be — at least not yet. In his testing, Rose says, the AI ​​hallucinated entire email correspondence that never happened.

It’s a step above Bing’s AI He tells Rose that he should leave his wife Earlier this year, but still, although Google claims it’s still working out some bugs, it’s not exactly a promising first step.

Train to nowhere

According to the article, the hallucinations began when Rose asked Bard to “analyze all my Gmail messages and tell me, with a reasonable amount of certainty, what my biggest psychological problems are.” Although it’s a bit of a strange question, it’s simple enough. Bard quickly got to work, reportedly telling Rose that he tends to “worry about the future,” citing an email allegedly sent by the writer, in which Rose expressed that he was “nervous about work” and “scared of failure.”

See also  The Hellcat-side-by-side swap could only end badly

the problem? Rose never sent that email. The AI ​​has found a newsletter in the writer’s possession Receive – About Elon Musk’s new autobiography, of all things — and he misinterpreted the quote Within that bulletin To coin an entirely new email, Bard claimed Rose sent it himself.

According to the columnist, Bard repeated this behavior, pulling an email in which Rose allegedly complained that he was “not cut out to be a successful investor” out of thin air. The AI ​​also repeatedly got airline information wrong, and even created a train that didn’t exist.

In response to Rose’s concerns, Jack Krawczyk, director of Bard at Google, confirmed that Bard Extensions are still experimental, and that this is the first iteration of the product.

Despite these reservations, the new extension still seems pretty undercooked, and it’s not confidence-inspiring that Google would dare launch a product. this Messy — not to mention the big data privacy implications caused by AI crawling through your personal emails.

Overall, it’s hard to shake the feeling that Google, in its mad rush to maintain dominance in the AI ​​industry, might end up making some serious mistakes that could end in disaster.

More about artificial intelligence: Tech Bros lectured in Congress about AI “like school kids” who weren’t allowed to raise hands

Leave a Reply

Your email address will not be published. Required fields are marked *