In 2013 a study of jobs at Harvard revealed that half of the jobs we know today are at risk of being automated by machines. It’s bad enough competing in the job market amongst humans that may be more intelligent or possess a higher qualification than you, but now we have to compete with machines? In an article from BBC, they created a tool where you can calculate how susceptible your job is to becoming automated. Early on in history, people have been worrying about machines replacing men. It was in 1772 that the British writer Thomas Mortimer complained “those [machines] which are intended almost totally to exclude the labour of the human race.” His particular concern was sawmills, which “if introduced into our dockyards etc. would exclude the labour of thousands of useful workmen.” The most powerful branch of Artificial Intelligence, machine learning, is responsible for this disruption.
Machine learning allows machines to learn from data as opposed to being explicitly programmed, thus allowing them to mimic the activities that people do. Machine learning first flourished in the 90’s; recognized as a separate field, it changed its goal from achieving Artificial Intelligence (AI) to tackling solvable problems of a practical nature. Machine learning shifted focus away from the symbolic approaches inherited from AI towards methods and models borrowed from statistics and the probability theory. It also benefited from the increasing availability of digitized information and the possibility to distribute via the internet. In the 90’s machines were assessing credit risk on loan applications and sorting the mail by reading handwritten characters from the zip code. Dramatic breakthroughs in machine learning allowed machines to perform more complex tasks.
In 2012, the Kaggle automated essay scoring competition aimed to develop algorithms that could automatically score students papers. The winning algorithm matched the exact grades that teachers had given papers. Another experiment was done, whereby images of the eye were taken and a series of algorithms were developed to diagnose an eye disease called Diabetic Retinopathy, and again the winning algorithms matched the diagnosis made by a human Ophthalmologist. Given the right data, machines will outperform humans at tasks like this. Teachers may read 10 000 essays in a 40-year career and a human Ophthalmologist may see 50 000 eyes in his career, however, a machine can do these tasks in minutes. We have no chance of competing against machines on frequent high-volume tasks. Are we saying that the occupation of being a teacher or doctor could become obsolete and ultimately be replaced by machines? There is an explanation for this.
Machines can excel at frequent, high-volume tasks, however, they have made very little progress at tackling novel situations. A fundamental limitation of machine learning is that they need to learn from large volumes of past data, while humans don’t. So while you stress that machines will steal all our jobs and make us insignificant, humans possess the ability to tackle novel situations, connect dispersed threads and solve problems we have not seen before.
A prime example of the way humans are programmed to tackle novel situations and come up with creative solutions to problems, is the case of Percy Spencer. Spencer was a physicist working on radar during World War 2. He noticed that the magnetron kept melting his chocolate bar and Spencer put two and two together, connecting his understanding of electromagnetic radiation with his knowledge of cooking, resulted in the microwave. This cross pollination of ideas was a remarkable example of creativity and you may see it as just one of those profound ideas that rarely occur. However, humans perform problem solving tasks in small ways every day. We are designed to solve problems and think out of the box. Machines cannot compete with us when it comes to tackling novel situations, this puts a fundamental limit on the human tasks that machines will automate.
So what does this mean for the future of our jobs? Think about it this way, consider to what extent the job can be reduced to frequent high-volume tasks or the extent to which it requires novel problem solving and creativity. We have learnt that machines are becoming smarter and more efficient at frequent tasks. Machines might conduct our audits and read the provisions in legal contracts but don’t worry, if you are an accountant or lawyer you are still needed. Accountants will be needed for complex tax structuring and lawyers for path-breaking litigation. As mentioned before, machines are not making progress on novel situations. The content of a marketing campaign still needs to grab a consumer’s attention and stand out from the crowd. The content of a business strategy needs to show gaps in the market, things that no one else is doing, thus, it will be humans that are behind our marketing campaigns and developing our business strategies. Human creativity and problem-solving capabilities will always be needed.
If you look back over time, machines and technology have automated many jobs that were previously done by manual labour, but isn’t it the point for machines to replace men in these sorts of routine tasks? After all, this is how our productivity has increased over time and our GDP’s have grown. Take the Spinning Jenny, it was invented in England precisely because wages were high, and thus it was worthwhile for mill-owners to invest in a machine that would allow them to reduce the number of workers needed to make yarn. There is more to it than just job loss, however. Workers in the industrial revolution had previously worked in small workshops. Due to the advancement in machinery and processes becoming automated, there was a shift towards factories as the common workplace. At the time workers revealed the ‘anxiety’ they felt in how machinery was re-shaping the workplace.
During the next few decades the notation of work and whether it is handled by a human or a virtual being will hinge on the predictability. As previously mentioned machines will manage routines while humans will take on the tasks that require problem solving, creativity and flexibility. However, that does not mean machines will not take on more sophisticated tasks and continue to grow their capabilities. Even today, machines take on tasks we couldn’t imagine 20 years ago, did a DJ pick that song or an algorithm? Using Statistical patterns in data, machines can ‘learn’ to improve the efficiency of processes such as customer care and toll collection on highways. As we deploy computers to make our world more efficient, humans will take on more of a problem solving role. Instead of dividing the work between humans and machines, a collaboration between the two will allow both parties to work together to tackle problems at lightning speed.