Unmasking Google’s AI Video: A Closer Look at the Gemini Demo

In the world of artificial intelligence (AI), showcasing groundbreaking capabilities is crucial to capturing the attention of consumers and investors. Google’s Gemini AI model recently made waves with a viral video that demonstrated its remarkable abilities. However, upon closer examination, it became evident that the video wasn’t entirely what it seemed. This article delves into the details of the Gemini demo, the controversy surrounding it, and what it means for the future of AI technology.

The Viral Gemini Demo

The Gemini demo video garnered significant attention, amassing 1.6 million views on YouTube. In the video, an individual engages in a seemingly real-time conversation with Google’s AI, using spoken-word prompts and video presentations. The interactions appear seamless, creating an illusion of the AI’s astounding capabilities.

Hands-on with Gemini: Interacting with multimodal AI

The Truth Behind the Demo

Upon closer inspection, it became clear that the Gemini demo was not an accurate representation of the AI’s functionality. Google admitted that it had altered the video to make the responses appear faster for demonstration purposes. Additionally, it was revealed that the AI did not respond to voice or video inputs as portrayed in the video.

In a blog post published concurrently with the video, Google provided a detailed explanation of how the video was actually created. It confirmed that the AI’s responses were generated by using still image frames from the footage and prompting the AI via text.

While the video showcases a back-and-forth conversation where the AI correctly identifies objects and answers questions, the reality is quite different. In each scenario, the AI was presented with still images and text prompts to guide its responses. This revelation raises questions about the authenticity of the video and the true capabilities of Gemini.

Dissecting the Gemini AI Video

Several moments in the demo video highlight the disparities between what was shown and how the AI actually operated:

1. Rubber Duck Identification: In one scene, the demonstrator holds up a rubber duck and asks the AI if it will float. The video suggests that the AI identifies the material by squeezing the duck and recognizing the squeaking sound. However, the reality is that the AI was shown a still image of the duck and asked about its material. It was then given a text prompt about the squeaking noise, leading to the correct identification.
2. Cups and Balls Magic Trick: Another impressive feat in the video occurs when the user performs a cups and balls magic trick. The AI seemingly tracks the movement of the ball beneath the cups. In truth, this was achieved by presenting the AI with a series of still images.
3. Game Invention: At one point, the user asks the AI to invent a game based on a world map and emojis. The AI responds by creating a game called “guess the country.” However, the AI did not invent this game independently. It was given specific instructions to play a game involving clues and pointing to a country on a map.

Gemini vs. OpenAI’s GPT-4

While Google’s Gemini AI model remains impressive, the use of still images and text-based prompts brings its capabilities closer to that of OpenAI’s GPT-4. The controversy surrounding the video’s authenticity raises questions about the true extent of Gemini’s capabilities and whether it genuinely represents a significant leap in AI technology.

The Context of the AI Industry

The release of the Gemini demo video occurred in the midst of significant turmoil in the AI industry. Just weeks before the video’s debut, Sam Altman’s dramatic firing and rehiring as CEO of OpenAI created unprecedented chaos in the AI space. The rivalry between Google and OpenAI has fueled speculation about which organization holds the upper hand in AI development.

The Complex Reality of AI

Bottom-line: Google’s Gemini demo video serves as a reminder that the world of AI is complex and multifaceted. While the video was undoubtedly attention-grabbing, its embellishments have generated discussions about authenticity and the need for transparency in AI demonstrations. Regardless of the controversy, both Google and OpenAI continue to push the boundaries of AI technology, leaving us to wonder what innovations lie ahead in this ever-evolving field.