Google's Gemini AI makes factual error in video search demonstration
What's the story
During the recent I/O conference, Google's Gemini AI made a mistake in its video search demonstration.
As part of a demo, Google presented a video of a malfunctioning film camera and asked the AI to identify the issue. However, the AI provided an incorrect explanation.
One notable suggestion from Gemini was to open the back door of the camera and gently remove the film, a course of action that could potentially damage any photos taken by exposing them to light.
Previous error
Not the first misstep for Google's AI
This isn't the first time Google's AI has been caught providing incorrect information.
In a previous instance, the company's Bard chatbot (now called Gemini) made a false claim that the James Webb Space Telescope was the first to photograph a planet outside our Solar System.
However, it was the European Southern Observatory's Very Large Telescope (VLT) that took the first pictures of an exoplanet in 2004.
The factual error cost the company $100 billion in market capitalization.
Video search
Search with video is still an experimental feature
Google's experiment with video search powered by their generative AI model, Gemini. This new feature allows users to search the web using videos instead of text or images.
Gemini analyzes the video in real-time, identifying objects and actions to provide relevant search results in form of AI Overview.
It is aimed particularly at younger generations who are more accustomed to using video for communication.
The feature will be available as an experiment in Search Labs for users in the US.