Artificial Intelligence and the future of work: Is ChatGPT coming for our jobs?
This is the first in a series of articles, Colin Davidson Senior Associate at Cole Khan Solicitors considering how Artificial Intelligence (AI) is affecting the world of work. The commentary can necessarily only be a snapshot in time, as the technology is advancing at an exponential rate, as such any detailed view will be superseded quickly as we all scramble to keep up.
However, the way we work is changing. The incorporation of this new technology into daily tasks is leading to productivity gains across diverse sectors. We are in the early chapters of AI and a sound understanding of the benefits and potential limitations of these tools will assist to build systems providing better services and replacing old ways of working as this technology evolves.
In this article I will begin by looking at the effect of technology such as ChatGPT (or Microsoft’s ‘Copilot’) and whether this will become a replacing or complementing /augmenting force to human labour.
In subsequent articles, I will set out significant concerns of discrimination and bias that can arise out of automated decision making and issues surrounding privacy where this technology is active. In later articles I will consider how this will affect specific sectors, such as financial services, the gig economy, and the judiciary as well as considering the evolving regulation of AI.
Intro
A report released by the International Labour Organisation (ILO) has provided granular detail on the types of tasks that can be automated by technologies such as ChatGPT (or Microsoft’s ‘Copilot’) across different occupations. This report provides significant insight into whether this tech is likely to automate a role in its entirety, or is more likely to augment it, leaving humans to undertake the role with the assistance of technology for the near future[1].
What is ChatGPT
There has been significant press around ChatGPT which has been somewhat polarising. Some view this tech as the forbearer to mass technological unemployment, while others consider it a magic bullet, eliminating the monotony of our daily tasks and freeing up time for more fulfilling work.
ChatGPT, released by Open AI in November 2022 (which has recently evolved to GPT-4o (“o” for “omni”)), has exposed the public at large to advancements in Machine Learning (ML) by introducing a ‘chat’ interface to ML models. This allows the user to provide instructions in natural spoken language, or pictures, as opposed to computer code, thereby opening this technology to the layperson.
Open AI (the creators of ChatGPT) have also released DALL.E which creates images from text captions and we are now seeing the first short movies created exclusively by AI through their service, Sora.
GPT stands for Generative Pre-Trained Transformers (GPTs). The ‘generative’ element is the ability to produce creative output such as sentences, pictures or moving images. The AI is ‘pre-trained’ on large amounts of data where it ‘learns’ how to carry out specific tasks, and the ‘transformer’ refers to the underlying mechanism by which it carries out this task (not a reference a beloved Saturday morning cartoon).
Replacing vs Complementing force
The ILO report analyses the ‘risk’ of jobs being automated by breaking the job down into its constituent tasks. It then assesses which of these tasks could be automated by ChatGPT or whether it requires a human to carry it out in whole or in part.
By way of example, in analysing the role of a primary school teacher, it was assessed that ChatGPT could quite easily develop lesson plans and daily schedules, however a human teaches expertise and nuanced understanding of their student’s needs as required to craft effective lesson plans and provide emotional connection and real-time adaptability. Furthermore, human interaction and physical presence is essential for effective discipline and enforce rules within the classroom. Accordingly, it was assessed that this type of technology would have an augmenting effect, automating the more routine tasks while leaving time for more gratifying and ‘human-like’ activities.
In addition to teachers, it was assessed that roles such as managers and legal and associated professionals are composed of some tasks that are difficult to automate and others that can be automated more easily. As such, these professions are likely to be augmented, rather than replaced by the tech currently in existence.
The occupational group the ILO considers is most at risk to automation (by largest share of tasks exposed to GPT tech) is clerical jobs. It was deemed that the broad occupation of clerical work is highly exposed to the technology taking over, with 24% of clerical tasks at ‘high risk’ of being automated and a further 54% at a ‘medium risk’.
This means that many clerical jobs may never emerge in developing countries where this technology is introduced along with the infrastructure which necessitates these jobs. Clerical work is also a sector which is traditionally predominated by woman. This presents a real problem where these roles have previously been a gateway for increasing female employment and therefore creates a risk to their participation in the labour market of these changing economies.
Ultimately, the primary findings of this report are that most jobs and industries are likely to be augmented rather than replaced by AI, for the time being. This is a finding reflected in a similar report by Goldman Sacs[2] which highlights there is going to be “significant disruption” to the US labour as a result of automation by the combination of this complementing and replacing force. While it considers 300M jobs will be effected, only 7% of current US employment will be substituted by AI, 63% will be complemented, and 30% will be unaffected.
