Bloomberg reported that Google has acknowledged that certain aspects of its demonstration video for Gemini, its new competitor to GPT-4, were staged.
The tech giant conceded that it altered the video’s speed and added a simulated interaction between a human voice and the AI user, raising questions about the authenticity of the presentation.
The original demonstration of Gemini was crafted using still image frames from the footage, accompanied by text prompts, rather than showcasing the AI’s response to real-time changes or drawings.
The revelation casts doubts on Gemini’s true capabilities, particularly because the demo did not include a disclaimer about these modifications. The lack of transparency in the video presentation has sparked concerns about the AI’s actual performance and effectiveness.
Gemini’s Leadership Defends the Video
Despite the controversy, Gemini’s co-lead, Oriol Vinyals, defended the integrity of the demonstration. Speaking on an unspecified platform (denoted as X in the original text), Vinyals insisted that “all the user prompts and outputs in the video are real.” He explained that the intention behind the video was to inspire and engage developers rather than to deceive viewers about Gemini’s capabilities.
The statement aims to reassure stakeholders about the authenticity of the interactions shown in the video, despite the admitted alterations.