December 10, 2023: Google’s recent demonstration of its Gemini AI model, designed to showcase the advanced capabilities of this new artificial intelligence tool, has been revealed to contain more than a touch of movie magic.
The video, initially presented as a real-time interaction between a human and Gemini AI, turned out to be a carefully edited portrayal, not an actual live demonstration.
This revelation raises questions about the current capabilities of Gemini AI and the ethical implications of presenting edited demos as real interactions.
The demo video, which wowed audiences with its portrayal of Gemini AI responding to various prompts and interacting with a human user, was found to be less than genuine. Google has admitted to editing the video to enhance the output speed and remove the actual voice interaction elements.
The reality was a demonstration created using still image frames and text instructions, a far cry from the real-time interaction initially implied.
This revelation by Google has sparked discussions about the readiness of Gemini AI for public use. The lack of proper disclaimers in the original video misled viewers, leading to doubts about the authenticity of the AI’s capabilities.
Oriol Vinyals, co-lead of Gemini at Google, stated, “All the user prompts and outputs in the video are real, shortened for brevity… We made it to inspire developers.”
However, this statement has been met with skepticism, as the portrayal in the video suggests a level of real-time interaction and comprehension not yet achievable.
Really happy to see the interest around our “Hands-on with Gemini” video. In our developer blog yesterday, we broke down how Gemini was used to create it. https://t.co/50gjMkaVc0
— Oriol Vinyals (@OriolVinyalsML) December 7, 2023
We gave Gemini sequences of different modalities — image and text in this case — and had it respond… pic.twitter.com/Beba5M5dHP
This incident is reminiscent of Google’s 2018 Duplex demo at Google I/O, where an AI voice assistant was shown making reservations at a salon and restaurant.
Like the Gemini demo, the Duplex demo was later revealed to be pre-recorded rather than live, raising similar concerns about transparency in AI demonstrations.
The editing of such demo videos is not uncommon in the tech industry, and in many cases, it is necessary to convey the potential of a product succinctly.
However, the lack of clear disclaimers and the portrayal of these demos as real-time interactions can be misleading.
This incident highlights the need for greater transparency in AI demonstrations and a clearer distinction between current capabilities and aspirational goals.