Effect on the legal sector
Employment in this sector consists of a mixture of low-skill, repetitive administrative tasks, as well as complex knowledge work. This is non-routine cognitive work, seeking novel ways to solve complex problems and utilising judgment based upon logical and emotional feedback. In this work there is a significant element of human interaction whereby an individual’s emotional needs require to be assessed through instructions and empathy is needed to form bonds of trust, more so with vulnerable clients. This engages logical-mathematical, linguistic, interpersonal and interpersonal intelligence. This is work that is not currently able to be automated by AI with a high degree of success.
Therefore, in the short- and medium-term, AI such as ChatGPT will likewise be incorporated to automate some tasks and augment others in the legal sector. This is being implemented already where routine clerical tasks such as drafting basic documents (sifting and sorting disclosure, contract analysis, routine correspondence, recruitment procedures, productivity monitoring, performance evaluation) can and is being significantly automated. The evolution and automation of these tasks can also lower the cost of services and potentially increase access to justice, however issues around implementing AI in these areas, and specifically recruitment, will be the focus of later articles.
Nevertheless, at this stage ChatGPT and similar technologies cannot provide the nuanced understanding of client’s needs to build trusting relationships through emotional connection and real-time adaptability to instructions. As such, it is submitted that this technology will have a complementing, or augmenting effect for the time being, automating the more routine tasks whilst requiring human oversight allowing the lawyers to carry out the more ‘human-like’ activities in the legal sector, much like in teaching.
Summary
It is clear this technology is here to stay and is evolving exponentially. AI and ChatGPT / Copilot are becoming a common feature in the world of work as a complementing force, as such it is important that we understand how this works and the potential pitfalls that arise from employing this tech across an organisational or society more widely.
Its implementation can have a widely disparate effect on different professions and even different genders. A key takeaway from the ILO report is that “any form of technological transition would have a strongly gendered effect, with a badly managed process disproportionately harming woman, and a well-managed transition potentially creating important opportunities in terms of woman’s empowerment” (page 34-35). This applies to the augmenting effect in developed economies as well as the replacing effect in developing countries.
In light of this, it is extremely important that the introduction of such technology is managed ethically taking into consideration power balance and the voice of workers. It is on this basis that the Trade Unions Congress (TUC) have formed a task force to consider the regulation of these technologies and produced a draft ‘Regulation and Employment Rights Bill[3]. We have also seen the introduction of regulation by the EU through the Artificial Intelligence Act[4].
At a macro level, Governments and industries will also need to consider policies around skills training to harness this new technology and the new roles created by it (emerging professions include AI ethicists, data curators, and algorithm trainers). There are also widespread discussions around social protection schemes to take account of frictional and structural unemployment during this period of transition. While many may welcome the opportunity to rid themselves of menial tasks taken on by AI, work can also be the source of great pride and meaning within society. Accordingly, this will require bold plans to manage this transition and may require policies such as increasing the job quality and remuneration in sectors such as care, as we see aging populations through the advance of medicine. This could provide roles for displaced workers at the same time as meeting societies need for more care work, work it is deemed at this time can and should be carried out by humans.
At a micro level, there are going to be significant ethical concerns about the implementation of AI in the workplace, including wide-ranging issues such as copyright over the output produced by AI, the difficulty in assessing decision making and liability for such decision and the possibility of bias in opaque decision making.
There is also significant concern on the effect exposure to new technology is having on individuals’ quality of life. In a study undertaken by the Institute for the Future of Work[5] it was found that interaction with new workplace technologies such as AI and ML, could have a negative impact of quality of life, albeit there was significant variation dependent upon level of exposure.
In light of this, it is imperative that the implementation of this technology is done in a collaborative and transparent way taking account of the requirements of different sectors considering the risks involved and ensuring non-discriminatory application. I look forward to exploring these issues amongst others in subsequent articles.
Should you wish to discuss any of the issues raised in this article, please contact Cole Khan at hello@colekhan.co.uk or the author, Colin Davidson, directly on ColinDavidson@ColeKhan.co.uk
[1] For an in-depth breakdown of the data across role see full report here . The data is caveated in that it is based upon technology in existence today without speculating what new technology or what new jobs might arise in the future.
[2] The Potentially Large Effects of Artificial Intelligence on Economic Growth (Briggs/Kodnani) (gspublishing.com)