Closer to the Machine: Technical, social and legal aspects of AI

The field of artificial intelligence (AI) has made amazing advancements in recent years. What was once science fiction is now becoming reality. Everything from restaurant recommendations to dental diagnoses are increasingly being powered by AI. But alongside the many benefits of AI, issues such as privacy, discrimination, accountability and transparency have surfaced. The full potential of AI, both good and bad, is looming on the horizon.

The world is currently considering how best to address these issues. In light of this, OVIC has brought together eight experts from different disciplines to explore the benefits and issues of AI, particularly in a public sector context, in the form of an e-book: Closer to the machine: Technical, social and legal aspects of AI.

Some of the key themes explored in the book include:

• AI is already generating positive benefits for all of us. It is making information more accessible, improving medicine by designing new drugs, detecting fraud, making energy generation more efficient, anticipating maintenance needs for infrastructure, and providing personalised services and strategic advice for policy decisions. But it is debatable if all the ways in which AI can be used are ‘good’. Predictive policing, jail sentencing and welfare distribution are all things that we can delegate to AI, but what are the implications of doing this?
• Human decision makers are increasingly being replaced by automated systems. But while humans are accountable for the decisions they make, the same cannot always be said for AI systems. Humans can justify and explain how they reached a decision. AI systems, on the other hand, make decisions based on statistical predictions from algorithms that are usually complex to the point of being incomprehensible to humans. This is a critical issue for government use of AI – government is expected to be transparent, and if the rationale behind an AI decision cannot be explained, that decision can hardly be described as transparent.
• AI is challenging existing laws and regulation, but those challenges are not insurmountable. While it may be intuitive to jump to making new and prescriptive legislation as a response, we need to first look at existing laws, such as those around product liability and consumer protection, because it might be that those laws are already sufficient to deal with AI. It may simply be a matter of applying those laws more vigorously, or producing more guidance on their interactions with AI. Alternatively, they could be adapted in nuanced ways to work with AI, rather than being circumvented entirely.
• The information that is used to power AI generally comes from the real world, but that information reflects the unavoidable biases and imperfections that come with real people. AI systems, and all the decisions and predictions made by them are influenced by the unavoidable bias that is in all data. Making unbiased and fair AI is as hard as making unbiased and fair people.
• The rapid advancements in AI pose challenges to the security of systems, leading to a cat and mouse game of vulnerabilities followed by fixes followed by more vulnerabilities. It is crucial that those who deploy AI are aware of the many constantly evolving security considerations with using AI in practice.
• The field of AI has been described as ‘alchemy’, in that people make AI systems, which often work wonderfully, without fully understanding how. They rely on tinkering and trial and error to get the results they are looking for. This has led to problems such as reproducibility, where one researcher cannot reproduce the results of another, and has likely exacerbated the transparency issues of AI.

On 30 August OVIC will launch Closer to the machine with an event consisting of presentations and panel discussions. We hope that this book is a valuable resource for anyone who wishes to increase their understanding of AI systems, the benefits that they can bring, and the technical, social, and legal aspects that must be considered when looking at AI.

Read or download your free copy of Closer to the machine here.