Google Said to Be Working on Reasoning AI, Chasing OpenAI's Efforts

Google is building artificial intelligence software to simulate the functioning of the human mind in regards to reasoning capabilities, like its version of OpenAI’s o1, marking a new battleground in the fight between the tech goliath and the rapidly growing start-up.

In recent months, Alphabet Inc.’s Google teams have enhanced its AI reasoning software, people knowledgeable about the subject and not authorized to speak publicly said, citing private nature of the matter. These programs now solve problems with more ease than they did before, including multi-step problems in disciplines such as mathematics and computer programming.

AI researchers are chasing reasoning models as they seek the next breakthrough in the technology. Like OpenAI, Google is trying to approximate human reasoning using a technique known as chain-of-thought prompting, according to two of the people.”. This technique – Google pioneered it – pauses for a matter of seconds before coming out with a written response to a question while, invisible to the user, behind the scenes it explores a range of related questions and then summarizes what appears to be the best response.

Google would not comment on the effort

The battle between Google and OpenAI has become very hot since its popular chatbot, ChatGPT, came out. Many investors fear that this chatbot will eventually make the use of Google search obsolete.

Google has been doing its best to regain the top spot through actions such as merging its premier research labs under a unit called Google DeepMind and further strengthening relationships between researchers and product teams. Still, it takes Google, the search giant, much longer to roll out AI products and pauses for thought on ethical problems arising in the course of an action, living up to its brand owned by the public that expects trust, besides juggling competing interests of a multi-faceted effort within a vast organization. Since OpenAI made public its o1 model, internally codenamed Strawberry, in mid-September, some people around DeepMind had feared the company may have slipped behind, said another person familiar with the situation. Employees don’t worry as much now as they did after the launch of ChatGPT, that person said, since Google has just showcased some of its own work.

Despite the slower rollout, said Oren Etzioni, a seasoned AI researcher and founder of TrueMedia.org, a nonprofit attempting to combat political disinformation, “Google is a big dog.”

“Technically it’s always been the case that Google’s capabilities were top-notch. They were just more conservative in rolling things out,” Etzioni said. “It’s a marathon, and it’s anybody’s race to win.”

In July, Google demoed AlphaProof, which specializes in mathematics reasoning, and AlphaGeometry 2, an updated version of a model focused on geometry that the company debuted early this year. The programs aced four of the six problems featured in the International Mathematical Olympiad, an annual competition in which students tackle topics such as algebra and geometry, Google said in a blog post.

At its developer conference in May, Google shared its first peek at an AI assistant named Astra, which can use a phone’s camera to see the world around it and answer questions, like telling a user where she had left her glasses. Google said that some features of assistant may come to its flagship AI model, Gemini, toward the end of this year.

“Advanced mathematical reasoning is a critical capability for modern AI,” Google DeepMind CEO Demis Hassabis wrote in a post on social network X in July.