Who is not talking about artificial intelligence?’ asks Ransith Fernando, chair of BCS Sri Lanka. Respectable CIOs are busy adding an AI strategy and road map to make their companies future ready. Most IT students sprinkle a bit of deep learning and some neural networks to make their CVs glow and improve their chances of getting employed.

Each time we search for something on the internet, the results will be based on machine learning applied to your past searches, to give you the most likely content that would match your needs. Machine learning is a part of artificial intelligence. We get suggestions for books, movies and fashion based on machine learning. AI is here to stay; a part of our lives whether we like it or not.

Many years ago, I remember seeing an old cartoon drawn in the 1950s of a future organisational chart. It had a very large computer, a dog and a man. The cartoon was an illustration of a quote by the American scholar Professor Warren Bennis, who stated that ‘The factory of the future will have only two employees, a man and a dog. The man will be there to feed the dog. The dog will be there to keep the man from touching the equipment.’

Science fiction vs fact 

This type of fear where computers take over the world vanished over time. When we hit the millennium, we were less concerned with HAL the famous computer on Dr Arthur C. Clarke’s 2001: A Space Odyssey and much more worried about the societal impact of having a two-digit year in software.

During the last few years, with renewed interest in artificial intelligence, the fear of super-intelligent computers taking over the world has surfaced yet again; it flits through the transoms of many minds each time it becomes visible.

Artificial intelligence is seen as a means of improving the ‘solution quality’ of every current software solution. It is also a solution to a few problems for which no solutions exist at present.

The cartoon of the dog and man is no longer relevant, since the AI is expected to be in the cloud. AI taking over the world is popular with film makers: the favourite plot being where, in the final seconds before unprecedented destruction, a human enters a well-guarded facility and pulls the one plug to stop the super-intelligent computer.

Hasta la vista, baby 

We still like to watch Terminator movies, where the super strong, AI-based intelligent robot keeps running behind humans with the intention of catching and destroying them. Of course, if the robot gets smart enough to catch the humans without running behind them, the movies would be a lot less fun to watch. If the robot used AI to predict human whereabouts and destroy them, the movies would be a lot shorter and extremely boring.

The Terminator type robot falls into the area of general intelligence, while machine learning falls into narrow intelligence. Narrow intelligence is limited to doing a well-defined task much better than humans. However, the task has to be very well defined. Progress in narrow intelligence is being made; machine learning being a good example.

With the recent advances in machine learning, a lot of jobs are expected to be made redundant in both developed and developing countries. The silver lining is that an equal number of new job opportunities are being created. The jobs are not only for the super-qualified, but also for unskilled workers. The expected job loss of unskilled workers may be offset by the creation of new job roles on machine learning based jobs.

The local labour market

Countries that have a large pool of relatively cheap, unskilled or semi-skilled labour would be affected, since a lot of the automation (due to advances in machine learning) is to replace jobs that are routine, repetitive and predictable. The impact will be greater on developing countries where a larger fraction of the workforce is engaged in work of this nature.

An example of possible job losses due to AI is in the apparel industry. The apparel industry employs a very large workforce in developing countries. If advances in AI result in total automation of manufacturing, factories will move back to markets where end-consumers live, cutting down the time from concept to consumer. For companies like Chinese clothing manufacturer Tianyuan Garments Company, who produces clothing for Adidas and Armani, automated sewing technology has enabled them to open their newest factory in Arkansas, and not China.

The rise of the machines 

The automated sewing robots reduce the need for sewing labourers. In the case of Tianyuan’s new factory, three to five people will work each of the 21 robotic production lines. This a labour decrease of 50-70% compared to the 10 workers on a conventional line. In addition to lowering costs, the robots will also increase production. A human sewing line produces 669 t-shirts in eight hours, compared to the robots which can achieve 1,142 t-shirts within the same period. That is a 71% increase in production, resulting in a total output of 1.2 million t-shirts per year.

Using robotics makes the cost of producing a t-shirt in the U.S. comparable to one that is produced overseas. For example, in Bangladesh the labour cost to produce a denim shirt is about $0.22. If made by U.S. workers, that labour cost jumps to $7.47, but with a robotic production line, it’s just $0.33 per t-shirt.

The reversal of job locations due to automation is inevitable. Governments in developing countries will need to face the new reality, accept the job loss and consider encouraging companies to move into new industries that emerge due to AI. One such job opportunity is image classification.

If, for example, machine learning is to be applied to recognise a bird in an image, then a learning dataset has to be created by marking thousands of images that contain birds, and another without birds in the images. The machine will then use this dataset to learn and identify a bird when presented with a new image. To teach a machine to learn, there must be large datasets of images that have marked areas of birds, and so jobs are being created to help produce datasets for machine learning.

Creating new jobs

A human will need to look at given images and mark the areas where certain objects appear - skilled labour isn’t needed for this type of dataset classification. These jobs will be best suited for countries with cheap labour and good internet connections. Therefore, for each of the jobs that are lost, new types of jobs will be created. Gartner predicts that by 2020, AI will create 2.2 million jobs, while eliminating 1.8 million jobs. That is a very small percentage compared to the amount to the workforce.

The major breakthrough on general intelligence - where an AI-based robot with super intelligence takes over all jobs - is much further away than imagined. Predictions can (and do) go wrong. Two years before their maiden flight, Wilbur Wright made a prediction to his brother Orville that man would not fly for 50 years. So, general intelligence may be sooner than we think.

As professional BCS members, we need to spend time learning AI. We should be able to identify hype from reality and be responsible for communicating to the general public and to our governments what is and isn’t possible using AI. We should make them aware of the type of jobs that will get created, as well as the types of jobs that will be lost in the short term.

These are my views on AI; I hope they will inspire you to learn more.

About the author

Ransith Fernando is the Managing Director of boutique data analytics company, Forestpin. He is also the Chair of the BCS Sri Lanka Section.