Meta Platforms, the company formerly known as Facebook, has disbanded its protein-folding team, a group of researchers who were working on using artificial intelligence to solve one of the most challenging problems in biology.
The move comes as Meta shifts its focus towards developing more commercial AI products and services, such as chatbots, virtual assistants, and augmented reality glasses, rather than pursuing blue-sky research that may not have immediate applications or revenue streams.
The protein-folding team was part of Meta’s AI Research (FAIR) division, which was founded in 2013 by Yann LeCun, a pioneer of deep learning and a Turing Award winner. FAIR’s mission was to advance the state of the art in AI and to foster collaboration with the academic community.
Protein-folding is a key problem for drug discovery and biotechnology
Proteins are the building blocks of life, responsible for carrying out various functions in living organisms. However, their function depends on their shape, which is determined by how they fold from a long chain of amino acids into complex three-dimensional structures.
Predicting how a protein will fold from its amino acid sequence is a notoriously difficult problem, as there are an astronomical number of possible configurations. Solving this problem could have huge implications for drug discovery, biotechnology, and understanding diseases such as Alzheimer’s and Parkinson’s.
In recent years, AI has emerged as a powerful tool for protein-folding, with several breakthroughs achieved by teams such as DeepMind, Google’s AI subsidiary, and OpenAI, a research organisation backed by Elon Musk and others. These teams have developed neural networks that can predict protein structures with remarkable accuracy, surpassing the performance of traditional methods.
Meta’s team made significant contributions to the field but faced internal challenges
Meta’s protein-folding team was led by Alexander Rush, a former professor at Harvard University who joined the company in 2019. The team consisted of about a dozen researchers, mostly based in New York and Paris.
The team made significant contributions to the field of protein-folding, publishing several papers in prestigious journals and conferences. They also participated in CASP14, a biennial competition that evaluates the performance of different methods on unseen protein sequences. The team ranked among the top performers in several categories, behind DeepMind and OpenAI.
However, the team faced several challenges within Meta, according to sources familiar with the matter. They said that the team struggled to get enough computing resources and support from the company’s leadership, who were more interested in AI projects that could generate revenue or enhance Meta’s products.
The sources also said that the team felt isolated from the rest of FAIR, which was more focused on natural language processing and computer vision. They said that the team had little interaction or collaboration with other researchers in FAIR or in Meta’s Applied Machine Learning (AML) division, which develops AI solutions for Meta’s platforms such as Instagram and WhatsApp.
Meta confirms the decision but says it remains committed to fundamental AI research
Meta confirmed that it had decided to disband its protein-folding team, but did not provide any specific reasons for the decision. A spokesperson for Meta said:
“We are incredibly proud of the work done by our protein-folding team and their contributions to advancing fundamental AI research. As we continue to evolve our research priorities and areas of focus, we have made the difficult decision to wind down this project. We remain committed to supporting fundamental AI research across a range of domains and applications.”
The spokesperson also said that Meta would continue to support the researchers who were part of the protein-folding team and help them find new opportunities within or outside the company.
The decision to disband the protein-folding team comes amid a broader reorganisation of Meta’s AI divisions, following the announcement of its new corporate identity and vision in October 2021. The company said that it would invest $10 billion annually in building the metaverse, a virtual environment where people can interact through avatars and digital objects.
As part of this vision, Meta said that it would focus on developing AI technologies that could enable more natural and immersive experiences in the metaverse, such as conversational agents, computer-generated graphics, and spatial audio. The company also said that it would leverage its AI capabilities to address social issues such as misinformation, hate speech, and privacy.
