The, at first glance, controversial demo video Google presented about the Gemini model for generative AI turns out to be largely fake. In particular, the video highlighted Gemini's multimodal nature – being able to handle audio, images, text and video – as an example of Gemini's capabilities.

In reality, a large part of the video appears to have been put together with loose material of which the presented context - Gemini responds with speech to an image that is presented - is not part of it. Opposite Bloomberg Google has admitted that it also worked with text prompts and not with speech, as the video suggests. (Ziptone/Bloomberg)

Featured, Technology