TSMC executive Y. J. Mii says AI can be extremely powerful while still lacking mature judgment. Speaking with National Taiwan University students, he argued that final decisions should stay with people. TSMC already uses AI heavily for coding, where simpler tasks let current tools perform well.
Chip design gets similar treatment. Mii said EDA works well with AI because design variables are defined and results can be checked. Optimizing huge transistor counts fits that setup. But leading-edge process research brings far more unknowns.
For A14 at 1.4 nanometers and future A12 work below 1.4 nanometers, missing data limits what AI can do. New materials and equipment may fall short, while AI has no proven information to fill those gaps.
And company data creates another limit. Mii warned that sharing sensitive information with AI could raise leak risks. For TSMC’s overseas expansion, he pointed to another hard problem: finding senior managers who can lead global teams.
Chip design gets similar treatment. Mii said EDA works well with AI because design variables are defined and results can be checked. Optimizing huge transistor counts fits that setup. But leading-edge process research brings far more unknowns.
For A14 at 1.4 nanometers and future A12 work below 1.4 nanometers, missing data limits what AI can do. New materials and equipment may fall short, while AI has no proven information to fill those gaps.
And company data creates another limit. Mii warned that sharing sensitive information with AI could raise leak risks. For TSMC’s overseas expansion, he pointed to another hard problem: finding senior managers who can lead global teams.
