Artificial intelligence (AI) will undoubtedly have a key role in how pandemics are managed in the future. Now, researchers have created a concrete plan for how the technology should be developed to get us there.
The COVID-19 pandemic turned our lives upside down – from testing centres to empty streets and closed schools. But next time, we can be better prepared – and AI could be the key.
In the next pandemic, AI might spell the difference between chaos and control. That is why researchers have created a blueprint for how the technology should evolve in the years ahead.
When millions of data points are collected - such as test results, mobile phone movements and disease rates – AI can find patterns and help decision-makers react quickly and purposefully, such as whether targeting specific population groups with a testing strategy is worthwhile.
However, before we get there, AI must first be developed in that direction, and researchers have created a plan for how to do that.
“Previously, models with fixed formulas and assumptions were used to predict how an epidemic will develop. With AI, we can instead find patterns directly in large quantities of data without knowing the rules in advance, which enables more flexible and situation-specific decisions,” explains a researcher behind the plan, Samir Bhatt, Professor, Section of Epidemiology, Department of Public Health, University of Copenhagen, Denmark.
Designing vaccines at digital speed
In the new article, the researchers present a concrete plan for how AI will be used in vaccine development, infection tracking, decision support and other things – with technologies such as language models, protein analysis and simulated networks.
A natural starting-point is vaccine development.
Researchers around the world are already using AI to analyse the surface proteins of viruses and predict where the immune system can best attack. This has revolutionised vaccine development and enabled vaccines to be designed that work precisely and have fewer side-effects.
So when the next epidemic strikes, AI could help to design a vaccine that offers the best protection with the fewest side-effects.
AI can trace chains of infection – like a digital detective working in real time.
During the COVID-19 pandemic, there was a strong desire to pinpoint outbreak sources. It does not take many cases to overwhelm the system – but tracking who infected whom requires enormous computing power, which AI can deliver without losing oversight.
“For example, you can simulate how one person spreads infection to others through the workplace, public transport and their household. AI can show this chain reaction in near real time and help authorities to intervene accurately and quickly. This will make it much easier for Statens Serum Institut and others to make decisions and recommendations in real time,” says Samir Bhatt.
Smarter decisions when crises hit
If an epidemic gets out of control, AI can also help decision-makers make optimal choices.
Large language models – such as ChatGPT – are advanced AI systems trained on vast collections of text. They are designed to understand and generate language that sounds convincingly human and are used to help decision-makers weigh strategies – such as whether to close kindergartens, whom to vaccinate first and how best to protect vulnerable groups – without shutting down society.
Large language models can also make regional recommendations, all with the aim of following a strategy such as the fewest deaths or least impact on daily life for the most people.
A critical goal is to maintain public trust – ensuring that people continue to believe that the authorities are acting in the public’s best interest, even under pressure. That way, a pandemic does not spiral into widespread distrust, as seen during COVID-19.
“AI can even help us to identify what questions we should be asking ourselves and AI," notes Samir Bhatt.
Knowledge where it is needed most
The use of AI as part of the epidemic toolbox can also take different forms across borders.
In a country like Denmark, where the amount of data on Danes is enormous, AI can help keep track of data and utilise it to make better decisions.
In low- and middle-income countries, AI can help fill the role of missing capabilities, such as a lack of epidemiological experts or vaccine researchers.
“Today, you can send experts to low- and middle-income countries in a crisis situation, but the expertise stays with them and then disappears again when those people leave. With AI, low- and middle-income countries can act quickly and accurately themselves – without being dependent on outside experts,” explains Samir Bhatt.
A global roadmap for future pandemics
Samir Bhatt says that the future scenario outlined in the Nature article is already on its way to being realised.
The necessary tools are already being developed. With this plan, he and his colleagues aim to unite global research and chart a shared course toward stronger pandemic preparedness with AI.
“This will happen in the coming years, but there are also some challenges to overcome. For example, many large language models are black boxes – we can see what is coming in and going out but not what is happening inside the machine at the time. This makes understanding and trusting the decisions difficult. And then some language models hallucinate and come up with answers that sound right – but have no basis in fact. These hallucinations can be dangerous in a crisis situation,” says Samir Bhatt.
However, he is convinced that these problems will be solved in the near future.
Large language models are still young – but their potential is immense. If we solve the challenges in time, AI could become our most powerful ally – and the key to the roadmap that guides us through the next pandemic.
