Another day, another development in the field of artificial intelligence (AI). Yes, the tech giant Google has teased about a new AI model capable of reasoning! Yes, you heard it right. The new ‘reasoning’ AI model has the potential to reason and explain the answers provided to your queries.
In an official blog, Google explained about its new “reasoning” AI model. However, as of now it’s in the experimental stages. Let’s have a look at what the new AI model brings to the table.
Google brings ‘reasoning AI’model
The new model is named ‘Gemini 2.0 Flash Thinking Experimental.’ The AI model is available in AI Studio, which is Google’s AI prototyping platform. Google describes their new AI model as the best tool for reasoning, multimodal understanding and coding/ It also added that it comes with the ability to ‘reason over the most complex problems.’ This is applicable for fields such as math, programming and physics.
Logan Kilpatrick, who leads product for AI Studio, in a post on X (earlier Twitter), highlighted Gemini 2.0 Flash Thinking Experimental as the first step in [Google’s] reasoning journey. Also, Jeff Dean, who is chief scientist for Google DeepMind, Google’s AI research division, promoted the product. He also made a X post, highlighting that the new Gemini 2.0 reasoning model is “trained to use thoughts to strengthen its reasoning.”
The way ahead
Looking at the future impact of the new AI model, “We see promising results when we increase inference time computation,” Dean explained. Looks like he was referring to the amount of computing which is used to “run” the model as it picks a question to answer.
The new model is built on Google’s recently announced Gemini 2.0 Flash model. So, the Gemini 2.0 Flash Thinking Experimental could be similar. So, how does the model actually work? The model can pause before responding. It tries to consider a number of related prompts. This helps it to explain its reasoning throughout the conversation with its users. After some time the model then summarizes what it considers to be the most accurate answer. Its like having a conversation with some subject expert.
Follow FE Tech Bytes on Twitter, Instagram, LinkedIn, Facebook