ChatGPT is Pushing Engineering Buttons

Alien intelligence, or a tool to take over engineering drudge work? Depends who you ask.

When a revolutionary new technology or service emerges, there is always a knee-jerk reaction. Understandably, when livelihoods are threatened people tend to build walls around their existing way of life and work. This is typically followed by calls for a technology, or a resulting service, to be banned. To some extent we’ve seen this with the likes of Uber and Airbnb. But these technologies and services emerge for a reason and that is usually because they tend to be useful.

OpenAI’s ChatGPT AI language model, for the moment at least, is no different. Although leading thinkers and tech observers such as Yuval Noah Harari suggest “we have summoned an alien intelligence,” in engineering and manufacturing, ChatGPT has already stirred-up considerable interest. Far from being a job killer, it is already being seen as a useful tool in analyzing and translating data and in saving time on generating text responses to complex engineering questions.

Of course, there are caveats. As we’ve already outlined previously on engineering.com, ChatGPT is not immune from what it calls “hallucinations,” or in other words, errors. Also, ChatGPT currently only works on data up to 2021. That should shape the thinking in how to use the technology. As Amit Jain, chief product officer at ServiceMax says, “it [ChatGPT] shouldn’t have the final word.” It’s far from foolproof and yet, that has not been enough to dampen enthusiasm.

“Engineers and manufacturers are finding ChatGPT to be a powerful tool for a wide range of applications,” says Angelo Sorbello, founder of Linkdelta.com, a generative AI platform. “One major advantage is its ability to quickly generate text responding to complex and technical questions. This can save engineers and manufacturers significant time and effort in researching technical issues.”

Sorbello adds that ChatGPT can also help with writing technical documentation, generating product descriptions and even automating customer service responses. This is perhaps more likely, at least in its current iteration. Of course, ChatGPT 4, the new and much more powerful version of the technology, has an ability to go much further. As Jain from ServiceMax points out, its ability to process and analyze unstructured data makes it an interesting proposition, especially for manufacturers and machine service engineers.

“ChatGPT can do much more of the drudge work for engineers,” says Jain. “It is incredibly good at consuming unstructured data and generating content. This has huge scope for field service.”

The language comprehension in ChatGPT means engineers can jot down a few notes as part of their debrief and the AI can use the content to generate a report, for example, adds Jain. This means engineers can automate the mundane work, such as writing and summarizing what they’ve just done when they complete a job, and generate a report quickly and easily.

“For service engineers, any practical tool that saves them time means they’ll use it,” says Jain. “Depending on how well ChatGPT delivers on its promise, such content generation could eventually be used for customer facing content, as well as educational, upselling and personalized cross selling content.”

An incredibly efficient assistant

Jain argues that this could in fact help to reshape machine and product service and support, by opening-up and empowering opportunities for self-service capabilities but also remote support teams. This could, says Jain, be transformational in how organizations structure their service strategy and service operations.

“If more equipment can be repaired or maintained remotely more often with the help of remote support engineers, it will save on truck rolls, overhead costs and CO2 footprint,” says Jain. “While human field service engineers will never be replaced, ChatGPT has the potential to remove the drudge work, focus on the subject expertise and enable them to truly sharpen and apply their engineering craft. Think of ChatGPT as an incredibly efficient assistant for now.”

This idea of efficiency is a common thread. For Modulus CEO Richard Gardner, this makes ChatGPT attractive to manufacturing. Again, the ability to analyze large amounts of data in real-time is important. Gardner suggests this can help to “enhance the quality control process in order to minimize waste.” In addition, Gardner talks about chatbots and the ability to enable customers to self-serve product and service information out of hours.

This idea of knowledge sharing can also be applied to documenting work for training purposes. For Jain at ServiceMax, ChatGPT could have a knowledge transfer role to play, especially among an aging field service and manufacturing engineer workforce.

“It can help organizations capture and organize the unstructured knowledge, experience of the experienced workforce and make it accessible in a format that is desirable for the next generation of workers,” says Jain. He suggests that as opposed to viewing paper manuals, calling a friend, and completing countless forms to close out a service job, ChatGPT can help engineers have access to relevant and contextual information via a query, with work instructions overlayed visually via augmented reality, for example.

“Work can be summarized and documented via generative AI,” says Jain. “While it might seem futuristic, it’s the type of vision for field service that organizations must have to create a marketable product for the next generation workforce.”

Not quite a replacement

That’s a key point given all the noise surrounding ChatGPT when it comes to taking jobs. This is not going to replace a workforce any time soon. If anything, it will only enhance an existing one and make them more competitive.

According to Rajat Kohli, managing partner at global consultancy Zinnov, ChatGPT can be a valuable resource for engineers working on a project to design a new product. It can, he says, “quickly provide information on materials and manufacturing processes, saving the engineer time and effort in the research phase.”

But of course, all of this comes with a warning. As Kohli continues, “ChatGPT’s responses may not specifically address the needs of the project and there may also be concerns around data security and privacy.”

Gardner too says that while large language models have incredible power to analyze data, “ChatGPT can misinterpret complex data, which can lead to liability for users, especially in the engineering space. Most wish list items right now are for better learning based on user feedback.” He adds that, in engineering, at least, AI systems which can understand non-textual data will be in demand.

They key here is that engineers have the power to use ChatGPT in areas where they feel they may need some support—creative ideas, report writing, concept comparisons and so on. It’s currently limited but a real insight into what is to come down the line.

As a blog from engineering software provider PTC recently claimed, for the foreseeable future ChatGPT can in no way “replicate the experience, intuition, and problem-solving skills of a human engineer.” It is a tool to help enhance those skills, neatly summed-up by Kyle McIntyre, head of AI engineering at Quiq, who is keen to calm the hype.

“ChatGPT is fundamentally a very impressive text generator, and it will undoubtedly have a large impact on business and on the future of AI. But it’s not as if we’re suddenly living in the movie Blade Runner.”