🤖 Future of Automation: Large World Models and Software-Defined Approaches 💡

[Süeda Asil]

The world of industrial automation is rapidly evolving, and giants like Schneider Electric, at the forefront of this transformation, are shaping the technologies of the future. Recently, at the Automate fair in Chicago, we met with Andre Marino, Senior Vice President of Industrial Automation at Schneider Electric, to discuss the hottest topics in the industry. This conversation, ranging from software-defined automation to physical AI and large world models, offers striking insights into the near future of automation.

─────────────────────────

🔓 The Rise of Open Source and Software-Defined Automation 🚀​

Marino began our conversation by emphasizing the importance of the word "open." At the core of software-defined automation lies the separation of hardware and software. This separation offers end-users more freedom, eliminating the problem of being locked into a single vendor. According to Marino, closed systems are no longer sustainable, and in the future, automation will become much more open.

So, what does this openness mean? Marino highlights the role of artificial intelligence (AI) in this transformation. For example, code translations between different automation platforms are now much easier thanks to AI. This provides a significant solution to the vendor lock-in problem that the industry has long experienced.

─────────────────────────

🎯 Expanding the Talent Pool: Automation with Modern IT Languages 💻​

The industrial automation sector is experiencing a significant labor shortage. However, software-defined and open-source approaches can solve this problem. Marino states that new talent capable of programming systems using modern IT languages can be attracted to the sector. Especially for young professionals eager to work in physical artificial intelligence (physical AI), these new technologies create a major attraction.

Schneider Electric's current work is based on the concept of decentralized controls. AI is integrated as close as possible to where physical processes occur, providing intelligence at the unit level, and then orchestrated across the entire plant.

─────────────────────────

📊 The Catalyst Role of Data Centers and Energy Consumption ⚡​

Marino emphasizes the critical role of data centers in these initiatives. Data centers, indispensable for running AI models, also stand out for their high energy consumption. This situation creates the need to develop new solutions for energy efficiency. Data centers act as a catalyst, bringing AI technologies to the industry and accelerating their adoption in manufacturing facilities.

─────────────────────────

🌍 Large World Models: The Future of Spatial Understanding 🧠​

One of the most interesting points of the conversation was the concept of large world models. Marino states that these models are vital for the orchestration of autonomously operating agents in the future. However, he emphasizes that these models are different from large language models.

According to Marino, large world models are based on a "spatial problem." For humanoid robots or autonomous vehicles to understand their environment, language alone is not enough; they need to understand 3D space, physical interactions, and their consequences. He clearly illustrates this difference by saying, "When I push someone, I know the consequences. Language cannot understand consequences."

Therefore, physical artificial intelligence aims to create models that develop spatial understanding and can simulate this understanding. While large language models serve specific purposes, they fall short in applications involving spatial problems, such as humanoids or self-driving cars. Developments in this area stand out as one of the most exciting and challenging fields shaping the future of automation.