While much of the AI race in recent times has focused on making models increasingly smarter in reasoning, OpenAI CEO Sam altman offers a different perspective. According to him, the factor that will create a real turning point for AI in the future is not logic, but memory.
Sharing in a recent podcast with technology journalist Alex Kantrowitz, Sam altman said that AI companies may have focused too much on the reasoning ability of large language models.
The next big step forward will come when AI systems have something almost like unlimited memory, which is the ability to remember users, preferences and all long-term interactions.
altman admits that ChatGPT has made significant progress in recent years, from accuracy to natural dialogue capability.
However, he also emphasized that memory is the factor that makes the real difference.
With memory, AI is no longer a tool to respond to disjointed questions, but can become a system that understands users in depth and persistently over time.
Currently, ChatGPT's remembering feature allows the chatbot to store information shared by users between conversations, helping to reduce the need to repeat familiar details.
According to altman, this is just the beginning. In the future, AI memory can develop to the point of remembering every detail in your entire life, from big interests to small habits.
Compared to humans, altman said that even the best personal assistant cannot remember everything that others have said.
In contrast, AI is not biologically limited, so it can theoretically possess unlimited memory. This opens up the prospect of AI systems being increasingly personalized, and can even create a feeling of long-term companionship with humans.
However, such extensive storage also raises serious concerns about privacy and data security. The fact that an AI system holds a large amount of sensitive personal information raises questions about control, security and limitation of use.
However, Mr. altman believes that if managed properly, humans can still accept and build a relationship of trust with AI.
In that context, OpenAI is also under increasing competitive pressure. Google recently attracted attention with Gemini 3, which was rated as exceeding expectations.
In response, Sam Altman declared a red alert early this month, suggesting the company is focusing on a new large language model, code-named Garlic.
These moves show that OpenAI is not only racing for performance, but is looking to reshape the future of AI, where memory, not just reasoning, but according to Sam Altman, will become the foundation for the journey towards general artificial intelligence (AGI).
