Industrial AI for the Physical World: Siemens’s Peter Koerte
Summary
Peter Koerte, Chief Strategy and Technology Officer at Siemens, discusses how industrial AI is profoundly transforming critical infrastructure like factories, transportation, energy grids, and buildings, often unseen by consumers. Unlike consumer AI, industrial AI demands near-perfect accuracy (99.9% or higher) due to its connection to the physical world, where errors have severe consequences. Training industrial AI models relies on proprietary, domain-specific data, such as time-series, construction, engineering, and simulation data, which is acquired through data-sharing agreements with customers who see tangible benefits like 30% energy savings in buildings or predicting train door failures 10 days in advance. Koerte emphasizes the critical role of domain expertise in understanding specific industry semantics and key variables, citing Siemens' partnership with Nvidia to accelerate complex engineering simulations from months to minutes, highlighting AI's role in transformation beyond just technology.
Key takeaway
For Directors of AI/ML or VPs of Engineering evaluating AI deployments in industrial settings, recognize that industrial AI's success hinges on achieving extremely high accuracy and integrating deep domain expertise. Prioritize building data-sharing partnerships that clearly demonstrate value to customers, as this is critical for acquiring the proprietary data needed for robust model training. Your strategy should account for AI as 80% transformation and 20% technology, focusing on workflow changes and addressing workforce anxieties to ensure successful adoption and maximize operational acceleration.
Key insights
Industrial AI prioritizes precision and domain-specific data to optimize critical infrastructure, demanding high accuracy and value-driven data sharing.
Principles
- Industrial AI requires near-perfect accuracy.
- Domain expertise is crucial for industrial AI success.
- Value creation drives data sharing in industrial settings.
Method
Industrial AI models are trained using proprietary, domain-specific data (e.g., time-series, CAD) acquired via data-sharing agreements, focusing on specific industry semantics and key variables to ensure high precision and reliability.
In practice
- Implement AI for predictive maintenance (e.g., train doors).
- Optimize building energy use with autonomous AI systems.
- Accelerate engineering simulations using AI-powered tools.
Topics
- Industrial AI
- Predictive Maintenance
- Data Sharing Partnerships
- Digital Transformation
- Worker Augmentation
Best for: Director of AI/ML, VP of Engineering/Data, Consultant
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Me, Myself, and AI.