From 15 hours to one minute: How AI/ML is speeding up GM's development
Summary
General Motors (GM) is transitioning its engineering and design processes into a "third epoch" by extensively integrating AI/machine learning, significantly accelerating development cycles. Sterling Anderson, GM's chief product officer, highlights this shift from empirical iteration and sequential virtual tools to a collapsed, probabilistic design method. For instance, traditional Finite Element Analysis (FEA) runs that previously took 15 hours now complete in one minute, enabling engineers to perform a much broader set of tests. Similarly, crash performance simulations, which typically required 15 to 18 hours, are now reduced to under one minute, allowing for rapid optimization of structural performance. This virtualization extends beyond core vehicle design to areas like HVAC system optimization, factory digital twins, motorsport, energy, defense, and GM's lunar program, fostering simultaneous hardware and software optimization and faster iteration across diverse applications.
Key takeaway
For engineering leads aiming to drastically accelerate product development and system optimization, GM's adoption of AI/ML for virtual integration offers a compelling blueprint. You should explore probabilistic methods to collapse sequential design functions, enabling simultaneous hardware and software optimization. This approach can reduce complex simulation times from hours to minutes, allowing your teams to iterate thousands of designs and harden products against real-world conditions much faster, freeing engineers for deeper innovation.
Key insights
AI/ML collapses sequential engineering functions into rapid, probabilistic virtual optimization, drastically accelerating design and development cycles.
Principles
- Collapse sequential design functions into probabilistic methods.
- Simultaneously optimize hardware and software virtually.
- Iterate broadly to harden designs against real-world conditions.
Method
AI/ML virtualizes complex engineering analyses (FEA, crash tests), enabling parallel execution and rapid iteration of physical and software parameters within a single virtual environment for simultaneous optimization.
In practice
- Reduce FEA run times from 15 hours to one minute.
- Cut crash simulation times from 18 hours to under one minute.
- Optimize HVAC systems in days, not months.
Topics
- AI/ML Engineering
- Virtual Integration
- Finite Element Analysis
- Digital Twins
- Product Development Acceleration
- Automotive Design
Best for: Executive, AI Engineer, Machine Learning Engineer, Director of AI/ML
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by AI - Ars Technica.