In connection with changes related to technology and technological development, a recurring question since the first industrial revolution is how the introduction of new tools and methods transforms the structure of the labor market. In relation to generative AI, there are basically two dominant narratives.
The dominant element in one of them is that the development of artificial intelligence may significantly reduce the need for human labor in some (if not all) professions. While it is true that AI is likely to create more jobs than it threatens, the pace of progress is much faster than the labor market can adapt to.
If you look at the pace at which generative AI has developed and changed in the year since ChatGPT was launched, it is easy to see the source of concern. Meta released Llama 2 in July of this year, Sanford Alpaca was released by fine-tuning Llama, and GPTAII is a complete ecosystem that enables the teaching and deployment of large language models. The above are just a few examples of the open-source language models published this year and the frameworks that support them. This alone clearly indicates that there is hardly a month when a large company or organization does not add something to the GAI palette.
In this environment, it is often difficult even for developers to keep up with what is new. However, adapting to new technologies is no longer just a question of technical knowledge. In the past, AI has also been characterized by paradigm shifts that required the acquisition of a whole set of hard and soft skills to adapt effectively to a changing environment. The crucial difference between the past and the present situation is that in the past, these were largely limited to researchers and developers. However, with GAI now mainstream, its impact extends far beyond the professionals who create it, a trend that is likely to affect a growing segment of the labor market.
It is also worth noting that many are predicting an exponential pace of development soon as an extrapolation of current trends. This begs the question: how long will the average worker be willing to keep up with the accelerating changes and the paradigm shifts that accompany them? And how long will they be able to do so?
The rise of automation has always eliminated jobs as well as created new ones. The 4th industrial revolution, or Industry 4.0, means a change in that the structural reorganization (according to the signs) is increasingly affecting white-collar jobs that previously seemed untouchable.
Another important factor that comes up regularly is the importance of human creativity, the key role of maintaining empathy, and the irreplaceable nature of human-human interactions.
Based on current trends, it seems quite certain that professions requiring general and systemic knowledge, such as medicine or law, will not be “under threat” soon. Today’s state-of-the-art tools still operate on a stochastic basis—their job is to produce the most likely output based on an input sequence. This is true even if, because of the millions or billions of elements of data, this operation is so sophisticated that it sometimes even seems human-like. In practice, however, the most probable solution is often not the right one, and in many cases, the right decisions are based on a kind of intuition. Not only is there no artificial modeling of such skills, but we’re not sure how these terms are grounded in human cognition.
What’s certain is that the proliferation of AI applications will increase the value of positions where the monitoring and judgment of AI decisions will be key. This can be either an ethical, a morally based judgment, or a way of filtering out the inherent problems of today’s tools (such as biases, prejudices, or even hallucinations caused by errors in the models’ training data). Such human-in-the-loop and human-on-the-loop positions are inescapable not only from a moral but also from a security point of view and are expected to represent a whole new segment of the labor market.
However, it also seems certain that masses of people have never had to adapt to changes of this magnitude so quickly before. Even if AI creates more jobs in the long run than it eliminates, the question is, in the time between, who will be thinking about those who cannot keep up with this change, or will be slow to adapt?
István ÜVEGES is a researcher in Computer Linguistics at MONTANA Knowledge Management Ltd. and a researcher at the Centre for Social Sciences, Political and Legal Text Mining and Artificial Intelligence Laboratory (poltextLAB). His main interests include practical applications of Automation, Artificial Intelligence (Machine Learning), Legal Language (legalese) studies and the Plain Language Movement.