How AI Will Change the Way Humans Invent
Instead of designing complex engineering by assembling individual components, AI will help evolve entire, optimal systems
Humans struggle to design complex systems. The requirement to understand every interaction across multiple diverse system elements is beyond anyone, or any team’s, capability.
We generally only master one area of expertise in a lifetime and even the most talented manage two or three at most. This is fine when you’re a musician or mathematician, but less helpful when trying to create the best possible design for something that requires multiple disciplines and the precise interplay between multiple variables. Within the field of engineering, few complex systems are ever truly at a system level today.
With advances in compute power and technology, the last couple of years have seen a new opportunity; by harnessing the power of AI, we now have the potential to mitigate human bias and design preconceptions whilst being able to look at system-level integration of complex systems.
Next Step in Engineering
The post-industrial revolution saw engineers rely on pen and paper, and this manual design process meant testing physical prototypes until a design worked– a very time-consuming and largely inefficient process. Enter the 1970s, which saw the introduction of computer-assisted design (CAD) software unlock human design capability by enabling speedy testing of multiple virtual 2D prototypes.
Fast forward to today and we have the most powerful tools at our disposal: AI and machine learning. Instead of designing complex engineering such as cars, planes and energy generators by assembling a multitude of individually designed components the way humans do, AI will help us evolve entire, optimal systems. These apex designs will be perfected for their intended function, designed to be as good as the laws of physics allow. Not only will this enable more purposeful invention, but it will also reduce years-long engineering projects to a matter of months or even weeks.
While this may all sound too good to be true, we are much closer than you think. But we must shift the AI conversation away from deep-fake videos and cheating on homework to focus on solving more salient problems with a bigger economic, societal and planetary upside.
Age of Domain-Specific AI
Until now, most AI research has focused on large language models (LLMs). Due to the vast amount of training data available “for free” on the Internet, this was a natural entry point. Still, it restricts the application of AI development to language, text, sound and images.
ChatGPT-4o might be able to teach math but surely AI innovation needs a more significant payoff to justify billions in investment, the effort of the smartest minds and considerable computing power. And then there is the terrifying amount of energy LLMs demand. We need to use this technology to address today's most complex challenges. Creating a sustainable future should be top of that list and optimizing engineering design will be vital to this goal.
Electric motors, complex systems that already use half of the world's electricity, and electricity generators, for which we rely on almost all clean energy, need to be as good as they can be. Innovations in either motors or generators will significantly impact the speed and viability of decarbonization. Even a 5% improvement in motors would remove gigatonnes of emissions - the equivalent of the total emissions of France and Germany combined.
If we want to enable AI to help us design better transport, energy sources, heating or cooling, LLMs are not the answer. They undoubtedly increase the current speed of innovation but we need new data sets to train on and domain-specific AI models that leverage industry data to create value in the years to come.
By collecting data from our AI-driven simulations today, we will see the emergence of the world’s first large engineering model (LEM) within the next four years. A more focused model like this will balance the energy required to power it with the payoff in society, unlocking new avenues of innovation throughout engineering and changing how we will invent forever.
About the Author
You May Also Like