Google’s new version of Gemini can handle far bigger amounts of data

According to Oriol Vinyals, a deep learning team lead at DeepMind, AI operates much like our brain by only activating certain parts at a time. This compartmentalization helps save computing power and allows for faster responses.

Oren Etzioni, former technical director of the Allen Institute for Artificial Intelligence, who was not involved in the work, finds the AI’s ability to fluidly switch between modalities impressive. He notes that this behavior is unprecedented.

An AI that can operate across modalities would mirror human behavior more closely. Etzioni highlights that humans naturally switch between speaking, writing, and drawing to convey ideas effortlessly.

However, Etzioni warns against putting too much trust in AI demonstrations. He mentions the cautionary phrase “Never trust an AI demo,” and points out that demonstration videos may omit crucial details or selectively showcase tasks. AI models, he adds, are generally fragile.

The release of Gemini 1.5 Pro is currently restricted to developers and enterprise customers. Google has not specified when it will be available to a wider audience.