[a: hover]: text- grey- 63]&, >,  , a: hover]: shadow- underline- dark black :]&, >,  , a: hover]: text- grey- bi dark :]&, >,  , a: hover]: shadow- underline- gray]&, >,  , a]: shadow- underline- gray- 63 dark :]&, >,  , a]: text- gray- bd dark :]&, >,  , a]: shadow- underline- gray “> Image: Google
Google recently unveiled Gemini, its latest and most advanced AI system. However, the company is facing accusations of misleading claims about its capabilities.
A report from Bloomberg alleges that Google may have misrepresented the capabilities of Gemini in a promotional video. During the recent announcement, Google showcased a captivating hands-on video demonstrating Gemini’s capabilities, which according to columnist Parmy Olson, appeared highly impressive, if not exaggerated.
The brief video highlighted Gemini’s bidirectional functionalities, showcasing its ability to process spoken commands in conjunction with image recognition. Gemini reportedly excels in instant image recognition, even with complex tasks like connecting the dots, providing rapid responses, and tracking objects in real-time, such as a cup and ball game. While individuals can perform similar tasks, the distinction lies in Gemini’s artificial intelligence that can anticipate future actions.
However, upon closer inspection of the video description on YouTube, Google included a disclaimer stating:
“For the purpose of this video, overhead has been minimized, and Gemini’s responses have been shortened for clarity.”
This statement raised concerns for Olson, who pointed out that Google’s demonstration did not occur in real-time with live verbal interactions as implied. Instead, the video utilized still image frames from raw footage with text prompts, to which Gemini responded. This revelation contradicted the initial impression conveyed by Google, suggesting a more interactive real-time experience with Gemini.
While it is common practice to edit and enhance video demonstrations for clarity and simplicity, Google has faced skepticism in the past regarding the authenticity of its AI showcases. Questions arose following the release of Google’s Duplex demo, where doubts were raised about the realism of the interactions due to the lack of background noise and overly cooperative responses. Pre-recorded demonstrations of AI technologies tend to evoke skepticism among viewers, as seen with Baidu’s Ernie Bot release and subsequent stock decline.
In response to the criticism, Google defended its approach. The Verge referenced a blog post by Oriol Vinyals, Vice President of Research and Deep Learning Lead at Google DeepMind, and co-lead for Gemini, explaining the rationale behind the video creation process.
According to Vinyals, “All user interactions and outputs in the video are genuine, albeit condensed for clarity.” The video aimed to illustrate the potential customer experience with Gemini and inspire developers.
While this perspective offers insight into Google’s intentions, the company’s credibility may be at stake, especially following recent comparisons with OpenAI’s GPT. To build trust and credibility, Google could consider providing hands-on experiences of Gemini to developers and journalists, allowing them to witness its capabilities firsthand. By offering open beta access and real-world demonstrations, Google can showcase Gemini’s true potential and foster greater understanding and appreciation for its AI advancements.
