Recently, Google introduced its new AI platform, Gemini, through a promotional video. This release quickly drew criticism from both users and the company’s own employees. The controversy surrounds the authenticity of the video, as insiders claim that Gemini’s actual performance differs significantly from what was shown.
Dissecting Google’s AI Demonstration
The video, initially presented as a live demonstration, was later revealed to involve significant post-production work. AI expert Rowan Cheung explained that the footage was edited, with researchers entering prompts off-screen rather than using live voice commands. A notable example includes the “duck video,” where Gemini’s responses were artificially shortened and the latency reduced, contrasting with its real-time performance. Although Google maintained the integrity of the words used, the overall representation was deemed misleading.
Cheung highlighted a specific segment of the demo as especially misleading. In it, a hand gesture is interpreted by Gemini as playing “Rock, Paper, Scissors.” However, Cheung clarified that this interaction was staged using preselected prompts and images, rather than live interaction. This revelation raises questions about the actual advancement of Gemini compared to its rivals, like ChatGPT.
Internal Responses and the Broader AI Landscape
Bloomberg reported that Google employees have expressed concerns about the misleading nature of the video, believing it sets unrealistic expectations for Gemini’s capabilities. Despite Gemini’s performance on some benchmarks, the advanced features shown in the video are still under development.
Peter Yang, a product lead at Roblox, pointed out that the controversy has shifted the focus away from Gemini’s strengths to the video’s inaccuracies, branding it a PR misstep. This incident has sparked parodies on social media, reflecting the growing skepticism.
In other AI news, the European Union has taken a pioneering step in regulating AI usage through the new ‘Artificial Intelligence Act.’ This legislation focuses on transparency, risk management, and human oversight for high-risk AI applications. Google, a major player in AI, recently launched its AI note-taking app, NotebookLM, in the U.S., featuring document transformation and Gemini integration. Additionally, AMD is making strides in AI hardware with its latest MI300X and MI300A AI accelerators.