news

google is working on developing artificial intelligence software with human reasoning capabilities, similar to openai's recently launched o1, according to people familiar with the matter, a move that heralds a new stage in the competition between the technology giant and the rapidly rising openai .

people familiar with the matter revealed that in recent months, multiple teams under google have made significant progress in the field of artificial intelligence reasoning software. this type of software has demonstrated excellent capabilities in handling multi-step problems such as mathematical puzzles and computer programming.

airesearchers in the field are actively exploring inference models in an effort to drive the next big leap forward in this technology. google is following suit, according to two people familiar with the matterOpenAI, trying to use a technology called "chain-of-thought prompt" to simulate the human reasoning process.

in the technology, pioneered by google, the software pauses for a moment before responding to a written prompt. unbeknownst to the user, it quickly analyzes multiple related prompts and combines them into what appears to be the best solution. good answer.

google declined to comment on this research and development effort.

the competition between google and openai in the field of artificial intelligence has become increasingly fierce, especially after openai released its chatbot chatgpt. this trend has become more obvious. some investors have expressed concerns that the popular chatbot could undermine demand for google searches.

to regain its leading position, google has taken a series of initiatives, including integrating its core research laboratories into the google deepmind unit and strengthening collaboration between researchers and product teams.

however, the search giant is obviously more cautious when it comes to releasing artificial intelligence products. they need to weigh ethical issues, meet public expectations for trust in their brand, and deal with competition among multiple similar projects within the organization.

since openai launched its o1 model (internally codenamed "strawberry") in mid-september, some at deepmind have worried that the company may have fallen behind, according to another person familiar with the matter. but the person familiar with the matter also said that as google has launched a series of products, employees' anxiety has eased.

senior artificial intelligence researcher oren etzioni noted that despite google's relatively slow release of products, it remains a strong contender in the field. he added: "technically, google has always been at the top. they've just been conservative with product launches. it's a marathon and everyone wants to win."

in july this year, google launched two artificial intelligence programs focused on mathematical reasoning-alphaproof and alphageometry 2, the latter is an upgraded version of the geometric model released by the company earlier this year. in a blog post, google revealed that the programs successfully solved four difficult problems in the international mathematical olympiad, an annual competition that challenges students to solve complex problems in algebra, geometry and other fields.

at its developer conference in may this year, google showed off an artificial intelligence assistant called astra. the assistant uses the phone's camera to see its surroundings and answer questions like "where are my glasses?" google revealed that some of astra’s features are expected to be integrated into its flagship artificial intelligence model gemini by the end of the year.

demis hassabis, ceo of google deepmind, emphasized on social media in july this year: "advanced mathematical reasoning has become an indispensable key capability for modern artificial intelligence." (specially compiled by tencent technology /golden deer)