AI Viral Video Scandal: Google Admits to Strategic Editing

• Google’s Gemini AI demo was edited, not real-time.
• Still images and text, not live prompts, were used in the demo.
• Critics question demo accuracy amid industry competition.

Google showcased the capabilities of its artificial intelligence (AI) model, Gemini, in a seemingly impressive demonstration. However, the tech giant later admitted that the video was edited, and the AI did not respond in real-time to voice or video prompts.

The Deceptive Demo: The video, which amassed 1.6 million views, featured a person interacting with Google’s AI, asking questions and presenting objects on screen. Notably, the AI correctly identified objects and even participated in a magic trick. However, Google later revealed that the responses were sped up for the sake of the demo, and the AI was not reacting to the live video.

How It Was Really Done: According to Google’s blog post, the AI was prompted using still image frames from the footage and text prompts, not live voice or video prompts. For instance, when asked about the material of a rubber duck, the AI was shown a still image of the duck and provided with a text prompt explaining the squeaking sound.

Staged Game Creation: The demo included a segment where the AI seemingly invented a game based on a world map and emojis. However, it was later clarified that the AI did not come up with the game. Instead, it was given specific instructions, and the user provided examples of correct and incorrect answers. The AI then generated clues and identified correct countries from stills of a map.

Google’s Explanation: In response to inquiries, a Google spokesperson stated, “Our Hands on with Gemini demo video shows real prompts and outputs from Gemini. We made it to showcase the range of Gemini’s capabilities and to inspire developers.”

Implications and Industry Dynamics: The video’s release comes amidst recent upheavals in the AI space, including leadership changes at OpenAI. While Google’s AI model remains impressive, critics argue that its capabilities, relying on still images and text prompts, are similar to existing models like OpenAI’s GPT-4. As the AI competition intensifies, questions arise about the accuracy and transparency of AI demos presented by tech giants.

Also Read

Shell Announces $5bn Investment Boost for Nigeria’s Offshore Oil

Shell to invest $5 billion in Nigeria's Bonga North oil project. This...