Artificial Intelligence: Opportunities, Risks and Implications

Artificial Intelligence, or AI, is making huge strides today, around the world and in Singapore.

The rapid developments in AI have engendered high expectations of the benefits it can deliver for business and for society at large. Yet there are concerns about AI replacing human input and thus displacing jobs.

In 2017, Elon Musk, the CEO of Tesla, sparked a debate when he said that AI poses a “fundamental risk to the existence of civilisation,” and called for regulators to be proactive rather than reactive to developments in AI.

On the other hand, others like Oren Etzioni, CEO of the Allen Institute for Artificial Intelligence, warned Americans that if they took steps to slow down progress on AI, other countries like China would overtake them.

In this publication, we examine the opportunities, risks and implications of AI use, particularly in the context of the accounting and finance industry, and how we can equip ourselves to deal with such future scenarios.

What exactly is AI?

AI is a rather broad term that has been bandied about in the media. At some level, data analytics and AI have considerable overlaps. One of the goals of data analytics is to interrogate data to obtain insights in order to make predictions, as in predictive analytics. Machine learning can be seen as an evolution of predictive analytics, albeit with vastly more advanced state of the art techniques and knowledge such as the use of neural networks.

Most of the AI application we hear of today is machine learning, where predictions are made by a computer system based on its exposure to data. In other words, the computer system automatically spots patterns from data sets; then, depending on the choice of algorithm, it will present a possible interpretation for a decision to be made. An algorithm is a set of rules used for the analysis.

Then there is deep learning which, as its name suggests, presents a deeper and higher level of machine learning. It imitates how the human brain works, in creating new patterns of data for decision making. It can “learn” without supervision, from unstructured data – data that is not already neatly organised according to pre-determined categories.

Robotic process automation (RPA) is a form of business process automation, which some say is not AI, strictly speaking. For example, RPA can be used by a financial institution to automate the processing of credit card application

One way to explain the nuanced distinction is that RPA mimics human behavior and actions, while AI mimics human thinking.

For today and the foreseeable future, as David Leow, Managing Director of Thaler Global, explains, the AI we have is termed “weak AI” and would fail the Turing test for strong AI, as their capabilities are easy to distinguish from those of humans. For example, current versions of Siri, Alexa, and Google Assistant have impressive voice recognition and are quite adept as scheduling appointments, but a real human assistant can think ahead to resolve logistical issues such as booking transport, and make and distribute concise meeting notes.

Neverthless, advances are being made fast enough that it may not be long before a stack of receipts and bank statements could be machine translated into a set of financial statements and tax returns ready for filing. As the goal of strong form AI is brought closer to reality, we will be able to take increasingly raw inputs to get increasingly more useful outputs.