How deep learning can help us capture infant biometrics for the millions of children globally whose births are not registered
Around the world, millions of children are not registered at birth, creating a barrier between them, their education and access to healthcare. Since 2017, biometric digital identity company Element has been managing the world’s largest infant biometric program to fight this problem. Co-founded by Yann LeCun, Element was the first modern AI company focused on digital identity, and laid the foundation for deploying deep learning on mobile devices. Collaborating with world-class healthcare institutions across Bangladesh, Cambodia, and Mozambique, Element has reached 10,000 infants and children so far, but their ultimate goal is to develop a biometric solution to link children to vaccinations, birth registration, and other essential health and social services on an even bigger scale. Below, they explain how deep learning can help.
Where is your team based, how big is the team, and who is involved?
Element is based in NY with regional teams around the globe. We have a diverse team with over 30 members including AI researchers and engineers developing a mobile digital identity platform. Among them, a handful of members focusing on infant biometrics, including Manoj Alwani, Wei Hong, Yang Wang, Adam Perold, and Yann LeCun, are involved in the AI XPRIZE project.
What's the global issue you're tackling?
Globally, an estimated 650 million children were not registered at birth – creating foundational barriers to accessing education, healthcare, and other essential services. For example, just 7% of children in the world’s poorest countries are fully vaccinated, in part because it is difficult to identify them and track their vaccinations over time. Without digital health records and an appropriate digital ID accurately and uniquely linking a child to that record, this will remain a persistent problem.
There are many innovative technologies addressing these challenges, from digital vaccination records to mobile birth registration platforms. Yet there are no commercially-available biometric solutions for infants and young children. Additionally, in such a context, developing robust privacy-led frameworks to ensure the protection of the data of participants is more important than ever, and this too must be solved.
Element aims to develop privacy-led, adaptive biometric identity solutions for the purpose of linking people to essential services. Our focus on growth markets led us to countries across Asia and Africa, driven by the rapid adoption of smartphones and expansion of digital ecosystems.
Walk us through your AI technology?
Element develops a mobile software-only solution to leverage the existing camera on mobile devices to image a biometric modality, using advanced AI methods for recognition – no specialized hardware or connectivity required. Element’s technology is based on the field of deep learning, in which models train themselves directly from data, rather than the conventional approach of a human expert imposing rules of classification. The technique was pioneered by Element co-founder, Yann LeCun, the research scientist credited with developing the method of Convolution Neural Networks (“CNN”), which helps achieve higher performance in natural signal processing.
Deep learning is uniquely suited to address the challenges of capturing infant biometrics, as the models learn directly from the natural features of infants, can be deployed to understand changes over time, and can be applied to multiple modalities at once. This creates a more robust method of identity recognition that cannot be achieved with typical approaches and creates a native security architecture that provides unique data privacy advantages.
What made you want to enter the $5M IBM Watson AI XPRIZE?
The application of AI to the global grand challenge of infant identity is novel with significant scope for services such as vaccinations and birth registration, and is the kind of work XPRIZE sought to uncover with this prize. Meanwhile the partnerships we created in pursuit of this are just as novel and important and, we think, demonstrate the promise of collaboration across technology companies, social impact enterprises, and governments to bring cutting-edge innovation to global development challenges.
What's been your team's biggest challenge so far?
Not having public datasets of the biometric modalities of infants and young children at the onset of our work for this type of large scale study was a significant challenge. As such, all datasets gathered for the project are primary data sources, collected directly by implementing partners at our project sites. This process yielded valuable insights into the end-user experience and field-level operations of our partners, which have helped us improve the technology performance and design.
Why is what your team is doing important now, and how do you see it scaling up in the future?
We believe the ability to prove who you are is a fundamental and universal human right. We envision a world where children can be linked to essential services in a manner that is inclusive, secure, and protective of their privacy. With our mobile software-only solution it is possible to achieve at scale. This could mean the platform is being used to create digital immunization records, ensuring children receive the full course of life-saving vaccines. It could mean supporting proof of birth registration, the legal framework which underpins the delivery of so many other essential services, like education. These are the digital systems of the future, but they are being built today.
How has the competition furthered your success? How has it changed you?
When the $5M IBM Watson AI XPRIZE launched in June 2016, Element was still very much in stealth R&D mode. We had just started conversations with Global Good about the potential adaptation of our technology for infants and children, but our partnership wasn’t formalized until December of that year.
From there, we were focused on launching the early-stage research in Bangladesh and Cambodia. When we discovered the XPRIZE opportunity, we felt encouraged to put forward our work in the Wildcard Round, despite a large set of competitors. That in turn launched us directly into the busiest months of the competition. While challenging, this became a valuable catalyst for our team to make further investments in the documentation of our work and to push the boundaries of our technical applications and vision for impact. Within our team, we found ourselves not referring to XPRIZE as a “prize” or “competition,” but instead as a “program.”
Outside of your work what's an area of AI that's exciting you right now?
Element has a mix of enthusiastic AI researchers and developers who are interested in pioneering various advances in deep learning, such as model optimization and compression, and leveraging them for the delivery of efficient software-only mobile solutions. There are many exciting frontiers of AI today. Self-supervised learning holds great promise for building models that can capture more powerful representations of the world, yielding higher performance results for tasks across disciplines. Attention mechanisms have also provided breakthroughs in NLP, computer vision and many other applications. It’s an incredible time in the field, with many profound topics for deep research.