Artificial intelligence (AI) technologies are designed to replicate human capabilities, and in some cases improve upon them. Lifelike robots are physical examples of AI technology, but it is the digital AI systems that already have a ubiquitous influence on our daily lives – from facial recognition software to decision-making tools used by banks, recruiters and the police. Too often, these systems can reflect preexisting social inequalities.

In this episode of the Physics World Stories podcast Andrew Glester investigates the ethical issues that can plague AI and machine learning technologies. He finds out about the concepts of deep learning and neural networks, why these systems can amplify problems in society, and who are the people adversely affected by these flaws.

It turns out that the physics community is part of the problem and potentially part of the solution. Directly and indirectly, physicists are involved in developing AI technology so are ideally placed to raise awareness of the issues. Featuring in the episode:

• Alan Winfield, a robot ethics researcher at the University of the West of England
• Julianna Photopoulos, a science writer based in Bristol, UK
• Savannah Thais, an experimental particle physicist at Princeton University, US

To find out more about the issue of bias in AI systems, take a look at this feature article by Photopoulos, which is summarised in the video below.