The most crucial facet of Artificial Intelligence (AI) is developing the technology without turning a blind eye to its consequences. AI is ultimately built by human beings, and humans can have very diverse motives for why they create something. Unfortunately, today there is a massive gap between people making these systems and those impacted by these systems. The changes that AI will bring to the jobs done by humans will have marked consequences on societies, economies and labour markets. For example, robotics has enabled us to do work that was too dangerous or impossible for humans or that is done more effectively by a machine.
Machine Learning and Prediction Problems
Machine learning is the study of computer algorithms that can improve automatically through experience and by the use of data. It is seen as a part of artificial intelligence. It is essentially a ‘prediction technology’; thus, its development reduces the cost of engaging with prediction tasks. In other words, ML advancements make it cheaper and easier to solve problems. But how is this revolutionary? The critical insight is to imagine how many aspects of our lives – personal and professional, can be recast as prediction problems? The answer is many; all we need is a bit of creative thinking. For example – when the problem of self-driving cars was recast as a prediction problem, i.e., instead of focusing on coding actions for all possible scenarios a self-driving car might encounter, the problem was reframed to have the machine predict what a ‘good human driver’ would do? (Alebit there are issues with defining ‘what a good human driver does’).
But what does this have to do with the ‘future of work’?
To understand how the technology might impact our lives and hence the future of work it is not enough to know what the core function of the technology is, we also need to clarify what type of technology ML is. Technologies are tools for humans and to learn how to use them we need to understand their full potential. Once we know that, then we can discuss when and how it is best to use the technology as a tool. ML is speculated to be a particular type of technology called General Purpose Technology (GPT). GPTs are rare; they hold the potential to affect economies and the social structures of society. Examples include the steam engine, electricity, the internet etc. ML is speculated to be a GPT because prediction problems are pervasive throughout the economy. But widespread potential use is not a sufficient condition to be GPT. GPTs are transformational because they also enable an ‘innovation feedback loop’ that sustains economic growth for a long time. Specifically, GPTs not only can be used to solve problems in several sectors of the economy, but in doing so, they stimulate new ideas and innovation in those sectors. In the case of ML, that would be equitable to use insights from innovation in these application sectors to continue to advance the core prediction technology toward the AI goals. But historically, GPTs are not of much use by themselves – their potential is realised when combined with other technologies, knowledge or insights. For example – electricity without light bulbs or without circuit boards is not very useful. Similarly, a tool that predicts well is not very useful in itself. Thus, without coordination between the GPT specialists and other domain specialists, the GPT potential is not likely to materialise.
The decision to Shape the Future of Work – Alignment and Incentives
The critical task here is deciding upon how to combine and coordinate different domains.
To do so for ML/AI requires making choices about: engaging in coordinated innovation across industry sectors and strategically selecting with whom to engage in such a collaboration; it also requires organisations to understand that the coordination effort takes time and is costly, and most importantly they must ensure there is alignment among different levels within the organisation, incentives need to be aligned internally to enable successful engagement in such a complex coordination effort.
To understand this in detail – Workers fear job displacement while top management is excited about the growth possibilities. How to achieve alignment? An approach is to recognise that the prediction technology can be developed as a tool that primarily substitutes for labour tasks, mainly focusing on automating current labour tasks. On the other hand, it can also be designed to complement labour tasks by embracing innovation and imagining new possibilities. The actual benefits will always rest with the latter. Hence organisations would benefit from carefully making and communicating these choices. Indeed, this is easier said than done. It all boils down to alignment and incentives but not understanding these dynamics can lead to missing out on the benefits that AI can bring or, even worse, following a path of technology developments that are detrimental rather than beneficial to the economy and society.
Looking at the adoption of ML across sectors today, the concern for the future of work due to substitution and little opportunities for complementarities can be explained at least in part by misalignment. These differences need to be reconciled among all parties involved and the incentives need to be aligned top to bottom in every sector. The path favouring complementarities is a natural choice when technology is used as a tool in innovation. In the private sector, it can give workers a lot more freedom by making it easier to perform the tasks and also allowing them to focus on other tasks of their choice. The end result – is an increase in productivity and social development of labour that can ultimately advance the quality of life. In conclusion, the recent advancements in AI hold tremendous potential but realising this potential is up to the choices we make; it is these choices that will shape our lives and the future of work.