Google Admits That Gemini AI Demo Video Was Staged: Everything To Know

Google's Gemini AI demo faces backlash as doubts arise over the veracity of a demo video

Highlights
  • Google faces backlash over deceptive Gemini AI demo, edited to speed up outputs and lacking real-time interaction.
  • Authenticity concerns arise as the demo used still frames and text prompts, casting doubt on Gemini’s practical readiness.
  • The incident adds to Google’s history of questionable demos, echoing past scepticism over the authenticity of the Duplex demo.
371209

Artificial Intelligence (AI) is rapidly advancing, with OpenAI’s ChatGPT leading the way. Major tech giants are fiercely competing in this evolving landscape, but the pursuit of innovation sometimes compromises authenticity. Google recently unveiled Gemini, its most potent AI model, but within 24 hours, the tech giant faced criticism for the authenticity of its demo video.

Gemini AI’s Deceptive Demo

Google’s demo video, showcasing Gemini’s capabilities as a large language model (LLM), was shared by CEO Sundar Pichai. The video depicted the AI’s ability to generate game ideas or provide translations for a rubber duck held by a user. The apparent conversational skills and real-time information updates impressed viewers. However, it has come to light that the demo was not entirely genuine.

Bloomberg reported that Google admitted to editing the video titled “Hands-on with Gemini: Interacting with multimodal AI” to accelerate the displayed outputs. The video description did acknowledge this editing but failed to disclose a more significant issue: the absence of actual voice interaction between the human user and the AI.

Contrary to the impression given in the video, Gemini’s interaction was not live. Instead, the demo was created using still frames and text prompts rather than responding or predicting real-time changes in the environment. The omission of this crucial detail raises doubts about Gemini’s practical readiness.

Google’s Response and Clarification

Oriol Vinyals, the co-lead of Gemini at Google, addressed the controversy in a post on X. He explained that while the user prompts and outputs were real, they were shortened for brevity. The video aimed to illustrate potential multimodal user experiences with Gemini, intending to inspire developers. However, the statement appears to carry an undertone of taking creative liberties.

Also Read: Google Gemini is Its Most Capable Multimodal AI: Coming First to Bard, Pixel 8 Pro

Google’s Track Record of Questionable Demos

This incident isn’t the first time Google has faced scepticism over the authenticity of its demo videos. The tech giant previously encountered doubts about the legitimacy of its Duplex demo, featuring an AI voice assistant making reservations at hair salons and restaurants. Concerns were raised due to the absence of ambient noise and overly cooperative employees.