Four things you should know about machine learning

How do babies learn things in the first few years of their lives? They observe and imitate everyone around them. But by the time they’re about 18 months old, they can infer intent as well.

If Mummy is stacking blocks that keep toppling over, Junior can still get what she is trying to do, even though she doesn’t succeed. Just like human babies, computers can teach themselves.

This field is called machine learning, and it has become a big trend in the public sector. Here are the four things you should know about machine learning and how governments are using it.

What is it?

Machine learning allows software to learn patterns from large amounts of historical data, which can then be used to predict future patterns in the data.

For instance, feed a machine learning algorithm data on past commuter travel patterns, like starting and ending locations, journey duration and time of day. The algorithm uses this data to form a model of where commuters travel to at certain times of the day and how long it takes.

It would then be able to predict how long it will take to travel from one location to another at different points in the day. These algorithms are not always accurate from the start, but “practice makes perfect”, as they say.

As they are trained with more data, they learning from their mistakes and correct their models when they make a wrong prediction.

Machine learning is often confused with the term ‘artificial intelligence’. Rather, it is a subfield of artificial intelligence, which is an umbrella term for any technology that can do what was previously uniquely associated with humans.

This confusion is unsurprising, since “modern artificial intelligence research is denominated by machine learning methods,” Lu Wei, Assistant Professor at the Singapore University of Technology and Design, tells GovInsider.

Why is this important?

Machine learning algorithms allow us to automate data analysis, which is ordinarily challenging for humans to perform due to the massive amounts and complexity of the data involved, Lu says.

Automation is not new, but the level of complexity was beyond human reach until machine learning came along. As Jeff Bezos, CEO of Amazon, recently put it: “Over the past decades computers have broadly automated tasks that programmers could describe with clear rules and algorithms.

Modern machine learning techniques now allow us to do the same for tasks where describing the precise rules is much harder.” “Machine learning algorithms can be useful in many practical tasks, such as extracting structured information, making predictions, finding hidden relations within data, and detecting anomalies,” Lu says.

For instance, “we can build machine learning algorithms to rapidly extract structured information from unstructured data such as texts from forums, news articles and social media,” he says.

These texts can be further analysed to find public sentiment towards particular organisations and products, he adds.

How does it affect government?

Governments are using machine learning to counter the shortage of labour or skills in industries such as healthcare, and improve productivity.

It also allows government to develop more accurate policies by predicting public response to changes.

“With the availability of massive amounts of data, useful information can be inferred, which will be helpful for making optimal and timely decisions,” says Lu.

He adds that “stable and continued funding support” from governments will allow universities and research institutions to carry out long-term research on machine learning, which can lead to commercial products that benefit society.

Machine learning can also be a powerful tool for management within government, Stephen Goldsmith, the Daniel Paul Professor of the Practice of Government at the Harvard Kennedy School, writes in an article for Government Technology magazine.

Since machine learning algorithms can infer and make future predictions based off of vast amounts of data, this ability can also be applied to analysing and predicting human behaviour. Supervisors and officers can use this information for “recruiting, training and management purposes”, Goldsmith writes.

What is happening in the region?

Machine learning has become a trending topic for government leaders across Asia. Singapore’s GovTech chief executive, Jacqueline Poh, tells GovInsider: “For 2017, we are developing our machine and deep learning capabilities to do more in the area of video analytics, unstructured data and cognitive assistance with chatbots and other AI-related tools.”

The Singapore Land Authority, for instance, is using machine learning and drones to identity unused state property that need maintenance, like fixing cracked walls, water ponding and illegal dumping.

And the city’s Changi Airport has built a S$50 million Living Lab using machine learning to improve customer experience, Lu from SUTD says. Meanwhile, Malaysia’s Government CIO said that machine learning is a technology “we find relevant because it can improve the way we deliver services to citizens”.

In the US, machine learning is being used to identify police violence. In August last year, the Charlotte-Mecklenburg Police Department in North Carolina started piloting a system that catches signs of officers at high risk of racial profiling, for example, New Scientist reported.

The algorithm analyses all the records of an individual - including gun use and previous misconduct - to identify officers likely to resort to violence.

It is evident that the machine learning wave is fast spreading across the world. To make a prediction of our own, we can imagine that soon, more and more governments will come on board.