Google pulls Gemini AI demo video after ad watchdog complaint

Google has delisted a popular but controversial demo video of its Gemini AI platform after an advertising watchdog group complained that the performance was inaccurate. This continues a streak of PR flubs to hit Gemini since the platform rebranded from Bard less than a year ago. In the demo video, which has amassed over 3 million views on YouTube since Google posted it in December, a person appears to be interacting with Gemini in real time, asking it questions based on video footage and receiving almost-immediate responses from the AI. But this portrayal is actually a dramatization of a far more stilted and slower process. The National Advertising Division (NAD) of non-profit BBB National Programs submitted the inquiry to Google in January because it felt that the representation could mislead viewers. NAD routinely monitors advertising for emerging technology to protect consumers. “It is crucial that consumers—including both end-users and sophisticated developers—receive truthful and accurate information concerning this technology,” said Laura Brett, VP of NAD, in a statement to Ad Age. “Google is pleased to accept NAD's resolution of this matter,” a Google spokesperson said. The tech giant was not forced to delist the video, but voluntarily chose to do so in agreement with NAD, the watchdog said. The demo will no longer appear in search results or be published as a standalone video in other channels. But it will still exist in its original blog post from December, in which Google spells out how the demo was actually constructed. Also read: Google conceals brands’ first-party data in new ad product This blog post paints a far different picture of Gemini’s performance than the one shown in the video, in which Gemini appears to instantly process speech, recognize visual patterns and perform analysis all at the same time—a capacity called multimodal prompting. For instance, Gemini never processed video footage as the demo suggests, but rather sequences of images accompanied by text prompts that encouraged the AI to process specific information. The demo also implies at several times that Gemini can intuitively understand how magic tricks work. Behind the scenes, however, Google needed to painstakingly walk it through the various activities for it to participate. These discrepancies caused controversy back when the video was first published and gaining traction online. Several news outlets reported that the demo was fraudulent, but Google stood firm, until now. NAD’s successful inquiry is the latest public flub to trip up Gemini. Earlier this summer, Google pulled an ad for the platform that was airing during the Olympics after viewers deemed it tone-deaf. The spot showed Gemini automating a child’s letter to an athlete they admired. And in February, Gemini went rogue after an update encouraged it to spout insensitive historical inaccuracies.

Google pulls Gemini AI demo video after ad watchdog complaint

Featured Stories