Google responded to o1
Google’s artificial intelligence prototype platform AI StudioThe model, available in , is described as one of the best in “multimodal understanding, reasoning, and coding.” Gemini 2.0 Flash Thinking Experimental is much more capable than the company’s other models in producing solutions to complex problems such as programming, mathematics and physics.
Chief scientist of Google DeepMind Jeff Dean, In his own post, he said that Gemini 2.0 Flash Thinking Experimental was “trained to use thoughts to strengthen his judgment.” Unlike traditional artificial intelligence, the model pauses before answering a question, considers multiple relevant scenarios, and explains its reasoning process. At the end of this process the most correct answer aims to summarize. However, it has been observed that the model gives incorrect answers in some cases.
On the other hand, reasoning models enable artificial intelligence to work with a self-verification mechanism. This feature aims to prevent mistakes that artificial intelligence frequently makes. For example, problems such as generating misinformation or misunderstanding the context are reduced in these models. But these innovations come with a price. Since reasoning models require more processing power to reach a solution, the process of producing results can take from a few seconds to minutes.
This news our mobile application Download using
You can read it whenever you want (even offline):
Source link: https://www.donanimhaber.com/google-openai-o1-rakibi-kendi-dusunen-yapay-zekasini-duyurdu–185600