Google is working on AI that can reason like humans
Google is developing software for artificial intelligence that will allow it to reason, similar to human intelligence. This development resembles o1 and opens a new front in the competition between the tech giant and rapidly growing startup OpenAI.
According to sources close to the company, several teams at Alphabet Inc.'s Google have made progress in developing software for reasoning based on AI. These programs are more intelligent in solving multi-step tasks in areas such as mathematics and computer programming.
AI researchers are studying reasoning models to approach human thinking. Google is trying to use a technique known as chain-of-thought (chain of thoughts), to achieve this. This technique allows software to consider a series of related clues and then sum up what seems like the best answer.
Google declined to comment on this initiative, but experts believe that the company will remain a powerful player in the field of AI. Oren Etzioni, an experienced AI researcher, noted that "technically, Google's capabilities have always been at the top level. They have simply been more conservative in their exploration."
Google and OpenAI are engaged in a fierce battle for dominance in the field of AI, especially after OpenAI's release of ChatGPT. Google has taken various steps to regain its leadership, including merging its leading research labs into the Google DeepMind division.
In July, Google presented AlphaProof, specializing in mathematical reasoning, and AlphaGeometry 2, an updated version of the model focused on geometry. The models successfully completed four out of six tasks presented at the International Mathematical Olympiad.
At its developer conference in May, Google presented a brief overview of the AI assistant Astra, which can use a smartphone camera to see the surrounding world and answer questions. Google stated that some features of the assistant may appear in its flagship AI model Gemini by the end of the year.
"Expanded mathematical reasoning is a key opportunity for modern artificial intelligence," wrote Google DeepMind's CEO, Demis Hassabis, on social media X in July.