Today we announced the AWS Impact Computing Project at the Harvard Data Science Initiative (HDSI) to identify potential solutions that can improve the lives of humans, other species, and natural ecosystems.
Technology and innovation are transforming the world at an unimaginable pace – changing society and economies, curing disease, and fundamentally re-shaping the way we live.
And yet, much remains to be done to combat global inequities, address climate change, ensure food security, and better anticipate global health crises. Thanks to advances in science and technology, we now deeply understand the velocity with which infectious diseases and pandemics can spread – supercharged because of climate change and globalization. For example, we know that a pathogen can travel from a remote village to major cities on all continents in 36 hours.
These are grand challenges. They are complex, highly interdependent, and dynamic. The solutions to these challenges must integrate many things from science, engineering, and technology, with policy, culture, and geopolitics.
What is this collaboration about?
I have spent my entire career thinking about the role HPC, and now AI, can play in solving the most vexing problems facing us. For the first time, because of the scale and capability of AWS, I can see a path towards meaningful progress. Scale matters – but understanding the depth and complexity of the use-cases matters too. Tapping into the brightest minds matters – but, asking the right questions, perhaps matters the most.
That’s why I’m thrilled about our collaboration with HDSI. The basic premise of the initiative posits that fundamental gaps in understanding the problem space, coupled with lack of accessible computational power and algorithms, have stifled progress.
It’s easy to think the solution is merely more compute. Unfortunately, it’s not. Important challenges in science, analytic methodologies, data, and accessibility must be addressed along the way.
Together, AWS and HDSI will engage in deep, cross-disciplinary data-science research to strengthen and expand our understanding of the problem space. We’ll leverage those insights to optimize and enhance our HPC and AI service portfolio to better support these unmet needs.
This is a classic example of what Amazonians do daily – working backwards from the needs of our customers. In this case, the requirements are coming from organizations working on large social and global challenges. Each of these areas, from climate change, sustainability, and food security, to drug discovery for orphan diseases and ensuring equitable healthcare, requires access to large shared data sets and easy access to significant computational power. Getting these resources into the hands of decision makers is vital. Our goal is to make data and analysis accessible to anyone, from field workers to policymakers, enabling them to deepen their understanding of the issues and make more informed decisions.
This collaboration will open a whole new area of impact-specific solutions and build the capacity for sustainable change.
How does the collaboration work?
One of the early projects we are exploring is an effort to predict — with reasonable accuracy — the maize yield in Africa in the context of climate change, driven by extreme heat waves. Fundamental gaps exist around this problem ranging from lack of high-resolution geospatial data to estimate land use, and algorithms and methodologies to combine land use and historical weather and climate data to predict crop yield.
To help with this, we plan to bring together the research methodologies from HDSI, the massive historical climate data from the UK Met Office, and geo-spatial data on land use from the Amazon Sustainability Data Initiative (ASDI) to refine and optimize the crop yield predictions.
These models could potentially be used by organizations like the World Food Program or the African Development Bank to plan effective response. A key focus will be to ensure the tools and methods we develop are easily accessible and usable by the people who need this local data to make decisions daily. From a technology infrastructure perspective, this project will involve developing data platforms to integrate large and heterogeneous datasets for climate, pollution, and environmental observations with high performance computing infrastructure so we can model and simulate future state scenarios based on the research algorithms.
While this example provides a glimpse into possible outcomes and associated impact, there’s also an equally important technology innovation aspect to this. Technology is a key part of most research projects, especially the ones based on analyzing and making sense of huge data sets, or others that require complex multi-variate analyses to infer the impact of hundreds of variables on a single event. The dependence of technology is so prevalent in these research projects that researchers may limit their analysis, simulation and modeling and investigations based on the availability of technology, either in terms of capacity or capability. In either case, the net effect turns technology infrastructure into a blocker to furthering scientific progress as opposed to a catalyst.
Necessity is the mother of invention
Every time a researcher is blocked….