You know that scene in every sci-fi movie where the computer gains sentience and immediately tries to destroy humanity?
Turns out the real version is way dumber.
Google's AI Overviews feature—the shiny new toy slapped on top of search results—recently informed users that humans should eat "at least one small rock per day" for minerals and vitamins.
This was just a language model confidently regurgitating a satirical Onion article like it was the Mayo Clinic's official dietary guidance.
The internet, being the internet, responded with the maturity of a middle school cafeteria.
Within hours, screenshots of the rock-eating advice were everywhere—X, Reddit, TikTok, your uncle's Facebook page.
People started asking the AI for other health tips just to see what cursed nonsense it would generate.
Someone reportedly got it to suggest adding "non-toxic glue" to pizza sauce to keep cheese from sliding off.
Another user coaxed it into claiming Barack Obama is Muslim.
It was like watching a toddler with a megaphone and a library card.
For those keeping score at home, this is the same company that spent billions developing AI and then shipped it without apparently testing whether it would tell people to consume geology.
Google has since said it's "taking action" on the weird results and rolling back some features.
Translation: they're in damage control mode while their PR team Googles "how to apologize without admitting you broke the internet." Here's the thing though—this isn't just a Google problem.
Every major tech company is racing to shove AI into everything.
Microsoft's Copilot, Apple's upcoming AI features, Meta's chatbots that are apparently designed to be as annoying as your most opinionated coworker.
They're all doing it because Wall Street demands it, not because anyone actually asked for it.
The result is a bunch of half-baked features that confidently lie to your face and then shrug when caught.
And before anyone starts with the "this is why we can't trust AI" hot takes, let's be clear: AI can be genuinely useful.
It can summarize documents, help you write emails that don't sound like a robot, and occasionally tell you a decent recipe that doesn't involve gravel.
The problem isn't the technology itself—it's the reckless deployment.
You don't hand a toddler a chainsaw and then act surprised when the family cat goes missing.
What's actually funny about this whole mess is how it exposes the gap between how tech companies talk about AI and what AI actually does.
We're told it's "transformative" and "revolutionary" and "the future of human civilization." In reality, it's a fancy autocomplete that will confidently tell you to eat rocks if it thinks that's what you want to hear.
That's the world's most expensive yes-man.
Maybe stop treating every AI announcement like it's the second coming of the printing press.
They're built by companies trying to beat earnings estimates, not by benevolent geniuses trying to save the world.
A healthy dose of skepticism isn't being anti-technology—it's just being an adult who doesn't take dietary advice from a chatbot.
The real question isn't whether AI will eventually get better.
It's whether we'll stop letting companies beta-test their products on the entire population.
Because right now, the only thing being disrupted is our collective faith in the tech industry's basic competence.
Final Thoughts
And no amount of "we take this seriously" press releases is going to fix that.