Date posted: 27/07/2018 8 min read

AI at an ethical crossroads

In a world where robots will one day write code for themselves, the rise of artificial intelligence and machine learning creates new ethical dilemmas, according to a new paper by CA ANZ.

In Brief

  • Accountants can help to test and build artificial intelligence industry standards for design, audit, algorithms and transparency
  • The EU is to draft AI ethics guidelines by the end of 2018 and Australia is to fund an ethics framework
  • Attend WCOA 2018 to gain insights into emerging technology and ethical issues

By Karen McWilliams FCA

From smart cars to smart phones, artificial intelligence (AI) has invaded every aspect of our lives. In the digital age, this new technology is raising ethical issues unlike any we have had to consider before.

To date, the media focus on AI has been characterised by two extremes. At one end, the focus is on the tremendous benefits and exciting opportunities AI can deliver for how we live and work. At the other end, robots are going to take over our jobs in a world of big brother surveillance. These differing perspectives show the need for an ethical framework to help shape developments. 

The fourth revolution

Machine learning has been referred to as the fourth industrial revolution. Machine learning is a subfield of AI which is focused on designing systems that can learn from and make decisions and predictions based on data. 

Recent advances in machine learning now bring us to an ethical crossroads where we need to decide the role AI and machine learning will play in shaping our future. To present a more holistic snapshot of this fast-paced technological movement, Chartered Accountants Australia and New Zealand has published Machines can learn, but what will we teach them?

“With the rise of artificial intelligence and automation, we are fundamentally changing everything about the way we live, work and understand our world,” Sarah Adam-Gedge, CA, Managing Director, Avanade Australia, says in the report.

According to the report, the rapid progress being achieved in AI means that super intelligent machines are now seen as the next development stage, where these machines will learn to write code for themselves. “This is how machines can possibly become independent of programmers and where, perhaps, even greater risks lie,” the paper says.

To date, however, there are no commonly agreed policies or accountability frameworks. Yet in the last two years we have seen the widespread use of drones, fingerprint technology, facial recognition, driverless cars and other significant AI breakthroughs.

Overall, the paper, authored by Attracta Lagan, principal at Managing Values, considers the ethical implications of AI and machine learning from different perspectives including that of society, business and the accounting profession. To inform the paper, key industry figures were interviewed across Australia and New Zealand to gain their insights. 

Ethical concerns

At the societal level, we need to consider how we will overcome issues such as the transparency and bias of algorithms. For business, there are a number of ethical concerns about designing and implementing technology, including fairness and privacy. In addition, interventions will be needed to transition the workforce to an AI-enabled economy, including upskilling and reskilling. The potential for humans to work alongside intelligent machines will provide the greatest opportunities for both increased productivity and increased human satisfaction from the new services and products that can be designed. Tomorrow’s business world will need to develop and nurture a balance of artificial and human intelligence.  

How accountants can help

The accounting profession has a pivotal role to play in ensuring that business information is sound and that business decisions are in step with wider societal values. They can help to build consensus around AI industry standards for design, auditing and transparency, as well as identifying techniques to increase public trust in these new technologies. 

“When organisations make assertions about an algorithm, the role of the auditor will be to test those assertions to ensure the algorithm does what they asserted it did.”
Peter Williams FCA Chief Edge Officer, Centre for the Edge at Deloitte Australia

As Peter Williams FCA, Chief Edge Officer, Centre for the Edge at Deloitte Australia notes: “AI can show us things, but we need humans to identify what we do about it. When organisations make assertions about an algorithm, the role of the auditor will be to test those assertions to ensure the algorithm does what they asserted it did. Software code is not infallible, mistakes can happen and on a vast scale. Accountants are in the box seat to continue to act as trusted advisers to interrogate the systems and processes that underpin the acquisition, management, analysis and disposal of this information.”

The profession will need to commit to continuous learning in the area of AI to ensure it has the expertise and knowledge to meet the fundamental ethical standards including duty of care and competency.

Ethical frameworks

There is also a need for a global agreement around developing an ethical framework for AI and we are already starting to see progress. For example, in April 2018, the EU set out an initiative on AI, which includes developing draft AI ethics guidelines by the end of the year. Also in April, the UK’s House of Lords Select Committee proposed setting up a cross-sector AI Code and suggested five principles to form the basis of the code. New Zealand has also called for more action. In May, the Minister for Government Digital Services and Broadcasting, Communications and Digital Media, Clare Curran, said an action plan and ethical framework is urgently needed to educate and upskill people on AI technologies. She made the comments when launching Artificial Intelligence: Shaping a Future New Zealand. 

Where Australia is at

In the 2018/19 Budget, the Australian government allocated A$30 million to grow Australia’s capabilities in AI and machine learning, as well as to establish a new national Ethics Framework and Standards Framework "to guide the responsible development of these technologies". Shortly after, Australia’s chief scientist, Alan Finkel, proposed the creation of a voluntary “Turing certificate” to enable consumers to identify businesses or products whose ethical standards and processes for AI have met a required standard subject to assurance. 

However, the development of ethical frameworks is typically slow compared to the pace of technological change, as Professor Nicholas Agar of Victoria University of Wellington points out: “Our pace of ethical reflection tends to be slow and deliberate, typically slower than the pace of technological progress. We need to have open conversations around the ethics of AI and share different ethical perspectives. These open conversations should address many different scenarios about what hasn’t happened yet but could in the future. We need to think creatively about what we are and can be. That way we can make regulations that protect what we really care about.”

It is clear that we have reached the point where we need to start considering these ethical implications so that we can choose the right direction when the time comes. Join us at the World Congress of Accountants (WCOA) in Sydney in November as we explore the new dimensions of ethics and trust in this digital age.

Karen McWilliams FCA is the Business Reform Leader at Chartered Accountants ANZ

Machines can learn, but what will we teach them?

A recent paper by Chartered Accountants Australia and New Zealand offers a timely warning that we are at an ethical crossroads where decisions need to be made about how AI will shape our shared futures.

Read the paper

World Congress of Accountants

Join us to explore the new dimensions of ethics and trust in a digital age

Find out more

Available from Your Library: How we can build AI to help humans, not hurt us

Cautions against the biases and blind spots we subconsciously program into artificial intelligence. Argues that if we want AI to help humans, we need to define the goals and strategies to do that now, or we risk replicating human prejudices in technology.

Read more