Software Engineering Daily Chip Design in the AI Era with Thomas Andersen
11 snips
May 29, 2025 
Thomas Andersen, Vice President of AI and Machine Learning at Synopsys, dives into the groundbreaking role of AI in chip design. He discusses the intricate nature of combining human creativity with AI capabilities to streamline complex processes. Andersen highlights the emergence of domain-specific large language models and their potential in optimizing chip design. He also addresses the challenges of extracting tacit knowledge from unstructured data, enhancing workflows, and the importance of balancing automation with human oversight in this rapidly evolving industry.
AI Snips
Chapters
Transcript
Episode notes
AI Accelerates Complex Chip Design
- Chip design takes 12 to 18 months because of its complexity and thousands of required engineers.
- AI reduces manual effort and accelerates this timeline but doesn't replace the human creativity required.
Reinforcement Learning for Chip Tuning
- AI automates tuning knobs for chip design performance, saving weeks of human effort.
- Models learn from previous chip versions, building a knowledge base beyond any single engineer's intuition.
AI Complements Verification Steps
- AI solutions in chip design complement existing verification tools by acting as companions, not replacements.
- Automated checks remain essential to ensure outputs are correct despite statistical nature of AI models.
