Google’s Super Bowl ad had to be changed last minute — and it’s all Gemini’s fault
Super Bowl commercials are a very big deal, and reports claim that each spot can cost upwards of $8 million — in part thanks to all the eyes that will be watching.
Needless to say you make sure everything is in order, and that ad isn’t going to cause any embarrassment. Hence why Google has had to change its Super Bowl commercial, thanks to its AI’s horrible habit of getting things wrong.
The Verge noticed Google has edited its Gemini Super Bowl ad, thanks to the AI stating an incorrect fact about cheese. Of all the ways Google AI could get something wrong, it ends up being about cheese. Specifically Gouda, which Gemini claimed “makes up “50 to 60 percent of the world’s cheese consumption.”
I’m no turophile, but even I know that Gouda is not going to be quite that popular. I’d have guessed cheddar, but Google search results suggest that it may be mozzarella. Which isn’t a huge surprise, considering how popular pizza is.
Image 1 of 2
Google Cloud apps president Jerry Dischler defended the original text on X, claiming the mistake is “not a hallucination” because “Gemini is grounded in the web.”
Dischler even notes that Gemini results feature sources for users to check themselves, but they don’t seem to be visible in the ad. The Verge speculates that this particular fact may have been pulled from a site called Cheese.com, which features lots of facts about cheese and few (if any) sources to back up the claims.
This fake Gouda fact was never spoken out loud in the original ad, but it was written on screen for those who cared to look. The site also notes that the business featured in Google’s ad actually had Gemini’s response on its website. This text has now been altered in both cases, now declaring that gouda is one of the most popular cheeses in the world, rather than relying on actual statistics.
Gemini does have a disclaimer in the ad claiming that the results are ‘not intended to be factual’
Gemini does have a disclaimer in the ad claiming that the results are “not intended to be factual.” Though that’s obviously a problem. It’s bad enough that Google’s AI is pulling incorrect information from untrustworthy sources, but presenting that information without clearer caveats is a sure-fire way to spread misinformation.
That’s something you should always bear in mind when searching the web, especially if an AI is only delivering a small snippet of the available information. Because you never know if it’s just feeding you some nonsense that happens to rank highly on search engine.
More from Tom’s Guide
Source link
