Google introduced the first AI model that reasons “out loud”

Google has unveiled a new experimental artificial intelligence model, Gemini 2.0 Flash Thinking, which is likely to compete with OpenAI’s o1 AI model. She is also capable of thinking and answering complex questions, but in addition to this, she can also explain the process of her “thoughts,” The Verge reports.

Image source: Google

Google DeepMind chief scientist Jeff Dean noted in his X post that the model is trained to “think” to enhance logical reasoning, and also has high processing speed thanks to updated Gemini Flash 2.0 technology. One example shows how a model solves a physics problem by sequentially “thinking” step by step before producing the correct solution. And although such a process is not analogous to human thinking, it allows the algorithm to break complex tasks into smaller ones, which leads to more accurate and reliable results.

Another example of Gemini 2.0 Flash Thinking, shared by Google product manager Logan Kilpatrick on his X account, showed how the model solved a problem that combined text and visual elements.

It is worth saying that recent weeks have been rich in news in the field of AI. Google recently unveiled an updated version of its Gemini 2.0 model, highlighting its commitment to developing agent-based AI, and OpenAI released an exclusive version of the o1 model for ChatGPT subscribers for $200, which can reason.