Recently, Masayoshi Son, CEO of SoftBank Group, made a shocking prediction that sent shockwaves through the industry: the deployment of super artificial intelligence (ASI) will be significantly accelerated, with its official arrival expected within the next two years.

Notably, this is the second time this month that Son has revised his estimate for the breakthrough point of this technology. Previously, he had adjusted his judgment several times, initially estimating it would take ten years, then shortening it to four years, and now confidently setting this critical technological threshold within two years.

This bold prediction stems from a fundamental transformation in the underlying R&D paradigm of AI. Son revealed that after in-depth discussions with the OpenAI team, he learned that the AI self-iteration closed loop has been initially completed. This means that existing AI large models have begun to move away from absolute dependence on human programmers, and can independently design and train the next generation of new models. With the start of this "recursive self-iteration" process, the intelligence level of AI will no longer increase linearly, but instead show an alarming exponential explosion, ultimately surpassing human research limits entirely.

Latest data continuously emerging from the industrial sector is strongly supporting this judgment. Currently, multiple leading global technology institutions have already detected related technical prototypes within their laboratories. The intuitive data released by competitor company Anthropic shows that over 80% of the code on its platform is now completely written by Claude AI autonomously. AI has already evolved from an auxiliary tool into a core R&D force. At the same time, OpenAI has also confirmed that its GPT-5.3-Codex model has achieved substantial self-iteration, capable of independently completing the entire R&D workflow, from training diagnosis and deployment monitoring to final result evaluation.

However, this technological singularity, driven by AI itself, has triggered collective anxiety across the tech community and even global regulators. In its latest published academic paper, OpenAI issued a public warning, stating that the current rapid evolution characteristics of artificial intelligence—“self-reproduction”—have far exceeded human society's expectations. The current global social system, laws, and regulatory mechanisms are not yet prepared for the arrival of super intelligence. Facing the closing self-iteration loop of AI, how to establish a complete supporting control and guidance mechanism to prevent technology from plunging into uncontrollable depths has become an urgent common challenge for the global tech community.