Siemens and Microsoft have announced the development of a new Siemens industrial copilot to improve human-machine collaboration in manufacturing. They are also working to integrate Siemens Product Lifecycle Management tooling with Microsoft Teams as an important step toward enabling the Industrial Metaverse. This will simplify virtual collaborations between engineers, frontline workers, and other teams.
The first fruits of the copilot will help users generate, optimize, and debug complex automation code. It will also reduce simulation times. The generative AI aspect will ingest automation and process information from Siemens’ digital business platform and enhance it with Microsoft AI services. Maintenance staff will be able to retrieve repair instructions, while engineers can get faster access to simulation tools.
Features like adaptive UI and command prediction can assist new engineers by providing suggestions and guidance on using the tools. This trustworthy AI assistance lowers the barriers to entry for new engineers and shortens the learning curve. Additionally, AI can make it easier for design and manufacturing engineers to find information and solutions quickly and reliably, improving overall workflow.
In the long term, the two companies are working on copilots across other industries including manufacturing, infrastructure, transportation, and healthcare. Other manufacturing copilots are planned for automotive, consumer packaged goods, and machine building. One early adopter is automotive supplier Schaeffler AG. The company is using the new industrial copilot to help engineers program robots and other industrial automation systems.
They are fine-tuning existing models via prompt engineering rather than training new ones. Siemens domain experts are fine-turning these models for different specific industries.
Warm up to digital twins
Although Siemens is a leader in digital twins, the initial copilot will focus on facilitating human and machine interaction for code debugging, which is not directly related to the use of digital twins. Siemens and Microsoft will plan to collaborate on other use cases where they can apply generative AI to enhance the use and creation of digital twins.
Dale Tutt, VP of Industry Strategy at Siemens Digital Industries Software, explains:
With AI increasingly capable of processing massive amounts of data, it will be possible to directly apply AI to the digital twin so that AI would be able to learn from the digital twin and suggest insights based on analysis of that data for engineers to apply. This will help them deliver smart, connected, and innovative products more efficiently. A first step will be applying AI to drive generative design and enhance simulation and testing. As there is increased trust in the AI, we will see more seamless integration of AI with human expertise to enable optimization of more products and processes, which include combining AI with a digital twin.
Data is a foundation of the digitalization of physical infrastructure. Digital twins add a semantic layer and provide a way of identifying how data is linked across various sources, formats, and use cases using a digital thread.
With AI increasingly capable of processing massive amounts of data, it will be possible to directly apply AI to the digital twin so that AI can learn from the digital twin and suggest insights based on analysis of that data for engineers to apply. The integration with digital twins will also be essential for validating code changes. A hallucination that results in bad code or unsafe designs could break expensive machines or risk human safety.
Simulation is a crucial step before the deployment to validate the results. For example, gen AI changes in automation code will be simulated to make sure that the output is valid.
This will be a gradual process. The first step will be applying AI to drive generative design and to enhance simulation and testing. Tutt expects more seamless integration once they have demonstrated increased trust in the AI, predicting:
In the future, AI will facilitate cognitive assistance for creative, real-time collaboration and feedback and enable AI-generated prototyping and fabrication. By processing large libraries of data including all the company’s product and process digital twins, AI will enable engineers quickly and efficiently evaluate and explore thousands of possibilities, so they can produce the most innovative and optimize processes and products.
The most popular innovations in generative AI started life trained on random text written across the Internet. Recent progress in multi-modal techniques has suggested ways that newer large language models could be trained on text, code, audio, and robot instructions.
Future innovations trained on digital twins will help make more efficient sense of complex systems like industrial equipment with vetted data. These kinds of tools could learn from the digital twin to suggest insights based on the analysis of that data for engineers to apply. In turn, the result could be tested more quickly to find problems. Additionally, generative AI will make it easier to query data about a complex system represented in a digital twin to find a quick answer, saving hours of research.