The shifting future jobscape is a good opportunity for the Indian government to invest in sectors like tourism and handicrafts — sectors that are unlikely to be affected by ML even in the long run.
Will artificial intelligence and machine learning lead to a huge loss in jobs, as the computers take over the tasks that keep so many Indians (and people around the world too) employed? Given the jobs crisis in India, this is a very important question, as is finding the balance between innovation and security.
The interplay between new technology and labour is as old as, well, technology and labour. The classic example of this is the impact of the steam engine on the labour market during the Industrial Revolution. During the Industrial Revolution machine power replaced muscle power and led to rapid increases in productivity of large enterprises, crushing out smaller firms. The Internet has had a similar effect — enabling large numbers of entrepreneurs while rendering many local businesses obsolete. Each technological revolution both replaces a large number of pre-existing jobs as well as enables new businesses that create new jobs. The expectation is that Artificial Intelligence will be no different. If the steam engine replaced muscle power, Artificial Intelligence is aimed at replacing brain power. The big question ahead of us is: will more jobs be created or destroyed by Artificial Intelligence (AI) and Machine Learning (ML)?
We first need to recognize that digitalization and machine learning are two different, though related, technological changes. Digitalization has led to online platforms where employers and consumers can connect directly to the labour market. Take for instance Uber, Swiggy, BigBasket in the Indian context. These platforms have replaced local vendors on the one-hand but on the other, have enhanced efficiency of service and created a number of new jobs. They have also changed the nature of work, from full-time to on-demand. However, merely replacing the local with the online, has nothing to do with with ML and these jobs (lost or gained) do not fall into the accounting that we’re interested in.
The destruction of jobs
For us to confidently point a finger towards ML, there has to be some “learning via algorithms” of a simple cognitive human task involved. Take for instance, the task of spotting an abnormality in a retina scan which may put a patient at risk of a serious eye condition. Previously this task required a trained human (an ophthalmologist), now such tasks are being increasingly taken over by machines. Another example: suppose you want to complain about a food order over a phone call, previously handling your calls was a task assigned to a human (a customer care executive), now you will likely end up “chatting” with a machine learning powered chatbot.
What makes ML a particularly efficient technology is that certain algorithms are becoming context-agnostic, meaning that some powerful algorithms perform very well across a wide variety of applications. For instance, the recent advancement in algorithms powered by Deep Learning and Neural Networks is being applied to fields as disparate as medicine and language processing. This makes the scale and scope of ML applications is unprecedented. Add to this the wealth of data being collected (due in large part to the digitalization of the economy) and growing computing power―and it’s no wonder that ML breakthroughs are occurring across fields at a furious pace. This explains why fears of job losses due to machine learning feel so real and urgent.
So will machine learning inevitably lead to the destruction of jobs that require simple cognitive ability? The answer is: to a large extent, yes. In fact multiple papers and reports have found that the greatest victims of ML related job losses are those who perform medium-skill and routine cognitive tasks, for example entry level accountants and customer service executives. In fact certain high-skill jobs requiring sophisticated pattern matching are under threat too. For instance, the National Health Service in UK is hoping to use AI tools to make up for the shortage of radiologists, which signals the loss of such jobs for humans in the longer run.
New kinds of work will emerge
What about the creation of new jobs directly because of ML? Here the answer is a bit mixed. Yes, there will be some new jobs created as a direct result of ML, but these may not be around for the long run. For instance, with algorithms replacing the job of a radiologist, automated medical diagnosis kits may proliferate to remote areas, where the patient would need a local individual who can explain the diagnosis to her. Another set of jobs being created by machine learning is that of the “annotator”. In order to produce accurate predictions, algorithms need large amounts of annotated or labelled data. An annotators jobis to do just that ― label the data. This could be labelling whether a photograph has a car in it, whether a handwritten note contains the number 4 in it, whether an audio recording contains the word “hello”. These jobs are set to proliferate, especially in low wage economies like India, however they are ephemeral in nature and may very well be the storm before the lull, as far as the job market is concerned.
Does this mean that we are staring into a future with little or no work for humans? Economists in the US have recently started studying the impact of automation (including digitalization, artificial intelligence and machine learning) on the labour market. Not surprisingly, one set of high-demand roles would include individuals who are highly skilled in developing and using ML algorithms. However, their findings suggest that the most widespread jobs of the future will be those which require “human qualities”, something which machines are still a long way from acquiring. Some examples include tourism professionals, teachers and nurses. Personal care and creative arts are also domains where ML is not a direct competitor to humans, at least not yet.
The Indian labour market
Indian policy makers would be foolhardy to ignore the potential effects of ML on the labour market. Joblessness has emerged as a key concern among voters in the run up to the 2019 general elections, and addressing the opportunities, as well as threats that ML brings is sure to creep onto the next government’s agenda. On the upside, much of the job loss fear is driven by the countries at the forefront of R&D—that is to say, we don’t expect self-driving cars to replace human drivers anytime soon in India. Having said that, the shifting future jobscape is a good opportunity for the Indian government to invest in sectors like tourism and handicrafts — jobs in these sectors are unlikely to be affected by ML even in the long run. On the flip side, our education system is just not equipped for the eventual labour market changes. With routine tasks becoming the greatest casualty of the ML wave, training the next generation of the Indian workforce in soft and cognitive skills needs to become a policy priority.
Across the political spectrum in India (and in fact the world over), conversations around Universal Basic Income are gaining prominence. It is no coincidence that this is the era of successful scalable AI and ML models. As machines start replacing simple repetitive human tasks, we cannot ignore the very real possibility of a future with rampant joblessness. The idea of universal basic income gains credibility when looked at from the lens of tackling the extreme economic inequality that would ensue. In an ironic twist, it may be that a simple capitalistic idea — increasing productivity via machines, is likely to provide the best argument for a large scale socialistic redistribution system.