Google has admitted to staging the Gemini AI demo video but denies any wrongdoing, saying it was only lightly edited.
In an opinion piece published in Bloomberg, Google admitted to staging parts of a demo video for Gemini, a newly released competitor to GPT-4. The company not only edited the output speed, but also added an implied voice-over between the human voice and the AI user.
The original demo was made to “use still frames of video and text instructions” rather than have Gemini react in real time to drawings or changes to objects on the table. This misleading video calls into question Gemini's capabilities, especially since there was no disclaimer in the demo.
https://youtu.be/UIZAiXYc
Gemini co-leader Oriol Viñals denied any wrongdoing on X, saying that “all user prompts and outputs in the video are real” and that his team produced the video “to inspire developers.”
We're really excited about your interest in our “Gemini in Action” video. Yesterday's dev blog detailed how Gemini was used to create this video: https://t.co/50gjMkaVc0
We fed Gemini sequences of different modalities (in this case images and text) and had it react to them… pic.twitter.com/Beba5M5dHP
Load the next story
