Real or Fake: Google’s Gemini AI Demo Sparks Controversy

Google’s latest AI model, Gemini, revealed in a “hands-on” tech demo, is now facing allegations of being “fake.”

The six-minute video, reaching 2.1 million views on YouTube and counting, demonstrated seamless interactions with a human operator, plus evaluating a drawing of a duck, hand gestures, and creating a game called “Guess the Country.”

But critics are now interrogating the authenticity of the demo. Google Deepmind executive Oriol Vinyals clarified that the video’s user prompts and outputs were genuine but had been “shortened for brevity.” Contrary to the video’s representation, Gemini’s interactions were text-based and took longer in reality.

NEW – Google's Stock Sinks Following Reports That Some of Their Gemini AI Demo Was Faked

"So that video demo of Gemini that everyone was kind of going crazy over…It was edited, it was cut to look like it was faster and more capable than it actually is…So I've been pressing… pic.twitter.com/9f7UmdLOlA

— Chief Nerd (@TheChiefNerd) December 8, 2023

Despite a disclaimer acknowledging the video’s alterations, social media has erupted with criticism. Software developers and Google employees alike have expressed skepticism, with accusations of editing to make Gemini appear more advanced than it truly is. Google CEO Sundar Pichai did not refer to the fine-tuning in a December 7 post, further fueling the controversy.

While some employees argue that marketing requires a certain degree of promotion, others maintain that the demo presents an unrealistic image of Gemini. A Google employee informed Bloomberg that individual words in Gemini’s responses were unaltered, with the voiceover capturing excerpts from actual text prompts.

Google LIED.

The AI demo flaunting Gemini's capacities was a FAKE.

Here’s the breakdown: pic.twitter.com/3x3fvydaLO

— Nelly R Q (@nrqa__) December 10, 2023

Initially lauded for its impressive capabilities, Gemini was positioned as a competitor to OpenAI’s ChatGPT. Google asserts that Gemini outperforms leading AI models in 30 out of 32 benchmarks, including GPT-4 in seven out of eight categories, testing reasoning, math, language, and other metrics.

The controversy surrounding the authenticity of the demo raises questions about transparency in AI presentations and the expectations set for emerging technologies. As debates continue, the tech community awaits further clarification from Google on the true capabilities of its Gemini AI model.

Related story: Google Updates Cryptocurrency Ads Policy