ChatGPT, AI and the labour market

Nice… but will I lose my job?

Is the hype around ChatGPT due to ingenious marketing or the disruptive power of AI on all our lives? What will it mean for the future of work and how will we have to adapt? SBE’s Jermain Kaminski and ROA’s Mark Levels look around and ahead.

ChatGPT has been variously described as the driver of a tech-utopia and an agent of doom for the creative classes. But how has the company behind it managed to monopolise the discourse and what will be its impact on work? Jermain Kaminski has been interested in AI and LLMs (Large Language Models) since his PhD at RWTH Aachen and MIT, where he looked into predictions based on machine learning, its inaccuracies and shortcomings in ethical decision-making. His perspective on ChatGPT is also an entrepreneurial one.

First mover advantage

“Allowing people to play around with ChatGPT for free through the chat interface was really important.” Kaminski likens it to the touch screen of the first iPhone, which made a lot of already existing technology broadly accessible. “ChatGPT is the first software ever to have around a million users in just five days after going public – they managed to set up an infrastructure and interface that’s very scalable.”

OpenAI offers white-label solutions, so APIs [Application Programming Interface] that can be embedded in other systems including but by no means limited to Microsoft. Different companies can customize the engine for their own purposes. “Many big companies will probably experiment with it soon. Obviously the more broadly applicable a programme is, the more economically viable it is.” On top of all the publicity, the ChatGPT API has recently been made ten times cheaper.

“They are securing the first mover advantage. Basically, you can’t make a good joke twice – even if Amazon, Google or Facebook come up with another model, OpenAI already has a foot in the door. They have created a buzz – that’s what platform building is all about.” In the public perception, ChatGPT has become synonymous for LLMs and chatbots.

Better at some things

“At the moment we are overwhelmed by the potential. I think, in general, AI excels at very structured tasks which are primarily a question of computing power, for example protein-folding in the pharma industry.” Yet, Kaminski points out that very few breakthroughs are the result of structural innovation. He cites the complex chain of serendipities amongst the chaos that led to coffee and eventually cappuccino. “It’s really unlikely ChatGPT would be able to invent something like that from scratch.”

Kaminski is currently writing on ‘computerpreneurs’, economic or entrepreneurial agents which aren’t recognisable as computers, acting on their own in the market. “You already have systems doing part of that work in stock trading for example, but you could imagine e.g. an AI analysing Amazon sales data, finding undersupplied parts of the market and then building a business around that. I think it will happen and it might need to be regulated.”

Kaminski says history shows that these leaps in technology increase jobs in the long run. “IBM Deep Blue can beat any human chess player – but chess players haven’t become obsolete. They have become better by using the machines as sparring partners.” He concedes that some jobs will fall by the wayside and that people should contemplate how they can excel at the uniquely human aspects of work. He also points out that we’re often willing to pay more for artisanal bread or handmade artefacts because of the cloud of associations and emotions around it, even though a similar product that is more obviously the result of an alienated production chain would be cheaper.

Overdue attention for a real problem

Mark Levels, who researches the impact of automation on the labour market, is less optimistic. In many ways, he embraces the media furore because it has brought a very relevant discussion to the fore. "We’ve researched the way automation affects work for years now. The development of AI has been an ongoing process and the public are only now catching up to it now – but that is good that they do. We can’t afford to sit on our hands any longer – we need to act before it’s too late."

Without discussing the philosophical niceties of what constitutes style and human creativity, ChatGPT’s output is certainly good enough for many commercial uses. It’s also worth pointing out that, while it might unable to do everything a worker does, it can now easily outcompete humans on large chunks of the repetitive tasks. “Once you needed a good idea and the ability to write or illustrate – now the good idea and the skills to prompt the AI are enough.” The craft often takes up more time than ideation – which makes for easy efficiency savings.

In some ways, this might augment workers – but only those with the skills to use these systems. Instead of ten translators, it might be enough to have one who is proficient at spotting the few but involuntarily hilarious errors Deepl has made. In other sectors, it might be worse. “Job losses due to automation in, for example, inbound call entres will be huge. Companies have been rolling out the technology for years now. Often, people there already act as intermediaries between customers and information generated by chatbots. The World Economic Forum also has worrying predictions for bookkeepers, accountant, administrators, factory and storage workers – everything with a lot of routine tasks.”

Lifelong learning the new bar

“The speed of innovation is one of the main determinants of the impact on the labour market – and that’s quite worrying here.” Levels agrees that making ChatGPT publicly available was ingenious marketing, adding: “the hype is a self-fulfilling prophecy. It influences where the money goes.” While the hype might still fail to materialise, Levels doubts that and warns that the internet’s impact had also been gravely underestimated initially.

Of course, no one knows what the future will bring. “There’s always uncertainty but there are some no-to-low-regret policies like redesigning the curriculum to involve digital skills and metacognitive skills for continuous learning.” Levels also advocates for cooperating with education institutions, employers and unions to provide an infrastructure for retraining. “It also makes sense to focus more on vocational training and jobs that are relatively immune to automation, like plumbing or care work.”

Overall, Levels thinks there will be a shift in the way we perceive of work. “The idea of getting training and then remaining in one line of work until retirement will likely become obsolete. Independently of your level of education, you will have to keep learning and acquiring new skills to remain attractive on the labour market. We are now building the Maastricht Academy for Lifelong Development, to take this responsibility and provide reskilling and upskilling opportunities at scale.”

This text was not generated by ChatGPT because I will not be complicit in my own demise for a cheap giggle…

By Florian Raith

Mark

Mark Levels is Professor of Health, Education and Work, programme director at the Research Centre for Education and the Labour Market (ROA), and dean of the Maastricht Academy for Lifelong Learning. He teaches macrosociology at University College Maastricht. Levels has set up TECHNEQUALITY, a European consortium researching how AI and robotisation will affect the labour market.

Also read

More news