WhatCan machines cause massive job destruction?? The debate is as old as the industrial revolution itself, even if the most optimistic positions have taken over. However, the outlook may be complicated even for them with the emergence of new technologies capable of performing tasks that until now have been considered ‘robot-proof’.
Many experts warn that the development of artificial intelligence it’s in a too rudimentary phase enough to be properly called as such. Nevertheless, the results already obtained by tools that combine machine learningbig data and hyperconnectivity They are already more than surprising (and disturbing) for the general public.
In the workplace, the problem is divided into two fronts. On the one hand, the increasing use of algorithms for decision-making, which can range from hiring to work organization and performance evaluation. Its use is so widespread that commercial models such as the gig economy and digital platforms rely entirely on them and have forced a complicated review of labor relations framework.
On the other hand, we talk about automation: that these tools no longer only serve to decide on human work, but that they replace it. In this sense, a very simple rule applies: “all this task simple and quite repetitive enough to be executed by a machine, it will eventually be executed by a machine”.
The question is to know what is meant by “sufficiently simple and repetitive task”. For most economists, labor market experts, authors and, of course, journalists, it was manual labor that required physical strength: knowledge jobs and jobs that require higher “cognitive skills”those who make up the so-called knowledge economy and who are the same ones these professionals practice, are safe.
The evolution of robotics, not only industrial, but also service, seems to prove them right. The advent of autonomous vehicles (with the promised impact on logistics) seems to be upon us. and although it is not foreseeable that our generation will have android butlers in their homes, already has autonomous vacuum cleaners and internet-connected devices.
Of course, although these robots are increasingly intelligent and autonomous, their function is associated with the performance of tasks, above all physical. for which they were specially designed.
However, this transformation of manual jobs did not result in massive job destruction, not even during the pandemic.
The OECD estimates that 14% of jobs are at risk of being lost due to automation. On his list, the most at risk are servers, followed by by agriculture, drivers, miners or operators.
Spain, by the way, leads the OECD with 23.7% of jobs directly threatened (and 35.3% forced to change), surpassed only by Slovakia, Slovenia and Greece.
The surprise comes when we see that, despite technological advances, between 2012 and 2019 employment increased in these sectors. Indeed, the two countries with the highest proportion of jobs at risk, Slovakia and Slovenia, have improved their productivity (although this is not the case in Spain) and those which have invested the most in robotization have increased their jobs.
Of course, this evolution has been affected by a context of employment recovery after the financial crisis: the analysis shows that the professions most at risk of automation create fewer jobs than the others and do so with jobs of lower quality. This also influences the fact that these are jobs performed by less qualified professionals.
How do new “artificial intelligence” tools change this scenario? Simply expand the range of what is considered”sufficiently simple and repetitive tasks” to cover professions that were not at risk.
The recurring and most obvious examples for the general public are illustrators threatened by image-generating software, or publishers seeing how a program can write a readable, SEO-friendly article or column in seconds.
But there are many others in other sectors, from programming to human resource management, where the emergence of this technology is beginning to replace human tasks. It is worth asking to what extent these had a “cognitive character”, but the fact is that it affects skilled workers more than a waiter.
Although we must bear in mind that in many cases we are not talking about AI itself, but about algorithms that combine the use of data and “machine learning”. Its functioning could almost be described as that of digital “robots”. Of course, much cheaper than any physical robot that wants to serve in a bar.
In this case, the fear of progress on this path to real AI is not just the loss of jobs, but the polarization of work: that is, that the middle class of workers will disappear. while only those in the low and high brackets of the salary bands move up. A phenomenon studied in Spain by UCM researcher Raquel Sebastián and which has already occurred before the current stage of digitization.
To what extent could this trend be aggravated by artificial intelligence? Over the past few weeks, the GPT Chat tool has made thousands of headlines for its incredible versatility and efficiency. Not only is he able to hold conversations, but he can also write all kinds of texts, including poems or scripts on any subject and much better than the algorithms used so far.
But the most striking abilities are those of writing a contract or writing programming code. The latter does not mean that it can “program itself”, but it can replace a large part of the work of a programmer.
“Repetitive cognitive tasks”
In any case, Chat GPT is a clear example that artificial intelligences are general purpose technologies (acronym for “General Purpose Technology”). This differentiates it from the classic concept of robotics (since each robot is designed and built for a specific task) and equates it, according to the OECD, with the steam engine or the Internet. And remember that Google has been working in this field for decades and we still don’t know the full potential of its patents.
This exponentially triggers its impact on employment. In this context, the OECD is starting to talk about “routine cognitive tasks” that can be replaced by these artificial intelligences, although it has not revised its calculations of jobs at risk of automation to adapt to this new reality.
Largely because the technology is not yet fully operational. But also because the organization considers that, as happened with the previous phase of automation, employment should not be affected as long as the professionals are ready to work with this technology.
But for that, of course, the key is not just that companies and the workforce can adapt, but that the country’s production model does. In the case of Spain, very vulnerable to classic automation and with a strong polarization of employment, an additional pressure is added so that this progression does not translate into a loss of employment for millions of people.