• A global competition has encouraged powerful predictions for what should happen to mitigate the effects of the pandemic.
  • Teams used AI to predict the impact on COVID-19 of communities across the world.
  • The work reflects AI’s major role in using data to solve real-world problems.

In 2021, artificial intelligence has found its stride. By replacing anecdotes, assumptions and gut feelings with real-world data and learning-based models, AI is powering a global pandemic response that rises above policy and privacy differences to unlock the value of data sharing and cross-organisational collaboration.

The XPRIZE Pandemic Response Challenge is a working example.

This four-month-long global competition brought together 100 teams focused on developing AI systems with two objectives: to predict COVID-19 infection rates, and to prescribe intervention plans to help regional governments, communities and organisations reopen as safely as possible.

The challenge was entirely data driven. Each team based its creations on AI models developed by Cognizant and data curated and compiled daily by the Oxford COVID-19 Government Response Tracker, which collects information on policy responses and then scores the measures into a Stringency Index.

The grand prize winners were announced on 9 March. Judges awarded first place to a team from Valencia, Spain, whose winning model successfully forecast epidemiological evolution through the use of AI and data science. Second place went to a Slovenian team that developed accurate predictors of COVID-19 infections by combining machine learning with a susceptible-exposed-infectious-recovered (SEIR) epidemiological model. The two teams split the $500,000 purse.

Competing teams fed data continuously into COVID-19 models and trained them daily to understand how the pandemic – as well as containment strategies for testing, treatments and vaccines – would impact specific communities across the world. The result: powerful predictions for what should happen, not what will. And these have the potential to shape ongoing mitigation strategies.

The global experience was collaborative and profound. It generated outcomes that are verifiable and reinforce AI’s role in understanding the spread – and management – of infectious disease like COVID-19, and its economic and societal implications.

More importantly, the Pandemic Response Challenge reflects AI’s application in solving real-world problems, offering proof of how policymakers can employ data – rather than hunches – in decision-making.

The volumes of data that impact health, environmental and societal issues has grown, and making optimal decisions requires the power of AI and simulation.

—Bret Greenstein

The learnings may well help illuminate how businesses, NGOs and governments can arrive more quickly at predictive measures and resultant policies for addressing the challenges that lie ahead.

And this has come thanks to an army of thousands of volunteers collaborating across the world through data science.

The power of data and AI

Around the world, governments and businesses consider the battle over AI to be an investment in the future. In 2017, the Chinese government laid out its blueprint for AI superiority by 2030, and Russia’s Vladimir Putin declared that world leadership depends on AI domination.

Meanwhile, US and European leaders continue to advocate for investments and policies that encourage both AI leadership and the ethical use of the same.

It's little wonder really. Many world problems are too complex to solve without this powerful technology, from health issues like cancer and COVID-19, to environmental issues like global warming, to societal issues like food insecurity. The volumes of data that impact such issues has grown, and making optimal decisions requires the power of AI and simulation. These are the tools to help us deal with complex, important business and societal issues.

While success with AI is all about having the right data (and more of it), most of the world’s data remains locked away, like oil deposits trapped under layers of rock, in data warehouses, under desks, in labs and in our homes.

No data, no privacy protection

Not only is there no easy way to share it, there’s also little motivation to do so.

This data stalemate creates tension between AI development and the protection of individuals’ privacy in terms of data sharing.

In countries with strong privacy regulations, data is locked down more tightly, and there are fewer opportunities to share it among stakeholders.

In countries that have less privacy regulation, data is available at scale, leading to more powerful AI systems. Look no further than China’s adoption of digital surveillance tools for track-and-trace efforts during the pandemic. China’s tools assigned individuals a “green, yellow or red” risk rating based on their personal information and recent travel and health status. It remains to be seen whether the country will continue using the system when the pandemic retreats, or perhaps even expand the use of personal ratings outside the health sphere.

Why collaboration and incentives matter

The Pandemic Response challenge shows what can happen when people, data and AI come together. On this occasion, stakeholders’ motivations aligned, and the urgency was felt by all.

But such collaboration is not always possible because companies are protective of their data and fearful of sharing it due to regulatory risks. As a result of GDRP, HIPAA, CCPA and other laws, they hesitate to share data.

Incentives will play a powerful role in changing that scenario. At the moment, companies have few incentives to share data. For a look at the future, however, an analogy to consider is how companies were first regulated and then incentivised to improve their environmental impact.

Initial regulations such as emissions standards for cars and trucks were put in place to the sound of limit environment impact. But they were followed by incentives – ranging from carbon credits to tax credits for environmentally friendly vehicles – that encourage the adoption of more sustainable options. As it stands, brands have a reputational incentive to demonstrate sustainability – and auto companies have made enough breakthroughs to make environmental vehicles cheaper to own.

With data, we are in a similar position.

Initial regulations ensured data was handled securely and with respect for privacy. But we are competing in a global marketplace with countries that not only have few data regulations but also promote government investment in companies, and this effectively encourages data sharing.

China’s mobile phone infrastructure was critical in enabling the country’s track and trace of individuals for COVID-19. Governments in the US and Europe lack access to similar data for public safety.

Giving data credit

One alternative for Western nations to consider is applying a concept like carbon credits for sharing data securely. Let’s call them data credits. This model views companies that share data as contributing to the benefit of society, through either increased public safety or the creation of greater economic value, and it incentivises them to do so.

A data credits model has the potential to advance data sharing by leveraging new technologies such as Snowflake’s Data Exchange. Additionally, it lays the groundwork for the reputational advantage of being seen as an open data company.

The Pandemic Response Challenge has set out a path ahead through which data and AI can help us transcend differences and solve global problems. In 2021, it’s the direction we all need.