How AI threatens white-collar jobs
- Algorithms using machine learning and image recognition can now work on increasingly complex tasks and compete with highly skilled professionals.
- Doctors, bankers and lawyers are among those who will have to hone their skills to work effectively with machines.
Which jobs will be replaced in the future? The faster artificial intelligence (AI) and robotics progress, the more brutal the response. A 2018 report by the Organisation for Economic Cooperation and Development (OECD) estimates that 14% of all jobs in developed countries are at risk of being automated. An earlier study from the University of Oxford, published in 2013, put at 47% the number of jobs threatened in the US.
Machine learning and image recognition have developed at breathtaking speed, says Pierre Vandergheynst, vice-president for Education at the École Polytechnique Fédérale de Lausanne (EPFL). “The technology for these applications has been around for quite some time,” he adds. “But today artificial intelligence has access to big data. Using millions of images, sounds or texts, it can analyse information and even solve problems.” That is why algorithms using machine learning and image recognition can now work on increasingly complex tasks and compete with highly skilled professionals. The job threat extends to some unexpected fields.
The speciality in which image recognition has the greatest potential is radiology. Comparing thousands of chest X-rays, for example, an algorithm developed by a US research team correctly detected tuberculosis in 96% of cases. Another recent study demonstrated that AI can diagnose health conditions using CT scans with 91% accuracy. These new opportunities have already had an impact in France. In June 2018, the French Society of Radiology announced that it would set up a data bank containing 500 million radiology files to develop AI-based image analysis. Meanwhile, a Google subsidiary has designed an algorithm that can predict an individual’s likelihood of developing heart disease using retinal images.
Financial algorithms are now used to manage portfolios, assess investment risk and oversee compliance with regulations. A study by the auditing firm Deloitte reported that 32% of financial institutions currently use AI, while another 40% are considering it. The CEO of UBS has stated that the Swiss bank could cut staff by 30% within the next 10 years thanks to “technological advances”.
Trying to impress someone in a job interview will soon be a thing of the past. More and more companies are opting for algorithms like HireVue to select the best candidates. HireVue’s software uses AI to assess video interviews of applicants. The programme then draws up a profile based on more than 200,000 criteria, such as tone and the words chosen in certain situations. Only after this first selection process is complete do human recruiters come in. More than 600 large firms already use this service, including Goldman Sachs.
A 2018 study showed that AI could detect errors in several contracts simultaneously in just 26 seconds, achieving 94% accuracy. It took a group of 20 US lawyers 92 minutes to perform the same task – with an accuracy rate of only 85%. When the job involves reviewing large amounts of text based on specific criteria, humans cannot compete with algorithms that use machine learning. In another example, researchers from University College London asked artificial intelligence to predict the outcomes of several hundred cases judged at the European Court of Human Rights. The algorithm got it right 79% of the time.
But there is no reason for professionals in these fields to despair. According to Ole Winther, professor and machine-learning expert at the Technical University of Denmark, these examples reflect mainly how jobs are changing. “Humans will have to learn how to work with machines,” he says. “Let’s take the case of radiologists. Algorithms will help point them in the right direction in their analysis and diagnosis.” He compares the development of AI to the advent of computers in the workplace. “Not everyone has to learn programming. But radiologists need to be able to express exactly what they want the machine to do. Similarly, lawyers should use algorithms like an assistant who speeds up the process of document analysis.”
At EPFL, the idea has been incorporated into the curriculum. All students now take a computational thinking course to learn how to work alongside machines. “That will be a crucial skill for most jobs over the next several years,” says Vandergheynst. “Meanwhile, schools and universities should encourage students to be creative. Humans are not about to be taken over by algorithms when it comes to creativity.”