AI, work and regulation

Artificial intelligence (AI) has become an increasingly important technology that has impacted the economy and jobs in various ways. Or at least that’s what OpenAI’s artificial intelligence chatbot ChatGPT told us when we asked it to summarise the effect of AI on the economy and jobs.

‍

The chatbot isn’t wrong about its increasing importance. Over recent months, ChatGPT and similar services have made headlines as students and civil servants alike have attempted to outsource complex tasks to this type of artificial intelligence.

‍

So what are these “various ways” that rapidly developing AI technology are impacting the economy and jobs? And can regulation really solve the issues it presents? 

‍

White paper on regulation. Last month, the Government published a white paper detailing “plans for implementing a pro-innovation approach to AI regulation”. Central to the proposals is an ambition for the UK to be a science and technology “superpower” with AI identified as one of five key technologies in this effort. However, the proposed regulatory frameworks are “underpowered” and risk “leaving prominent AI harms unaddressed”, warns the Ada Lovelace Institute.

• Old regulations, new technologies. Artificial intelligence is by no means new, the term itself is well over half a century old, and yet the regulations around its development and use are still in their infancy. This is particularly alarming in the context of the drive to create AGI (Artificial General Intelligence), and the potential imminent arrival of  “God-like AI”, as argued by AI investor Ian Hogarth recently. This iteration of AI, of which some argue OpenAI’s ChatGPT and DeepMind’s GATO are precursors, “learns and develops autonomously, that understands its environment without the need for supervision and that can transform the world around it”. But the issue of regulation isn’t just about the future, it's about how the tools that exist now are being integrated into our lives. The Government’s White Paper, which includes a consultation that is open until the 21st June of this year, does seek views on the topic of general-purpose AI “for future action”. However, as it stands, the white paper relies heavily on existing regulatory frameworks and expects not to implement the new AI framework or assess its effectiveness for at least a year. As noted by the Ada Lovelace institute, this rapidly developing area of technology is already being “deployed in high-stakes contexts such as recruitment and healthcare”. 

‍

AI and work. The adoption of AI in workplace settings is one of the frontiers of this debate on adequate oversight and regulation. A reliance on AI for “hiring, firing, pay and promotion” could make workers in several sectors increasingly vulnerable, according to the TUC, with the TUC’s Kate Bell describing the Government’s white paper as “flimsy” at its AI@work conference yesterday. Recruitment in particular has been identified as a key area in which harms may already be present and must be addressed much more quickly than the expected timeline would allow. According to the TUC, AI systems are already being used “to draw conclusions from candidates’ facial expressions and their tone of voice in video interviews”. The Ada Lovelace institute argues that key risks in recruitment are either not covered by regulation or are covered by “a patchwork of regulators”. For example, it is unclear that any regulator is responsible for ensuring that AI videos used in recruitment processes are used transparently. A lack of “statutory backing” and “adequate resourcing for regulators” are also identified as key risks for monitoring the use of AI in these areas.

• Power and control. Ultimately, unions are exercised about AI because it has the potential to blur lines of accountability and power relationships in the workplace. Decisions will be being made about workers based on algorithms that often have very little transparency, and are barely understood by the managers and HR departments adopting them. In response, unions are arguing for new laws and regulations, such as the expansion of collective bargaining rights to include employee data, and for clear accountability for decisions made by AI so they can be challenged. 
• Job losses. Of course, the impact of AI on work is about the availability of jobs as well. A new study has found that around 80% of the workforce could have at least 10% of their work tasks affected by the introduction of LLMs (Large Language Models which includes Generative Pre-Trained Transformers like ChatGPT), and approximately 19% of workers may see at least 50% of their tasks impacted. Research by Goldman Sachs found that AI could replace up to 300 million jobs across the UK and Europe. The report also found that 46% of tasks in administrative and 44% in legal professions could be automated. 

‍

So what happens next? The response to the Government’s white paper has focused on regulation, with hopes that promptly updated regulatory frameworks with proper statutory oversight and resources could mitigate existing and emerging harms. While this is needed and can be helpful in protecting existing workers and job seekers, wider structural concerns about the impact on the job market must be addressed. Speaking in a US context, the founder of Stealth Startup JB Rubinovitz argues that governments “need to start building safety nets now”.

• Rights at work. In addition to specific new rights and regulations referenced above, there is a more old-fashioned point about power dynamics which needs to be considered every time there is a new technology or development promising to ‘disrupt’ work. With union membership and collective bargaining coverage still at relatively low levels, large amounts of the UK workforce have limited ability to shape or control the way that AI will interact with their working life, leaving them vulnerable to redundancy or even practices like ‘fire and rehire’ which might be deployed by employers in response to AI changing job roles. Extending both union membership and collective bargaining coverage is key if workers are going to have more of a voice in shaping the future of AI at work.
• Social Security. Another aspect is the ability of the social security system to cope with the impact that AI might have on the labour market. One issue is whether the social security system is robust enough to ensure people's incomes in and out of work can’t drop below a certain threshold, meaning calls for reform could gain traction. Another issue is churn, where ideas such as a new system of employment insurance to help give workers more resilience to bounce back quicker from periods of unemployment could gain momentum.‍
• Working time. Finally, it is worth asking the question of whether the productivity gains from the introduction of AI into the workplace should not be realised in the form of more leisure time for workers. That’s what Oxford economist Carl Benedikt Frey thinks - arguing that "Any technology that increases productivity, ChatGPT included, makes a shorter workweek more feasible.”