One of the core building blocks of the CDEI’s COVID-19 response is our repository - a database for novel use-cases of artificial intelligence and data specifically being used to counter and mitigate the effects of COVID-19 around the world. The first public release of the repository can be found here.
We will continue to update the repository on a regular basis, publishing an updated snapshot each month with any new use-cases that have been brought to our attention. To that end, we strongly encourage anyone who has visibility of developments we are yet to identify to contribute to the database. Further use-cases can be sent to email@example.com, which will be monitored by members of the CDEI team.
Key findings from the April repository
- The variety of use-cases is wide-ranging, covering both the deployment of algorithms (e.g. automated content moderation systems) and the collection and sharing of data (e.g. through open banking models, allowing users to disclose their income to HMRC in order to receive new welfare entitlements - N.B. this is still in beta mode).
- Although there are a number of entirely new uses of data and AI, approximately half of our April database relates to extension or pivots of existing activity to a new context, or a new synthesis of existing data sources. We would expect to see an increase in novel use-cases as time goes on, after further development and testing time.
- The majority of our existing use-cases (two-thirds) arise in the healthcare and digital sectors. Again, this is to be expected given the priority of managing the immediate public health crisis (70% of the uses in the repository are directly related to managing this), and also the advantage held by Big Tech companies in their ability to pivot existing activities.
- However, we have identified noteworthy applications across a number of different sectors, and expect to see more as the nature of the crisis turns to focus on re-opening the economy, and looking to future resilience.
- An underlying development across all of these use-cases is that data is being shared between organisations, both public and private, at unprecedented speed and scale.
- Alongside government-led attempts to use technology to address the pandemic, consumers and businesses are deploying digital and data-driven applications of their own accord to cope with lockdown measures. We are seeing an increased use of video conferencing, digital entertainment, social media, automated recruitment, online marketplaces and delivery networks, and automated content moderation.
- The lifespan of these applications is uncertain. Many of the challenges posed by COVID-19 have already begun to subside (e.g. product shortages in supermarkets), which may reduce the appetite for technology-driven interventions (i.e. algorithmic systems designed to distribute products where they are most in need).
- Having said this, we may see that now the applications have been implemented there is continued appetite to use them to improve resilience moving forwards. It is also plausible that we see repeated waves of infections in coming months, which means that at the very least the adoption and relevance of these technologies will rise and fall in waves.
- In many of the cases, we have little choice but to rely on AI and data-driven technology to manage the effects of the pandemic, such as automatic moderation of content for social media platforms where human moderators cannot enter the workplace. However, there are some use-cases with credible alternatives to the use of AI and data-driven technology. For example, using wearables to maintain distance in the workplace may not be necessary if companies lean on traditional methods such as floor markings, or changes in workplace layouts.
- Levels of technology adoption vary from country to country, with some nations being more open to new innovations than others.
- Finally, the impact of AI and data-driven technology in mitigating the effects of the pandemic may yet be overstated. Many of the most novel use-cases are either still being scoped or in the early stages of development, and so we cannot say how effective they will be moving forwards. Some rely on improved science elsewhere (i.e. digital health certificates will rely on improved testing), and it’s also often the case that successes demonstrated in the lab do not always translate into the field (this is of particular note for diagnostic tools).