AI is designing OpenAI's next model in a sign of 'super intelligence': SoftBank's Masayoshi Son

Masayoshi Son, the CEO of SoftBank, has made headlines by asserting that OpenAI's forthcoming model is being crafted by an existing AI, suggesting a move towards so-called 'super intelligence.' In a recent interview with CNBC, Son discussed his conversations with OpenAI's CEO Sam Altman and the company's engineers, who revealed that an AI model is currently engaged in designing its successor. Son expressed his belief that this trend will extend to other major AI models, indicating that human engineers may soon be outpaced by the capabilities of AI. He claimed, "Once that happens, the model generates the next model... and it's going to be exponentially smarter than all of us. That's a super intelligence." While OpenAI declined to provide specifics on unreleased models, they did confirm their ongoing use of AI in model development processes. Notably, earlier this year, OpenAI announced that its GPT-5.3-Codex was pivotal in its own creation, utilizing early versions of the model for debugging and deployment tasks. Son’s remarks come amid a larger discourse on artificial super intelligence (ASI), a concept he previously defined as AI that could surpass human intelligence by a factor of 10,000. He initially projected that ASI would emerge within a decade but recently revised that estimate, stating, "I thought it was coming in four years instead of 10 years. Now, I say it's coming in the next two years." The SoftBank leader also shared his personal experience with AI, noting that he utilizes OpenAI's ChatGPT for two to three hours daily, claiming it outperforms him in most subjects. He predicts that within a few years, AI could surpass human intelligence in 70% to 80% of topics, potentially being tenfold smarter than average individuals in those areas. Son has been a staunch advocate for AI, positioning SoftBank strategically within the tech sector through investments in OpenAI, chip design with Arm, and advancements in robotics and autonomous vehicles. He has characterized the ongoing AI revolution as 50 times more significant than the dot-com boom. The potential risks associated with advanced AI systems were highlighted recently when Anthropic published a blog discussing 'recursive self-improvement' (RSI), a phenomenon where AI can autonomously design and improve itself. While Anthropic acknowledged the positive implications, they cautioned about the risks of humans losing control over such systems, advocating for a collaborative effort among AI labs to slow development for safety. While it remains uncertain if Son's comments directly relate to RSI, an OpenAI research paper from June indicated that early signs of RSI are already present in current systems, warning of the competitive and governance challenges that may arise as these technologies evolve. The paper emphasized the need for societal frameworks to guide AI development in alignment with human interests.