SoftBank CEO Masayoshi Son Predicts ASI Arrival Within 2 Years

Artificial Super Intelligence (ASI) is no longer a distant theoretical concept but an imminent reality. SoftBank CEO Masayoshi Son now predicts this milestone will arrive within just 2 years, drastically shortening his previous 10-year estimate.

This acceleration is driven by a critical technological shift: AI systems designing their own successors. Son revealed that OpenAI is actively using AI to design and develop future models, creating a recursive improvement loop.

The Shift to Self-Designing AI Models

The core of Son's prediction lies in the methodology used by leading AI laboratories. He stated that OpenAI is advancing "AI designing AI," where artificial intelligence participates in the creation of subsequent models.

Son engaged in detailed discussions with Sam Altman and multiple OpenAI engineers to understand this process. They confirmed that current AI models are capable of designing future iterations, a capability that human engineers cannot match in speed or complexity.

This approach marks a fundamental departure from traditional software development. Human engineers previously wrote code and designed architectures manually. Now, AI agents optimize these structures autonomously.

Codex: The First Self-Creating Model

OpenAI highlighted this capability earlier this year with its GPT-5.3-Codex model. This specific iteration was described as the first model to "participate in creating itself."

The Codex team utilized early versions of the model for critical tasks. These included debugging training processes, managing deployments, and diagnosing test results.

Such automation allows for rapid iteration cycles. Errors are identified and corrected by the AI itself before human review. This reduces the bottleneck of human oversight significantly.

Son believes this pattern will expand to other mainstream models soon. Human engineers will increasingly struggle to design stronger models independently. The complexity of neural networks has surpassed human cognitive capacity for manual optimization.

Redefining the Timeline for Superintelligence

Son's timeline adjustment represents one of the most aggressive forecasts in the tech industry. In 2024, he predicted ASI would emerge within 10 years. He now considers that view overly conservative.

He defines ASI as artificial intelligence that is 10,000 times smarter than humans. This metric goes beyond simple processing speed to include reasoning, creativity, and problem-solving capabilities across all domains.

The reduction from 10 years to 2 years suggests exponential growth in AI capabilities. This aligns with the concept of an intelligence explosion, where each generation of AI improves the next at an accelerating rate.

• Previous Prediction: ASI arrival within 10 years (announced in 2024)
• Current Prediction: ASI arrival within 2 years (announced in recent CNBC interview)
• Definition of ASI: AI possessing 10,000x human intelligence levels
• Key Driver: Recursive self-improvement via AI-designed models
• Primary Example: OpenAI's GPT-5.3-Codex self-development capabilities
• Expert Insight: Direct conversations with Sam Altman and engineering teams

Personal Adoption and Daily Reliance

Son's confidence in AI stems from his personal usage habits. He disclosed that he uses the ChatGPT chatbot for 2 to 3 hours daily. This extensive interaction provides him with firsthand insight into the technology's evolving capabilities.

He openly admits that the AI is smarter than him in most subjects. This humility highlights the shifting dynamic between human experts and machine intelligence. It is not merely a tool but a superior intellectual partner in many contexts.

This level of reliance among top executives signals broader market trends. If leaders of major conglomerates depend on AI for strategic insights, enterprise adoption will accelerate rapidly.

The integration of AI into daily workflows is becoming standard. Tasks that once required hours of research are now completed in seconds. This efficiency gain drives the economic value proposition of AI investments.

Industry Context and Competitive Landscape

The race for ASI is intensifying among Western tech giants. While OpenAI leads in consumer-facing models, competitors like Google DeepMind and Anthropic are pursuing similar recursive training methods.

Microsoft and NVIDIA are also heavily invested in this infrastructure. Their chips and cloud services enable the massive computational power required for self-designing AI systems.

Regulatory bodies in the US and Europe are watching closely. The rapid approach of ASI raises urgent questions about safety, alignment, and control. Governments may need to intervene sooner than anticipated.

Investment flows reflect this urgency. Venture capital and corporate spending on AI infrastructure have reached record levels. The potential for ASI promises unprecedented economic returns, driving risk-taking behavior.

What This Means for Developers and Businesses

For developers, the role of coding is changing fundamentally. Writing boilerplate code will become obsolete as AI handles implementation details. Focus must shift to architecture, logic, and ethical oversight.

Businesses must prepare for operational transformation. Processes reliant on human cognition will be automated. Companies that fail to integrate AI-driven design risks falling behind competitors.

• Adopt AI-First Workflows: Integrate LLMs into every stage of product development immediately.
• Prioritize Data Quality: Self-designing AI requires high-quality, unbiased training data to avoid compounding errors.
• Focus on High-Level Strategy: Shift human resources from execution to strategic oversight and creative direction.
• Monitor Regulatory Changes: Stay updated on AI safety laws in the EU and US to ensure compliance.
• Invest in Compute Infrastructure: Secure access to powerful GPUs/TPUs needed for running advanced models.
• Ethical Alignment Protocols: Establish strict guidelines for AI decision-making to prevent unintended consequences.

Looking Ahead: The Next 24 Months

The next two years will define the technological landscape. We can expect rapid advancements in model autonomy and reasoning capabilities.

Human-AI collaboration will evolve into AI-led innovation. Humans will act as validators rather than creators. This shift requires new educational frameworks and workforce training programs.

The societal impact will be profound. Job markets will adjust to accommodate superintelligent systems. Policymakers must address displacement and inequality proactively.

Gogo's Take

• 🔥 Why This Matters: This prediction validates the 'recursive self-improvement' thesis, suggesting we are entering a phase where AI progress becomes autonomous. For businesses, this means the competitive window to adopt AI is closing fast; waiting another year could mean falling behind by a decade in capability.
• ⚠️ Limitations & Risks: The primary risk is the 'black box' problem. If AI designs AI without full human transparency, understanding failure modes becomes nearly impossible. Additionally, a 2-year timeline leaves minimal room for regulatory safety nets, increasing the risk of misaligned or unstable systems reaching production.
• 💡 Actionable Advice: Do not wait for ASI to arrive. Start auditing your current AI dependencies today. Implement rigorous 'human-in-the-loop' validation for any AI-generated code or strategy. Invest in tools that provide explainability and audit trails for AI decisions, ensuring you maintain control even as systems become more autonomous.