Google is currently under scrutiny for the demonstration video accompanying the launch of its latest artificial intelligence model, Gemini. In a move that coincided with the approach of the year’s end, Google introduced Gemini, touting it as their most extensive and advanced AI model to date. The demonstration video, shared with both media outlets and the general public, has become a focal point of controversy.
The six-minute video showcases spoken interactions between the user and a chatbot powered by Gemini. Additionally, it highlights Gemini’s proficiency in recognizing visual images and physical objects, distinguishing between them accurately. Several notable capabilities were demonstrated, including Gemini’s capacity to articulate verbal descriptions of drawings, such as distinguishing between a realistic duck and a rubber duck, among other examples.
The company’s YouTube description includes a brief statement that mentions, “For the purposes of this demo, latency has been reduced, and Gemini outputs have been shortened for brevity.” Notably, this disclaimer is absent in the actual video.
Subsequent to the launch, the company clarified to Bloomberg that the demonstration was not conducted in real time. Instead, it utilized static images and provided text prompts to which Gemini responded, a fact previously highlighted by The Information. The author emphasized the substantial disparity between this approach and what Google seemed to imply: the possibility of having a seamless voice conversation with Gemini in real-time as it observed and responded to its surroundings.
Following numerous requests for comment, the company issued a statement to CNBC on Friday, asserting, “The video is an illustrative representation of the potential interactions with Gemini, derived from actual multimodal prompts and outputs during testing. We anticipate witnessing the creative applications people will explore when access to Gemini Pro becomes available on December 13.”
While edited demonstrations are not uncommon, the revelations following the Gemini demo evoke a sense of déjà vu for the search giant. Earlier in the year, Google encountered public and Wall Street criticism for a demonstration of its AI chatbots that its own employees labeled as a “rushed, botched” presentation. This incident occurred during the same week that Microsoft had planned to showcase its Bing integration with ChatGPT.
Earlier this month, The Information revealed that Google abandoned its initial plans for in-person events to introduce Gemini, ultimately opting for a virtual launch. The tech giant finds itself in intense competition with OpenAI’s GPT-4, backed by Microsoft, which has held the position of the most advanced and successful model thus far. This week, Google published a white paper asserting that Gemini’s most potent model, “Ultra,” exhibited incremental superiority over GPT-4 across various benchmarks.
