Preparing for the rise of the machines


Singapore, December 1, 2016


Preparing for the rise of the machines

WHEN AlphaGo defeated Lee Sedol – the world’s top-ranked professional player in the ancient game of Go – earlier this year, the world sat up and took notice. Not only was it the first time a computer programme had beaten a human in the game, it was also an excellent display of how far machine learning could be stretched. AlphaGo analysed data from 30 million moves made by human players in previous games until it could accurately predict human moves 57 percent of the time.

Other successes of machine learning abound. Financial firms, for instance, have been using machine learning to execute high-speed trading, while e-commerce websites are well-known for employing machine learning to recommend products to the consumer based on browsing and buying patterns.


The key drivers behind machine learning

Machine learning as a scientific discipline emerged in the 1990s, at a time when cheaper computing power became more widely available to researchers. Since then, interest in harnessing machine learning to solve problems has grown. 

The dramatic fall in the cost of computing technologies has made it possible for organisations to take advantage of faster computers and storage to crunch large volumes of data.

And data itself is becoming much more widely available. With growing digitisation and greater use of the Internet, more and more data is being collected by governments, businesses and individuals, and this is being used to train machines to make accurate predictions and in some cases, the right decisions. 

These factors have led to dramatic advancements in neural networks, which are a key enabler of machine learning. “Neural networks provide a framework for machines to emulate the way our brains work, such as how we store and process information,” said Mr Liang Seng Quee, Director, Future Technology Office, NCS. 

“The lowest hanging fruits in machine learning applications are repetitive tasks that require minimal decision-making.” - Mr Liang Seng Quee, Director, Future Technology Office, NCS

In IT operations, machine learning can help to recognise patterns in hardware failure - for example, a consistently faulty network router and shut it down. “These simple decisions can be automated to improve the productivity of IT staff who can be more focused on making higher order decisions,”
said Mr Liang.

“While we’ve had some success in harnessing neural networks in machine learning such as making semantic sense of call centre calls through voice recognition, more research is needed before machines can make decisions with a high level of accuracy,” he added.

Emerging applications

Today, machine learning applications are starting to move beyond simple decision-making and gaining traction across a wide swathe of domains. One of these is the area of public safety.

“In public safety, a machine can make use of video feeds and police data, as well as location-based information to predict crime rates in a given area. This will help the police in resource planning,” said Dr Clifton Phua, Director, Smart & Safe City Centre of Excellence, NCS. 

“Another area where machine learning can be put to good use is cyber security. A lot of data being collected about networks including data packets and user browsing behaviour can be crunched by computers using machine learning to detect anomalies and malicious activities by cyber criminals and hackers.” Dr Clifton Phua, Director, Smart & Safe City Centre of Excellence, NCS

Machine learning can also be applied in the area of facility planning. Dr Ying Li, Chief Technology Officer, DataSpark, gave the example of a new sports stadium, where machine learning can be used to identify the flow of people using active data streams from those who are using their mobile phones as well as passive data streams from those who are simply logged on to the cellular network.

“We can use algorithms not only to compute crowd density but also predict routes that people could take to get to the stadium. This would help in deciding the service level at the entrances to the stadium.” Dr Ying Li, Chief Technology Officer, DataSpark

Smaller enterprises can also make use of machine learning in commercial applications. For example, KAI Square, a video analytics software firm based in Singapore, has developed a solution that uses license plate recognition to identify and count the number of cars that enter a car wash shop.

“Video analytics solutions can be used to prevent car wash helpers from siphoning payments, and to track down car owners who did not pay for the car wash services.” Mr Victor Goh, Chief Operation Officer, KAI Square. 

The way forward

As machine learning gains wider acceptance across industries, organisations will have to start thinking about how they can bring together the right technologies and talent and manage issues such as workforce redesign in order to take advantage of these new developments. 

“For organisations that are starting to dip their toes into machine learning, it is important to do so with a clear idea of the problems that they want to solve,” said Dr Phua. They will also have to take steps to ensure the quality of their data, failing which the accuracy of a machine learning application will be compromised.
 
One way that organisations can position themselves for the machine learning age is to set up innovation centres that bring together domain experts including data scientists, developers and business analysts to create machine learning applications. This will help to ensure that the applications meet the needs of the business, said Dr Phua.

Getting the right talent, though, can be challenging in a competitive market. Organisations must be prepared to hire machine learning experts who must continue to improve their skills as the field evolves, said Dr Li of DataSpark. Companies should also promote a data-driven culture - from collating data, to analysing that data to glean useful insights, to taking action based on those insights.

Another area that organisations should look into as they delve into machine learning is human-machine interaction. As Mr Liang pointed out, “People are often enamoured by the fact that machines can do things for us and make decisions, but how do we interact with them and correct their actions as circumstances change over time?” 

Indeed, in the age of machine learning, job scopes may have to be reframed in the context of man-machine collaboration. With improvements in productivity and cost-savings expected as machines take over previously labour-intensive, repeatable or even dangerous tasks from humans, organisations should be looking into ways to better tap on the unique capabilities of human beings to create and innovate.  

As with any technology disruption, organisations should brace themselves for the changes that machine learning will bring. “The nature of today’s jobs will change. Workers will have to be retrained so they can innovate and take on higher value jobs,” said Dr Phua. “For now, it is still humans who decide the problems that are to be solved, and it’s up to us to innovate and figure out how we can work together with machines to improve our lives.”

What is machine learning?

Machine learning is an approach to achieve artificial intelligence (AI) which attempts to program machines to think like humans. 

In machine learning, algorithms are developed to process and learn from data, with the aim of making a decision or predicting an outcome. There are several techniques for implementing machine learning, one of which is known as deep learning. 

In deep learning, attempts are made to mimic the human brain and the interconnections between neurones through what is known as neural networks, which are made up of multiple layers, connections and data flows.

The uses of machine learning are diverse, from chatbots in call centres that take on customer queries to autonomous vehicles and drones that ferry passengers and deliver goods to consumers. 

While science fiction dramas and movies have depicted human-like robots that can think and even feel like humans do, machine learning in its current form still has some way to go before it can take on activities that involve managing and developing people, or that apply expertise to decision-making, planning, or creative work, according to management consultancy McKinsey.

 

Full Article here.

Source:Business Times © Singapore Press Holdings Limited. Permission required for reproduction