A "hands-on" tech demonstration of Google's new AI model, Gemini, has gone from being the talk of the town to being accused by critics of being "basically completely fake."
The six-minute video, which has garnered 2.1 million views on YouTube since its release on Dec. 7, shows him seamlessly interacting with a human operator in seemingly real time, including analyzing a drawing of a duck, gestures with hands and inventing a game called "Guess the country" with just a picture of the world map.
However, Oriol Vinyals, a Google Deepmind executive, has since clarified that while the user prompts and results in the video are real, it has been "shortened for brevity." In reality, Gemini's interactions were text-based, not voice-based, and took much longer than was depicted in the video.
Google even admitted as much by uploading the video to YouTube, with a disclaimer saying: "For the purposes of this demo, latency has been reduced and Gemini outputs have been shortened for brevity."
This, however, has not caused an avalanche of criticism towards Google on social networks.
"Google LIED. The AI demo that flaunted Gemini's capabilities was FAKE." saying a software developer, "Nelly RQ", software developer in an X (formerly Twitter) post on December 10.
"It was edited, cut to make it look faster and more capable than it really is," said "Boss Nerd," another software engineer.
NEW: Google shares sink after reports that part of its Gemini AI demo was faked
"So that Gemini demo video that everyone was going crazy over... It was edited, cut to make it look like it was faster and more capable than it actually is... So I've been pushing. .. pic.twitter.com/9f7UmdLOlA
- Chief Nerd (@TheChiefNerd) December 8, 2023
Even Google employees themselves have reportedly taken issue with the video.
A Google employee said Bloomberg that the video shows an unrealistic image of Gemini and shows how easy it is to make the artificial intelligence tool appear more advanced than it really is. The company's CEO Sundar Pichai also did not mention the adjustments made to the six-minute video in a December 7 statement. mail.
However, another Google employee said he wasn't too surprised by the demonstration, as a certain degree of marketing is required to promote these types of products.
Another Google employee told Bloomberg that individual words in Gemini's responses were not changed, and the voiceover captured excerpts from Gemini's actual text. "The user voiceovers are all actual excerpts from the actual prompts used to produce the Gemini output that follows."
Related: AI-powered fake nude services skyrocket in popularity: research
At its launch, Gemini's "hands-on" tech demo garnered impressive reactions from viewers.
“Google's new Gemini AI watches a man draw a duck and explains what he's doing every step of the way; Not only literally or mechanically, he can infer what the man is trying to do, why he is doing what he does. He feels… very human.” saying Armand Domalewski, data analyst at Palo Alto Networks in a December 7 post.
Launched as a competitor to OpenAI's ChatGPT, says Google Gemini outperforms leading AI models in 30 of 32 benchmarks that assess reasoning, mathematics, language, and other metrics, including GPT-4 in seven of eight of those benchmarks.
Magazine: Real Cryptocurrency AI Use Cases: Cryptocurrency-Based AI Markets and AI Financial Analysis
Post navigation
