The next frontier: enter the Large Engineering Model

Jaroslaw Rzepecki, PhD, CTO of Monumo, further explores.

LLMs leverage massive, freely available Internet data sets, requiring vast server farms and enormous compute power. They aim to 'replicate human behaviour' and generate text, stories, images, and videos – with the output often being subjective. If we’re looking for new solutions to help us solve some of the world’s most complex problems, from climate change to drug discovery, we must look elsewhere the answer – smaller models requiring less data, less compute power, and more upside.

Introducing LEMs – Large Engineering Models

In the engineering industry, there is no room for interpretation and subjectivity. Engineering operates under the rigorous rules of physics and calculus, requiring factual precision.

Large Engineering Models (LEMs) are a specialised subset of machine learning models designed to tackle complex engineering and physics challenges. Unlike traditional machine learning models that focus on general tasks like image recognition or language processing, LEMs are tailored to simulate and forecast outcomes that help to design systems, anticipate potential issues, and optimise performance. Trained on smaller, specialist datasets, LEMs are more cost-effective and arguably more helpful in addressing critical engineering challenges, such as the push towards decarbonisation.

Unlocking system-level optimisation

From automobiles to aircraft, we are surrounded by impressive feats of engineering. But the reality is that there is still scope for a wide range of engineering systems to be truly optimised. There is always a compromise – size, weight, heat, pollution, or the overuse of environmentally damaging materials.

Humans struggle to optimise design at a 'system level' because the potential permutations of these complex systems are too vast to process.

Here’s a simple example: if an engineer was tasked with looking at the design of the motor in a small passenger car, but the only things they can change are the length and the radius of the motor, and there are 40 different options for length and 10 possible widths for the radius, then they can very quickly find the optimal solution for by 'simply' looking at just 400 permutations (40 x 10). In reality, motor design is vastly more complex, and even in this simple example, we quickly end up with an equation that looks more like 40 (length) x 10 (radius) x 5 (magnet configuration) x 10 (stator configurations) – 20,000 permutations. By adding just two more parameters, the number of computations you need to make increases by a factor of 50 – a task beyond human capability and the current industry's standard tools.

LEMs can dramatically reduce the time required to analyse and solve problems, thereby accelerating the design and development process. Instead of designing complex engineering systems by hand and assembling several individually designed components, LEMs can explore millions of prototype designs in hours. The precision of LEMs ensures that solutions are reliable, which reduces the risk of errors that can lead to costly revisions or failures in real-world applications. By optimising designs and processes, LEMs help reduce material waste, improve energy efficiency, and lower production costs.

The path forward

While the LLM hype has made artificial intelligence and machine learning 'mainstream,' we are yet to see any real benefit. Smaller, industry-specific models like those in drug discovery and LEMs in our industry are the key to unlocking real economic, societal, and environmental benefits. LEMs will fundamentally change the level to which you can optimise engineering designs – all delivered in a fraction of the time of current approaches. It is time for media, analysts, and company CEOs to turn their attention to where the actual value of this groundbreaking technology lies.