Son Says OpenAI Uses AI to Design Next Model as He Sees Superintelligence Within 2 Years
Updated
Updated · CNBC · Jun 5
Son Says OpenAI Uses AI to Design Next Model as He Sees Superintelligence Within 2 Years
3 articles · Updated · CNBC · Jun 5
Summary
OpenAI engineers told Masayoshi Son that one AI model is designing a future model, which the SoftBank CEO said shows AI is nearing “super intelligence.”
2 years is now Son’s timeline for artificial superintelligence, down from the 10 years he cited in 2024 and shorter than the four years he said he privately expected.
OpenAI declined to discuss unreleased models but pointed to February’s GPT-5.3-Codex, which it said helped debug its own training, deployment and evaluations.
Anthropic thrust the issue into focus Thursday, warning that recursive self-improvement could raise the risk of humans losing control and arguing AI labs may need to slow development.
SoftBank’s stance matters because Son has placed the group at the center of the AI boom through Arm, OpenAI and bets on robotics and autonomous driving.
As AI begins designing its successors, are calls for a global pause a vital safeguard or a strategic ploy in a high-stakes tech race?
If superintelligence is two years away, what governance plans exist beyond corporate papers to manage this unprecedented global power shift?
Recursive Self-Improvement in AI: OpenAI’s Autonomous Model Development, Superintelligence Timelines, and the Global Governance Challenge
Overview
OpenAI is now using its own AI models to help design and improve the next generation of AI systems, marking a major shift from traditional human-led development. This approach, known as recursive self-improvement, allows AI to identify efficiencies, optimize its own architecture, and even generate code for future models, creating a feedback loop for rapid progress. As AI becomes more involved in its own evolution, the pace of innovation is expected to accelerate, raising both exciting possibilities and important questions about safety, control, and the future role of humans in AI development.