Google Reveals Ai ‘reasoning’ Model That ‘explicitly Shows Its Thoughts’

Google has introduced a caller AI “reasoning” exemplary tin of answering analyzable questions while too providing a rundown of its “thoughts,” arsenic reported earlier by TechCrunch. The model, called Gemini 2.0 Flash Thinking, is still experimental and will apt compete pinch OpenAI’s o1 reasoning model.

In a position connected X, Google DeepMind main intelligence Jeff Dean says nan exemplary is “trained to usage thoughts to fortify its reasoning,” and too benefits from nan velocity that comes connected pinch nan faster Gemini Flash 2.0 model. The demo shared by Dean shows really Gemini 2.0 Flash Thinking goes astir answering a physics problem by “thinking” done a bid of steps earlier offering a solution.

This isn’t needfully “reasoning” successful nan measurement humans execute it, but it intends nan instrumentality breaks down instructions into smaller tasks that tin nutrient stronger outcomes.

Another example, posted by Google merchandise lead Logan Kilpatrick, shows nan exemplary reasoning its measurement done a problem that involves immoderate ocular and textual elements. “This is conscionable nan first measurement successful our reasoning journey,” Kilpatrick says. You tin effort retired Gemini 2.0 Flash Thinking connected Google’s AI Studio.

There personification been alternatively a less notable updates successful nan AI abstraction arsenic of late, pinch Google revealing its upgraded Gemini 2.0 exemplary earlier this play arsenic information of nan company’s push into “agentic” AI. Meanwhile, OpenAI made nan afloat type of its o1 reasoning exemplary disposable to ChatGPT subscribers.