Google presents a new AI capable of reasoning

Google has announced Gemini 2.0 Flash Thinking, a new “reasoning model” powered by artificial intelligence (AI). The proposal is in an experimental phase, but it is not ruled out that it will be launched on the market in the future to compete live with OpenAI o1.

Jeff Dean, chief scientist at Google DeepMind, describes the model as a tool that “explicitly shows your thoughts.” The system is trained to break down complex requests into more manageable parts, allowing it to “think” before generating a response. This feature significantly improves your analysis and processing capabilities.

Google has shared few details about the internal workings of the model. However, it is known that it is built on the basis of the Gemini 2.0 Flash multimodal model, known for its speed and high performance. Logan Kilpatrick, product division leader at the company, notes that “this is just the first step in our thinking process.” Gemini 2.0 Flash Thinking is available for free through the Gemini API, which is can be accessed from the Google AI Studio and Vertex AI platforms.

The presentation of this model highlights the ambitions of the Mountain View technology company to lead the emerging and increasingly competitive artificial intelligence market. A report of The Informationpublished in October, revealed that more than 200 researchers at the corporation are dedicated to developing computer programs with advanced reasoning capabilities.

Google increases its bet on AI

Google has expanded its portfolio of AI solutions with innovations that seek to rival OpenAI’s flagship products. This week it launched Veo 2, an engine capable of generating videos of up to 2 minutes in 4K resolution from text instructions, in obvious competition with Sora. It also presented an update to its Image 3 graphics generator, which now offers greater precision and a new function called Whisk, designed to create images based on others proposed by users.

The big tech They are focusing their efforts on developing AI models with more advanced reasoning capabilities. These technologies stand out for their verification capacity, which improves the accuracy of responses. However, they face challenges related to scalability due to the high computational requirements they imply.

Analysts suggest that Gemini 2.0 Flash Thinking will need to balance its reasoning ability with greater operational efficiency to compete effectively. This will be crucial to face OpenAI o1, whose leadership has been consolidated with its integration into the new ChatGPT Pro subscription plan.