By: Prof Tshilidzi Marwala
SA, and the world, must adapt now to changes already under way.
As I have travelled across SA and other countries, speaking about the fourth industrial revolution (4IR), one thing has become obvious. Industry 4.0, as the 4IR has become popularly known, is misunderstood and its impact is underestimated. In countries like China, Singapore, Poland, the US and Italy the scale and magnitude of this revolutionary change is almost always misconstrued and underestimated.
The original Industrial Revolution was about the use of water to mechanise production. It happened in Britain, and people known as Luddites opposed it because it was destroying their jobs. But the technology marched on to revolutionise Britain and ultimately the world.
The second industrial revolution was catalysed by the scientific revolution that unified electrical and magnetic forces. This gave us electricity and the electric motor, and these ushered in mass production.
The third industrial revolution was catalysed by semiconductors, which gave us the transistor and the electronic age.
The 4IR, driven by technologies such as artificial intelligence (AI) and biotechnology, is merging digital, physical and biological systems into one.
The first three industrial revolutions created massive numbers of jobs and drove urbanisation, but the fourth will make human beings irrelevant.
What will happen to jobs in the 4IR? Will we still have conventional jobs in this era? For an insight into this, it is important to revisit the work of Hans Moravec, who came up with what is now known as Moravec ‘s dilemma. This states that the level of difficulty it takes to automate a task depends on how long humans have been performing that task on an evolutionary time scale.
For example, humans have been kind to one another for tens of thousands of years and therefore it is incredibly difficult to automate kindness.
Another example is that human beings have been writing for only 6,000 years, so it is easier to automate writing than kindness. In fact, due to the ease of getting a computer to read as well as check spelling and grammar, companies no longer employ people to do this because Microsoft Word is now performing these tasks.
What Moravec taught us is that when it comes to intelligent machines taking over jobs, white-collar workers will be affected more than blue-collar workers. If anyone doubts this, try to find a robot that can prune the trees in your garden.
As it is known now, the 4IR is driven by AI , a technique that is used to make computers intelligent. There are three types of AI:
● Computational intelligence is the use of biological systems such as the flocking of birds or colonies of ants to build intelligent machines. It has been used to create systems such as Google Maps, which identifies the shortest distance between two points;
● Soft computing, which does not require large amounts of data to train; and
● Machine learning, which is the statistical approach to making intelligent machines. An example is deep learning, which requires huge amounts of data to train. Deep learning has enabled functions such as face and voice recognition. This is used when a person loads photos on Facebook, and an app identifies and labels the people in them. Because deep learning depends so much on data, data is emerging as the most valuable commodity, surpassing oil.
At the University of Johannesburg we have used AI in a range of systems that can:
● Identify epilepsy in a person by reading brain activity;
● Check for pulmonary embolisms by looking at images of a patient’s lungs;
● Allow patients who have lost their voice boxes due to cancer to use their tongues to produce speech instead;
● Monitor the condition of dams and bridges using low-cost cameras and sensors; and
● Perform credit scoring, which is important for allocating loans.
Most tasks that are being carried out by AI were traditionally performed by engineers, doctors, bankers and so on.
Given these advances, what does the future hold?
The world of work will shrink as more and more jobs are taken over by intelligent machines, which will increase inequality. Aggregate demand will drop, depressing the markets. In the first three industrial revolutions, humans invented trade unions to fight exploitation in the workplace. The 4 IR will usher in the problem of human irrelevance, which will have a huge impact on our identity.
The 4IR’s dependence on data is changing the concept of democracy. Companies that collect huge amounts of data are using it to influence national elections, thereby usurping power for their own benefit.
Given all these problems, what is to be done? First, we need legislation that will regulate technologies of the 4IR, including — to protect our citizens — the ownership of data. The law should ld specify regulatory and taxation measures for platform companies such as Uber, Facebook and Twitter.
Second, we should create a system that will incentivise the collection and ownership of data.
Google Maps does not pronounce indigenous street names well because the data used to train its systems is gathered from North America,
Europe and Asia. If we do not collect this data then we shall be locked out of the 4IR.
Third, we need to educate our people so that they are able to understand these developments.
The way education was structured during the first three industrial revolutions is not suitable for this one. Different skills are required; the 4IR demands critical thinking rather than memorising facts.
Because the remaining jobs will require a human touch, empathy and understanding will be important assets. We need to offer education that is multidisciplinary. Those who are studying technological subjects must understand human and social sciences, and vice versa.
Prof Marwala is the Vice-Chancellor and Principal of the University of Johannesburg and the co-author of Deep Learning and Missing Data. This article was first published in the Sunday Times on 23 December 2018.
* The views expressed in this article are that of the author/s and do not necessarily reflect that of the University of Johannesburg