Is AI breaking democracy?

As an engineer, I’ve worked on the safety of software-controlled systems for around 40 years and for the last decade that has included artificial intelligence. My perspective is that the question of whether democracy can survive, is dependent on the extent to which we can safely harness, manage and regulate AI.

What do we mean by democracy?

Let’s start with what we mean by democracy. Democracy is a system of government where political power is vested in the people, exercised either directly or through elected representatives. When I talk about democracy surviving, I don’t mean will we continue to have elections, I’m sure we will. I mean governments acting for the public good.

This requires two things: aligning what is genuinely good for the public with what is politically achievable and retaining the support of the electorate.

On the first of these, what is the public good? There is no simple answer and perhaps because of that complexity, we have seen a rise of single-issue parties, arguably the Greens and Reform among them.

How AI distorts public opinion 

Social media and AI are powerful factors here, helping to create “echo chambers” where specific views are reinforced and those who hold differing views are demonised. Cookies on websites and AI are good at personalisation of data and so we see adverts that try to sell us particular products, reflecting our recent browsing, or information that shape our views, likely making those views narrower and more polarised.
 
AI tools are unreliable but they are designed to be convincing even when wrong and can be sycophantic, reinforcing polarised views. More perniciously we have AI generated “deep fakes.”

The result is that AI and social media are having a profound impact on our ability to have an honest conversation about the public good. Without that, democracy cannot function as it should.

AI as an agent of growth 

At this point we should turn to economic growth as it is economic wellbeing that sustains public support for democratic institutions. Put simply, the better you feel about your own economic position the more you will be prepared to accept redistribution to help those in need or to fund other activities for the public good.

The UK has economic challenges, including stubbornly low GDP per capita, but AI has the potential to be an agent of growth. The OECD estimates that UK labour productivity growth from AI could reach 0.4–1.2 percentage points annually over the next decade, placing it second only to the US among G7 economies. Other figures suggest a £500bn boost by 2035.

Part of the reason for this is that knowledge intensive services account for about a quarter of UK GDP, however the use of AI is significant for physical systems, including self-driving vehicle, robotics, and critical national infrastructure, albeit with a need to assure safety.

There are enormous potential benefits, but there are risks too. The regulators have a massive task to keep pace with this fast-moving technology, something we at the Institute for Safe Autonomy and Centre for Assuring Autonomy are trying to help them with. 

There are other forms of risk from large-scale take-up of AI. First, the major AI developers are huge. Alphabet (which owns Google and DeepMind) and Nvidia have greater turnover than countries such as Finland and Chile, and there have been suggestions that nations will be denied access to such technologies if their laws aren’t favourable to the companies.

Second, data centres which train and run AI models consume a significant percentage of electricity generation – perhaps 1.5 to 2% of global energy consumption. They are expected to grow significantly in the next five years, stretching the capacity of the UK electricity network.

Regulation is key

Engineers like to provide solutions and I see regulation and regulatory innovation as the key.

First, regulation needs to control the adverse impacts of AI, including physical safety risks and activities that can distort public opinion. I believe that the best approach is a national institute that supports all the major regulators to provide consistency and to provide a locus for international harmonisation as the problems that need to be addressed are trans-national.

Second, good regulation can stimulate growth, but that has to include accelerating the introduction of new energy generation capacity as well as support for AI itself. I’m supportive of the mission of the Regulatory Innovation Office which is working to adapt regulatory mechanisms to enable growth. 

However, there are potential conflicts here. Controlling AI as well as encouraging its safe use takes us back to the question of the public good and what is politically achievable, which is a complex ethical trade-off.

So, for democracy to survive we need mature conversations about the safe, socially acceptable and sustainable use of AI, including how we power the data centres it depends on. I think it is essential to have a National Centre of Excellence to support this public debate.

Professor John McDermid is Co-Director of the Institute for Safe Autonomy and Director of the Centre for Assuring Autonomy