The thinking computer: cognitive technology in the public sector

©iStock

As the Budget made clear, the only way to improve services while spending less is to increase productivity. Could cognitive computing help?

Cognitive computing, also called artificial intelligence or AI, involves using self-learning systems to mine big data, recognise patterns and process natural language. The aim is to copy how the brain works while processing information and making better decisions more quickly. Cognitive computing is also claimed to be capable of anticipating problems and modelling potential solutions.

An example is Watson, developed by IBM. In 2011, it took part in US game show Jeopardy, beating two former winners to take the $1m prize. To achieve this, Watson had access to 200 million pages of structured and unstructured data, including the full text of Wikipedia, but was not connected to the internet. It cost IBM a lot more than $1m to get Watson to achieve this feat.

The power to boost productivity comes from the growing ability to process huge amounts of unstructured data – text, images, voice files – and to communicate with people using ‘chat’.

You have probably come across virtual personal assistants such as Apple’s Siri and Microsoft’s Cortana, which use this type of technology to help you find information using your voice. Google is using AI so self-driving cars learn to drive as we do, through experience. Many banks are deploying it to spot fraudulent activities and to ask customers to confirm details before completing transactions.

Cognitive computing could offer huge potential in areas such as healthcare and regulation where there is a lot of documentation from many sources. It can, for example, be difficult for doctors to keep up to date with research and techniques. A system that allows them to ask questions, synthesises the latest thinking and gives precise answers can be very powerful.

The Memorial Sloan Kettering Cancer Center in the US has been trialling this type of technology with a health insurer since 2013 for lung cancer treatment and 90% of its nurses in the field now follow its guidance. Some hospitals in the UK are now testing such approaches. In Singapore, a virtual assistant called Jasmine helps people to navigate its tax system and file returns.

We are, however, still at the very early stages. Watson started swearing when taught slang to sound like a real person. Microsoft’s Tay, a bot intended to mimic a 19-year-old woman using social media, learnt the language it encountered; within 24 hours, it became an embarrassment by making lewd and racist comments.

At present, cognitive systems are mainly being used to automate low-end, low-value conversations. The next generation could dramatically reduce the number of staff needed for all sorts of tasks including surveillance, driving, planning, healthcare and even report writing.

They are also beginning to be used to determine priorities and allocate resources, without emotion and biases, such as scheduling daily tasks for the 1,000 maintenance staff working on Hong Kong’s Mass Transit Railway.

The experts promoting these technologies tell us that the productivity improvements are limited only by our imagination.