- Google's Gemini in Search demo video released on Tuesday contained factual errors.
- Gemini suggested opening a film camera without using a dark room, as the photos would be ruined.
- Google's demo videos have made similar mistakes in the past.
In two days of big announcements, OpenAI and Google unveiled their latest artificial intelligence projects to the world.
They created an impressive demo video showcasing all the new features OpenAI's ChatGPT-4o can do and how Google's Gemini will revolutionize search as we know it.
But Google's Tuesday video shows one of AI's big pitfalls: not just bad advice, but bad advice. In the first minute of the flashy, fast-paced video, Google Search's Gemini AI presented a factual error first spotted by The Verge.
The photographer takes a video of a malfunctioning film camera and asks Gemini, “Why doesn't the lever move completely?” Gemini will immediately present a list of solutions. This includes destroying all of his photos.
The listing's video highlights one suggestion: “If your camera is jammed, open the back door and gently remove the film.”
Any professional photographer, or anyone who has ever used a film camera, knows this is a terrible idea. If you open the camera outdoors where the video is being shot, exposure to bright light can ruin some or all of the film.
Google faced similar problems with previous AI products.
Last year, a Google demo video showing a bird chatbot incorrectly claimed that the James Webb Space Telescope had photographed the first planet outside our solar system.
Earlier this year, Gemini's chatbot came under fire for refusing to create photos of white people. He was criticized for being too “woke” and for deploying photos that outraged historical inaccuracies such as Asian Nazis and black Founding Fathers. A Google executive apologized, saying he “missed the mark.”
Tuesday's video highlights the dangers of AI chatbots causing false predictions, hallucinations and giving users the wrong advice. Last year, users of Bing, Microsoft's AI chatbot, reported strange interactions with the bot. He called users delusional, tried to gaslight them about what year it was, and even confessed his love to some users.
Companies using such AI tools may also be held legally responsible for what the bots say. In February, a Canadian court held Air Canada liable for a chatbot that provided passengers with false information about bereavement discounts.
Google did not immediately respond to a request for comment outside standard business hours.
On February 28, Axel Springer, the parent company of Business Insider, joined 31 other media groups in filing a $2.3 billion lawsuit against Google in Dutch court, alleging losses caused by the company's advertising practices. I woke you up.
