OpenAI has taken another step in the artificial intelligence (AI) arms race.
The preeminent generative AI company recently introduced OpenAI o3, a system capable of “reasoning” math, science and computer programming problems. The new system is a successor to o1, a reasoning system that debuted earlier in 2024 and is now available to businesses and individuals; o3 is more than 20% more accurate than o1 and is currently being tested by safety and security experts.
“This model is incredible at programming,” OpenAI CEO Sam Altman said in an online presentation that acted as an introduction of o3. Based on the same core technology as the original ChatGPT, o3 is designed to think through problems by deconstructing a problem into segments and then logically taking steps to solve it. However, like ChatGPT, o3 can hallucinate and get things wrong — and the new system requires larger amounts of computing power that can be expensive.
The new system, built on what OpenAI calls “reinforcement learning,” carried out a series of common programming tasks and even out-performed the company’s chief scientist, Jakub Pachocki, on a programming test.
Ultimately, OpenAI’s latest chatbot illustrates a larger effort by the company and the industry in general to build AI systems that can reason through complex tasks, leading to technology that can be used by computer programmers who rely on AI to write code. To that end, Google last month unveiled Gemini 2.0 Flash Thinking Experimental, which boasts similar capabilities to o3.
The AI reasoning race — particularly between OpenAI (backed by Microsoft Corp.), Google and Anthropic (backed by Amazon.com Inc.) — is among the crucial factors in determining which company establishes itself as an early leader in a potentially multitrillion-dollar market.
Indeed, OpenAI last month announced its plan to restructure as a public benefit corporation (P.B.C.) from nonprofit status.
“The P.B.C. is a structure used by many others that requires the company to balance shareholder interests, stakeholder interests and a public benefit interest in its decision making,” OpenAI said in a blog post. “It will enable us to raise the necessary capital with conventional terms like our competitors.”
The plan “would result in one of the best resourced nonprofits in history,” OpenAI said in the blog.
OpenAI’s corporate transformation comes a year after the company board unsuccessfully tried to fire Altman. Subsequently, Altman helped lead a plan to pave OpenAI to for-profit status.
