The article discusses recent advancements in AI-designed chips, showcasing how these chips have become so complex that their designs are not fully comprehensible to human engineers. It references the work of Adrian Thompson, who used evolutionary algorithms in the 90s to create highly efficient circuits. This shines a light on the potential of deep learning in chip design. While past successes with evolutionary algorithms exist, this new wave of AI designs presents unique challenges, especially regarding validation and real-world manufacturing. The conversation raises important questions about the definition of AI in engineering and the innovative potential versus regulatory constraints human designers face.