SOPA Images
- Google's much-lauded demo of Gemini showed the AI responding in real-time to events on screen.
- However, the company confirmed it edited out the human intervention that made Gemini work.
- "We made it to inspire developers," said Google DeepMind's head of research.
An "incredible" demonstration of Gemini released by Google was edited to remove the human intervention that made the AI look slick and intuitive.
In the video, Gemini appeared to be responding in real-time to what it could see and the human's voice. Google touted this "multimodality" — or the ability to process information like video and audio — as a key feature of Gemini.
At one point in the demo, a humorous interaction unfolds when the AI, after remarking on the rarity of blue ducks, is shown a blue duck toy by a human.
Seeing some qs on what Gemini *is* (beyond the zodiac :). Best way to understand Gemini’s underlying amazing capabilities is to see them in action, take a look ⬇️ pic.twitter.com/OiCZSsOnCc
— Sundar Pichai (@sundarpichai) December 6, 2023
Surprised, Gemini exclaimed: "What the quack! I was just talking about a blue duck, and now you're holding one! It looks like blue ducks are more common than I thought."
However, a Google representative told Bloomberg on Thursday that the demo involved "using still image frames from the footage, and prompting via text."
A Google blog post published Wednesday gave a glimpse into how this process worked in practice.
In short, there was no real-time conversation or reaction. Gemini works much like rival chatbot ChatGPT does, through text prompts and uploaded images.
Gemini's demo was shared widely on Google platforms, while CEO Sundar Pichai's post on X racked up more than 7 million views. However, the disclaimer that "latency has been reduced and Gemini outputs have been shortened for brevity" only appeared on the video's YouTube description.
"All the user prompts and outputs in the video are real, shortened for brevity. The video illustrates what the multimodal user experiences built with Gemini could look like," Oriol Vinyals, vice-president of research at Google DeepMind, wrote on X. "We made it to inspire developers."
Since its release, the demo has drummed up immense hype for Gemini.
One user resharing the video wrote that Gemini showed "broader intelligence than a non-zero fraction of adult humans." Another use wrote that watching the demo made their jaw drop.