Google Found Faking It, Lambasted Over Gemini AI Demo Video

Tech giant Google received backlash for a promotional video the company published for its new AI platform, Gemini. The technology doesn’t quite work the same way as the video portrayed, as company employees have noted. 

Google launched its “ ChatGPT killer ” artificial intelligence model last week , lauding its capabilities in a demo video . 

Google Artificial Intelligence Video Slammed 

After heavy criticism from users and company employees, the firm has admitted it faked it. 

The video was not recorded live, and the prompts were not given by voice. In reality, frames were extracted, and researchers entered prompts off-camera, noted AI expert Rowan Cheung. 

In the duck video, Gemini shortened its responses and reduced latency compared to its actual functioning. Google stated that while they didn’t change individual words, the depiction wasn’t completely accurate.

Cheung picked part of the video demo that he found “particularly shocking.” On-screen, a hand made gestures, and Gemini responded, “I know what you’re doing! You’re playing Rock, Paper, Scissors!”

He said, in reality, they used prompts and still images behind the scenes and just used the video for show, before adding:

“Was this just a simple PR miscommunication, or is it farther behind ChatGPT than initially thought?”

Read more: AI Stocks: Best Artificial Intelligence Companies To Know in 2023

According to Bloomberg , some Google employees are also calling out those discrepancies internally.

One employee told the outlet that, in their view, “the video paints an unrealistic picture of how easy it is to coax impressive results out of Gemini.”

Furthermore, Gemini exceeds other AI models on some benchmarks, but more advanced capabilities shown in the video remain in research.

Product lead at Roblox, Peter Yang, commented : 

“All signals show that Gemini is a great model but now people are writing articles about this video instead. This seems like a PR fail.” 

Parodies were already being posted on social media. 

AI Latest News 

The European Union has become the very first continent to set clear rules for the use of artificial intelligence.

The new ‘Artificial Intelligence Act’ aims to ensure oversight of high-risk AI through transparency, risk management, and human reviews.

Google, which is heavily invested in AI , has released its AI note-taking app, NotebookLM, to US users. The move includes several new features, including document transformation and Gemini integration.

Meanwhile, semiconductor firm AMD has made a substantial leap forward with the release of its new MI300X and MI300A AI accelerators.