Changes are being detected in the ecosystem of OpenAI, the developer of ChatGPT. The company recently revealed the accident process of the newly launched "o3 mini" to users, and allowed the use of "ChatGPT Search" without requiring membership registration. This move by OpenAI is interpreted as a response to the shock from the Chinese artificial intelligence (AI) startup DeepSeek, which amazed the world with its innovation in expense and performance.
On the 7th, according to foreign media including TechCrunch, OpenAI confirmed that users can verify the specific "chain of thought" feature of the latest AI model, "o3 mini." Previously, the model's thought process was provided only in a summarized form, but this update has disclosed a more detailed thought process. A spokesperson for OpenAI said, "The current 'chain of thought' of 'o3 mini' makes it easier to understand how the model thinks," adding, "Through this, users will be able to grasp the AI's logic and increase the credibility of the answers."
This appears to be a measure aimed at responding to the recent DeepSeek shock. OpenAI boasts that "o3 mini" derives fast reasoning speeds with minimal expense, suggesting that it is targeting DeepSeek's AI model, which is celebrated for its cost-effectiveness. In particular, DeepSeek's AI model, "R1," reveals the entire thought process to users, which is regarded among AI researchers as a way to enhance reliability and transparency. While OpenAI previously provided only a brief overview of some thought processes in the "o1" and "o1 mini" models, this opportunity signifies a relaxation of such policies.
Additionally, OpenAI announced on the 5th (local time) that it would open access to "ChatGPT Search" without the need for membership registration. "ChatGPT Search" is an interactive search engine. Similar to standard ChatGPT, users can ask questions in natural language, much like chatting with a friend, and the AI chatbot finds and provides suitable information from the internet in real time. OpenAI launched the ChatGPT search feature for paid subscribers in October last year, and two months later expanded access to free members, eventually opening its doors to those who are not signed up for ChatGPT.
Until now, OpenAI has maintained a closed operating style under the pretext of AI safety. This has drawn criticism for diverging from its original promise to disclose AI research-related data in the name of benefiting humanity.
Sam Altman, OpenAI's Chief Executive Officer (CEO), hinted last month at the possibility of transitioning the company’s AI models to open source. In response to a user's question on whether the company would consider revealing some of ChatGPT's technology on the U.S. online community Reddit on the 31st of last month, he replied, "We also think we need to look for an open-source strategy." He noted, "Not all OpenAI employees think this way, and it is not the highest priority," but added, "I believe we have historically taken the wrong path."