While there are legitimate concerns about the potential risks of artificial intelligence, we have to keep in mind that this multifaceted technology has for years been a powerful tool in many areas of science and research — including the not-at-all artificial fields of biology and the life sciences. An organization that knows this as well as any is the Howard Hughes Medical Institute, one of the world’s largest funders of basic biomedical research, and an institution with a substantial history of harnessing AI technology for research.
That’s why HHMI’s recent announcement of a $500 million, 10-year commitment to drive AI research in the life sciences was not a surprise, despite the considerable number of dollars involved.
Founded in 1953 by aviation industrialist Howard Hughes, HHMI is one of the world’s largest funders of basic biomedical science, having spent approximately $4.4 billion in the last five years on research and science education. And with $24.2 billion in assets at the end of fiscal year 2023, it’s one of the country’s wealthiest foundations overall. Now, after some 15 years of experience working on AI and machine-learning projects to support biological research, HHMI is doubling down on its commitment to AI in biomedical sciences.
Dubbed AI@HHMI, this new, 10-year program will support AI-driven science at HHMI’s Janelia Research Campus in Ashburn, Virginia, and at more than 300 HHMI-affiliated labs around the country. Janelia, a research campus HHMI created in 2006, is a fascinating part of the HHMI story. It was established to find new ways to advance science, created in the tradition of research and development hotbeds like Bell Labs and the Laboratory of Molecular Biology in Cambridge, England. Currently, more than 650 scientists, research technicians and other employees work at Janelia.
In an effort to encourage both creativity and collaboration, Janelia consists of many small labs with just two or three members, mostly run by recently minted Ph.D.s and postdocs. The labs are small by design to encourage collaboration, fluid team building and sharing of resources, such as computing and software, or chemistry and biology.
Beyond Janelia, however, HHMI also funds scientists who work at universities and research institutions around the country, and the AI commitment is designed to create novel collaborations between teams at Janelia and scientists in the HHMI-affiliated labs nationwide.
“We want to address problems for which existing funding mechanisms don’t really exist, and make sure that we can use Janelia at its best, namely at this proximity between experimental design and execution and working on computational problems,” Stephan Saalfeld, Janelia senior group leader and head of computation and theory, told me recently. A key strength of Janelia is the collaboration it enables between theorists, experimentalists, computational scientists and engineers, allowing them to develop theoretical and computational models and collect high-quality experimental data.
While not all science funders enjoy HHMI’s massive means, those that can set up their own in-house research hubs (the Simons Foundation’s Flatiron Institute also comes to mind) may be putting themselves in a position to leverage their own institutional knowledge to deliver on one of science philanthropy’s common selling points — that it can contribute to the research landscape in ways that public or corporate funding do not.
The AI@HHMI program will help scientists and engineers place AI and machine-learning tools at the center of Janelia’s research projects, building and deploying AI systems to accelerate every aspect of the research process — aiming, says HHMI, to explore “the full extent of what is possible when AI is integrated into the research process.” AI systems will be used to help design experiments and create learning models capable of inferring underlying principles and gaining useful insights about the world hidden within the data.
As the largest private biomedical research organization in the U.S., the directions HHMI takes have a big impact throughout the biological and life sciences fields, in academia as well as industry. HHMI’s Janelia has previously partnered with Google to apply AI to biology research, working with Google’s Connectomics group, which focuses on neuroscience, and Google’s DeepMind, which develops AI technology across a range of scientific and engineering fields.
The new AI@HHMI program will not only cover AI computational tools, but also microscopy, sensor design, chemistry and essentially any other tool to be used in the research process. Over the last decade and a half, labs at Janelia have developed numerous AI solutions to enable research not otherwise practically possible, Saalfeld said. AI has helped scientists to analyze large data sets, for example, that would require decades to complete without AI technology. Researchers have also developed AI microscopy that can extract data beyond the range of human eyesight, identify individual neurons and the neurotransmitters inside them, and that contributed to the creation of the first detailed map of an adult fly brain.
In the first phase of its $500 million investment, through October of this year, HHMI is inviting proposals for AI-based research projects led by its HHMI Investigators, Freeman Hrabowski Scholars and Janelia Group Leaders. (Under its Investigators program, one of the organization’s signature initiatives, HHMI employs scientists at institutions around the country, providing them with about $11 million in support over seven years with an aim to enable creative research. HHMI launched the Freeman Hrabowski Scholars Program in 2022 to support the careers of diverse researchers.) Accepted projects will be fully funded by HHMI and executed at Janelia in collaboration with HHMI labs as well as a new team of AI scientists, AI engineers, robotics engineers and data scientists.
“We want to focus on research where we need the AI people and the experimental people and the engineers in one place to build something — and couldn’t be done easily elsewhere,” Saalfeld said. That may be common in an industry context around product development, but here, Saalfield said, “our product is to further human knowledge and to understand fundamental principles of biology. And we also want to make these results and intermediate steps and the pipelines extremely accessible.” HHMI also maintains a commitment to open science and will publish its AI tools as open source.
Much AI coverage in the last couple of years — including here at IP — has focused on fears about AI, and on doubts about its capacity to broadly benefit society and humanity. These are issues that must be addressed, particularly in commercial applications of AI that could lead to bias or discrimination. But to scientists, AI technology holds incalculable promise. “What excites me the most is that we can use these AI models to uncover a structure in data that is not immediately accessible,” Saalfeld said. “It’s just a massively better statistical analysis tool than what we’ve historically had.”
