Artificial intelligence and white-collar work
As artificial intelligence and machine learning increasingly reveal their potential, not only in terms of what they can and cannot do, but also in relation to removing entry barriers through the development of MLaaS (Machine learning as a Service) platforms, debates are taking place about replacing workers with machines or to be more accurate, by algorithms in all kinds of areas.
The first approach to these substitution processes can be optimistic, along the lines of “we’ve seen this before a few centuries ago, and after the replacement and loss of jobs followed a reinvention of previously non-existent tasks that reshaped the labor market, and generated other occupations”. Or it can be more pessimistic, on the basis that the surplus productivity machines will generate will make a significant part of the workforce redundant and not needed.
It’s interesting to see how the lines of thought in this regard tend to assume a more rapid replacement of jobs that fall within the so-called three Ds: Dull, Dirty and Dangerous, belonging to the blue-collar sector. Advances in autonomous driving will soon mean that anybody who earns their living behind the wheel will be looking for a new job.
We tend to hear less about the replacement of white-collar workers, perhaps because there is a widespread belief that their jobs are safer, based as they are on qualities we tend to consider inherent with human reasoning. Administrative work in offices or corporate environments, leadership roles in general, are often regarded as safer from machines, or at best, to need help from them: the question is often posed as how managers can do their job better with the help of machine learning, or how machine learning should be introducen in corporate environments, but with no managers being harmed in the process.
But this view is not borne out by reality: planning advertising campaigns across different media, for example, in principle a white-collar job, is increasingly being impacted by programmed advertising models in which a series of algorithms negotiate real-time ad prices for a given medium based on sociodemographics or the behavioral characteristics of users. Where once an executive made decisions and assigned his or her impressions to each format based on different criteria (not all of them especially clear or transparent), a machine now does this much more accurately, professionally and undoubtedly, with more accurate results… in other words, what we are seeing with driving: machines can do the job more productively, without the need for rest times and without distractions, and of course with far fewer accidents.
To what extent might managerial work be replaceable by a machine or an algorithm? Most leadership roles tend to be about decision making in complex situations, involving negotiations or other responsibilities that, in practice, could fall within processes deep learning can systematize. And by systematizing we no longer mean applying a formula or automating a decision, but just the opposite: studying a complex situation and proposing a decision that maximizes the return, either immediately (from a move) or in the course of a long-term strategy. As soon as we start talking about machine-learning techniques, we identify the tasks performed by a machine not just as simply replacing the human that made them, but instead as improving what humans are capable of.
The fact that companies like Google, Facebook or Microsoft, not usually regarded as blue collar bastions, are being remodeled around artificial intelligence, should give us a clue. In many cases, those algorithms will improve their products or services in ways that white-collar humans working closely together could not possibly imagine. And consequently, this will result in the replacement of people previously considered managers.
Given the current state of technology, with algorithms able to understand and analyze speech better than most humans, in different languages and within well-defined contexts framed within social networks interactions, an algorithm’s ability to monitor what is going on in an industry, a company or an area of activity can be already far superior to that of any human. At this rate, it is only a matter of time, value proposition and interest before we see the development of processes based on machine learning decision-making.
(En español, aquí)
