Could generative AI help to fill the skills gap in engineering?

AI in the workplace

According to the survey, 58% of companies have introduced AI tools into engineering teams and 42% of them only use AI tools in different parts of the business. The most common AI tool used is a large language model (LLM), with nearly 60% of businesses using this.

Meanwhile, nearly a third of companies use machine learning and productivity tools, such as Microsoft 365 Copilot, to assist in their work.

Generative design tools, such as those used in simulations to optimise designs or identify potential faults, are less common, with less than a fifth of organisations using them. Computer vision and neural networks are even fewer, with just over a tenth using them.

Nearly a third of the survey’s respondents use AI tools for written tasks, such as emails and pitches. Meanwhile, approximately a quarter of respondents use AI for data analysis. However, AI’s use in data analysis is expected to grow, as nearly 60% indicated they would accept AI assistance.

The tasks that engineers would most like AI to be used for are those for simulation and tools that can improve productivity. AI tools for design optimisation, predictive maintenance and research followed close behind. It is worth noting that nearly two-thirds of the respondents believe that AI tools will automate mundane and repetitive tasks, which will make engineers more productive and enable them to focus on complex or creative tasks.

“In the short term, AI is going to be operating mostly as a co-pilot for engineers. What we’ll see with AI is the ability to start utilising this technology to automate mundane tasks that might have been time-consuming, allowing engineers to move on to more interesting activities,” says King. “There is a big opportunity here, but we have to be careful so that we don't lose the human-based knowledge.”

Concerns remain

There is concern (37%) that widespread adoption of AI will result in engineering roles being replaced with AI tools. Just over a quarter believe that engineers would be replaced. Likewise, over 40% of respondents do not feel that AI tools would result in maintaining the same level of engineers.

There is also the concern (66% of respondents) that widespread adoption of AI tools will lead to reduced project oversight. This is partly due to AI tools being akin to a black box, where there is insufficient transparency to understand how AI derived a solution.

“The AI world can be a bit like the Wild West, but in an engineering context, that doesn't work. You've got to have systems that are reliable, provide the right answers, are safe, and behave in an ethical way,” says King.

“If we look at the sort of framework that we've used for years, especially in areas like aerospace or nuclear engineering, there are very strict rules and guidance. We almost have to take some of that learning and apply it as safeguarding principles to any AI systems that we're introducing.”

The lack of understanding behind an AI’s design methodology, together with being unable to properly interrogate the solution, could cause problems with the verification of designs. With a growing number of solutions generated by AI systems, it will become even more vital for skilled engineers to interrogate these designs to ensure they are suitable and appropriate.

Over a half of respondents also raised concerns about the potential security risks of AI tools, as well as nearly 50% being concerned about potential historical bias in the data. Overall, nearly 55% of respondents are not comfortable with AI being used to make critical decisions in engineering.

Companies using publicly accessible LLMs, such as ChatGPT, are especially at risk. Not only could they be exposing themselves to poor datasets and misinformation when importing AI generated content into their networks, they are also potentially leaking sensitive information.

There is a strong feeling among the respondents that regulatory oversight is needed to ensure AI is deployed and used appropriately in engineering. However, given the speed of technological development in AI tools and the comparatively slow legislative processes, this is easier said than done.

Some AI regulations are being developed, such as the European Union’s Artificial Intelligence Act, but there is a significant risk that legislation could rapidly become obsolete.

“AI developers are applying reinforcement learning with human feedback - when they see the models do something, they'll say whether or not they think the models behaved in the right way. That's based on their perceptions and biases, but somebody who is sitting in the Middle East or Russia might have a very different view about how the model should have responded,” says King.

“You've also got to look at the data they're training the LLM on, which is usually scraped off the internet and often in English. If you're only training on English-language websites, there's a chance that it's biased towards Western cultures.”

The future of AI in engineering

The roll-out of AI tools in engineering is already well underway, but carries potential pitfalls.

This thinking was clearly articulated by one survey respondent, who noted: “A computer should be able to more easily and more quickly identify patterns and check against known problems. On the other hand, human nature will encourage people to believe blindly the results of any AI task, which could be a problem.”

Companies can also learn from the previous deployment of new technologies to identify potential risks. A key element is that different countries have different engineering regulations and guidance documents.

As such, an AI tool developed for one region may be incompatible, or least require re-learning, before it can be deployed in a different region.

“My one hope for engineering is that it doesn’t try to use AI as a way to save money, but as a way to accelerate performance,” says King. “In the long term, AI creates an inflection point for us all, because we’re able to develop systems and products faster and better, you should then see an acceleration of that technology like we never have before. It should open up huge breakthroughs.”

Although AI tools have clear benefits for the automation of mundane and repetitive tasks, engineers will still need to learn new skills to fully engage with AI, ensure safety and maximise the benefits.

There will be a need for engineers that have training in coding and prompt engineering to work with AI systems, while critical thinking will become an essential skill for interrogating AI-generated solutions.