Google CEO Sundar Pichai shows off capabilities of next-gen Gemini 1.5 Pro, says it is as good as Gemini Ultra

Just a week after Google started to roll out Gemini Ultra 1.0, the company has already made some progress and has released the next-generation Gemini 1.5 –– and the first Gemini 1.5 model that Google has released for testing is Gemini 1.5 Pro.

“It shows dramatic improvements across a number of dimensions and 1.5 Pro achieves comparable quality to 1.0 Ultra, while using less compute,” writes Google CEO Sundar Pichai in a blog post.Â

Pichai says that Google has made Gemini 1.5 smarter by giving it the ability to understand a lot more information at once. Now, it can handle up to 1 million pieces of information at a time, which is the most any big model has ever been able to do.

What can Gemini 1.5 do?

The Gemini 1.5 uses a new Mixture-of-Experts (MoE) architecture, which Demis Hassabis, CEO of Google DeepMind says is smart because it only uses the parts of the model’s brains that it needs for the job. This makes it work better and faster.Â

“Our latest innovations in model architecture allow Gemini 1.5 to learn complex tasks more quickly and maintain quality, while being more efficient to train and serve. These efficiencies are helping our teams iterate, train and deliver more advanced versions of Gemini faster than ever before, and we’re working on further optimisations,” Hassabis wrote in a Gemini 1.5 blog post.

Interestingly, during their testing, Google gave the Gemini 1.5 Pro model a 44-minute silent movie, and it was apparently able to “accurately analyse various plot points and events, and even reason about small details in the movie that could easily be missed”.Â