How scientists are leveraging artificial intelligence to solve the pandemic
Thanks to sci-fi, the world is convinced robots are coming to get us. Of course, it’s unlikely they ever will. For now, AI is on our side, helping us in various corners of life and work, not least around getting things back on track after this difficult year.
Since coronavirus first appeared in China, researchers have looked to artificial intelligence for many different answers. In fact, machine learning (ML) systems and computational tools are why we were able to start testing potential vaccines as early as March 2020 — far sooner than humans could ever hope to create one without the help of such technologies.
Why? Because these technologies allow scientists to draw quick insights from vast data sets that previously would have taken huge teams years to analyse. As is the case in so many applications of AI, the best results come from human-machine collaboration — I said as much in a recent Instagram post on using AI to combat fake news.
In terms of Covid-19 vaccine development, ML and computation play their part, but there’s no substitute for human lab work. Many of these collaborations have produced vaccines showing favourable results in phase III clinical trials with several now approved for use across the world.
One of these is the RNA vaccine designed by pharmaceuticals giant Pfizer alongside German biotech firm BioNTech. The BioNTech story is a fascinating one. Married Turkish-German scientists Özlem Türeci and Ugur Sahin set up in Mainz, Germany in 2008, with the aim of developing new types of immunotherapy against cancer. A mere twelve years later, BioNTech is worth over £16 billion and is leading us to the end of the coronavirus era.
After hearing about the Wuhan outbreak, the couple set about their Covid vaccine research project (Project Lightspeed) in January 2020. Sahin, who felt the development of a vaccine was a “duty,” said:
“There are not too many companies on the planet which have the capacity and the competence to do it so fast as we can do it.”
Although we don’t know exactly how BioNTech developed its vaccine, we do know that in 2019 Sahin said they mixed:
“biology with bioinformatics, robotics and artificial intelligence.”
Binbin Chen, at the Stanford Institute for Human-Centered Artificial Intelligence (HAI) believes every vaccine was developed using computational models. ML tools can predict — based on training data sets from known pathogens — which pieces of the virus the immune system is most likely to recognise, giving each target a numerical score. This speeds up the process because immunologists can focus on fewer potential “targets” with the highest scores and bring those to the lab.
BioNtech’s recent partnership with London-based InstaDeep suggests the need for even greater expertise in AI. Both firms say:
“The AI innovation lab will combine InstaDeep’s advanced capabilities in the areas of artificial intelligence, machine learning, and digitalization along with BioNTech’s deep domain expertise in precision immunotherapies and its access to a wide variety of internal and external datasets.”
Some also think AI should be used more widely — for example, to decide who should be vaccinated first. This follows the logic that moving broad at-risk groups like the elderly to the front of the queue is “short-sighted” because it ignores the healthcare, social, and environmental complexities that place people at greater risk of catching or dying from Covid-19. The vaccine rollout is somewhat of an ethical dilemma, and I suppose AI and data-driven decision-making can help make the process appear fairer.
If it helps get us back to some sense of normalcy, I’m all for human-machine collaboration. A final word of warning though: some coronavirus AI systems were applied without first being tested or were trained on small or low-quality data sets. This was possibly justified by the horrors of the pandemic and a life-or-death timeframe, but future regulation of such systems must be tighter or they could do more harm than good!