By Lukas Tuggener (ZHAW)

Reposted from

There currently is much talk about the recent developments of AI and how it is going to affect the way we live our lives. While fears of super intelligent robots, which want to end all human live on earth, are mostly held by laymen. There are other concerns, however, which are also very common amongst insiders of the field. Most AI experts agree that the short to midterm impact of AI developments will mostly revolve around automating complex tasks, rather than “artificially intelligent beings”. The well-known AI researcher Andrew Ng put it this way:

“If you’re trying to understand AI’s near-term impact, don’t think “sentience.” Instead think “automation on steroids.”

Exemplary of that is the 1998 landmark publication “Gradient-based learning applied to document recognition” by Yann LeCun et al, which is one of the first applications of convolutional neural networks, a technology that is at the core of computer vision to this day. The goal of the application discussed in this paper is to automatically process handwritten checks. So naturally the question on the mind of many researchers working on AI applications, such as self-driving cars or fully automated customer service, is: “will my research cause mass unemployment?”

It’s not about the color of your collar

While we are used to many drivers of efficiency, from the assembly line to digital communication, the jobs, which fell prey to automation where usually manual labor jobs of blue collar workers. This will change during this AI-driven revolution of the labor market. AI is very good at automating routine tasks, no matter if this is a complex cognitive task, performed by a highly trained individual or a simple manual operation.

For example: today’s computer systems are significantly outperforming pathologists at detecting cancer in medical imaging data. This is because the analysis of a patient’s biological tissue is a routine task, even though it takes a lot of training to do so.

This means currently the safest jobs are those that require creativity and ability to improvise, which humans have and AI will lack for decades to centuries (depending on who is asked).

History tells a different story

While fears of the human workforce becoming obsolete are as old as the oldest steam engines, this never happened. This is partly because it in the human nature to never be completely satisfied. If we suddenly have free resources due to an efficiency gain, we tend to allocate them on new tasks. The occupation of a dog’s hairdresser, for example, can only exist because we are satisfying our basic needs with such high efficiency.

Another – arguably more important – effect is that a massive efficiency gain in the production process of product X leads to higher availability and lower cost of X. This in turn will lead to higher demand for X and thus more jobs in the non-automated parts of the production process for X. For example: all the jobs in the car industry can only exist because massive automation has been able to create a broad market for cars. This means automation tends to create jobs “around it”.

Let’s revisit the software system that can detect cancer better than any trained human can. Although this means that the pathologists will no longer be needed to conduct cancer screenings, this will lead to a massive price drop for said service. So it’s not far-fetched to assume, that we all will have our tissue checked for cancer regularly in the near future. This in turn would lead to much more work for general physicians or whomever would conduct these screenings.

A prepared society

So if history’s pattern repeats itself, next industrial revolution will not cause mass unemployment. It will, however, eat up outdated jobs quicker than anything we have ever seen before and will disrupt the labor markets quicker and harder that every preceding industrial change.

Even if I anticipate no pandemic unemployment, there will be still a lot of people which have a skillset that is no longer sought after. Job switches will become more and more frequent, so there is also a growing population between an old and a new job.

This means that as a society it is still very worthwhile to prepare for this job-market disruption (and the many more, which will surely follow).

First and most importantly, it is pivotal that we make a very clear commitment to lifelong learning. The days are gone where one specific education is enough to carry a whole thirty- to forty-year long career. It needs to be the main goal of basic education to teach people how to acquire new skills. It is also necessary to put a framework in place, which provides financing and resources to people in order to be able to go back to school later in their careers.

Secondly, because there will probably be more and more people which are between jobs or have a transition period from one career to another. A social system is needed to alleviate pressures from these situations, such that a job loss is no financial tragedy and proactive career changes are encouraged.

Make AI work for everyone

If history does not repeat itself and AI does actually destroy more jobs than it creates, it is of the utmost importance that we have a serious look at more drastic measures like unconditional basic income and taxes on robots. Otherwise society steers toward – almost feudal – system where a growing part of society would be robbed of the prospect for a dignified and worthwhile life. The resulting growing social tensions and an inevitable collapse of social peace would be a disastrous outcome for everybody.

I conclude that, the short to midterm impact of AI is a social challenge, rather than one of flying killer robots.


[personally signed contributions reflect the opinions of their authors and not necessarily those of datalab]