TSMC Executive Vice President and Co-Chief Operating Officer Y.J. Mii said artificial intelligence remains highly capable but immature, comparing it with a "3-year-old Superman" at a forum at National Taiwan University on the 15th. AI has delivered gains in programming and electronic design automation (EDA), where inputs, outputs and datasets are well defined, but Mii said it remains unreliable for semiconductor process research involving unknown parameters, real-world materials and high-stakes tradeoffs. TSMC is therefore permitting uses such as coding assistance and EDA while remaining cautious about applying AI to the research and process-development work behind its technological edge. Sensitive R&D data, including manufacturing parameters and process recipes, could be exposed when AI tools use external systems. The caution comes as TSMC expands manufacturing capacity for AI chips used by Nvidia, Apple and other customers, suggesting demand remains strong even as productivity gains in proprietary R&D may take longer to emerge. Mii also said advanced semiconductor development has offered no established answers to copy since the industry moved beyond the 7nm process node, making human judgment, rapid debugging, commercial evaluation and cross-functional discussion essential. He discussed his move from IBM Research to wafer-fab work at TSMC, overseas hiring and management challenges in Arizona and Kumamoto, and the importance of humanities, communication and people management in leadership. Leaders at OpenAI and Anthropic have also called for measured AI development amid concerns about unexpected system behavior.