Stay updated with free Meta Platforms news by subscribing to our newsletter.
Receive a daily email digest, known as myFT Daily Digest, containing the latest Meta Platforms news every morning.
Meta has discontinued a project called ESMFold, which utilized artificial intelligence to create a comprehensive database of over 600 million protein structures. This decision signals the company’s shift away from purely scientific endeavors in favor of developing profitable AI products.
The project involved a group of approximately twelve scientists who trained a large language model to process vast amounts of biological data and predict protein structures. The initiative received praise from professionals working in drug development and treatments.
Earlier this year, the ESMFold team was disbanded as part of company-wide layoffs. While the protein-folding team was relatively small compared to Meta’s extensive AI workforce, its elimination reflects the company’s focus on revenue-generating AI projects, according to insiders.
Yaniv Shmueli, a former research scientist and engineering manager at Meta AI who worked on ESMFold, stated, “Meta has shifted its research strategy to focus on creating advanced intelligence that can benefit the business, rather than pursuing purely curiosity-driven projects.”
Meta has recently undergone significant restructuring, including job cuts affecting around 20,000 employees, as part of its “year of efficiency” initiative. The company, an early investor in AI, established its Fundamental AI Research (Fair) lab in 2013 and hired top academics in the field.
While Meta has been recognized for AI advancements and research publications, it has fallen behind competitors such as OpenAI, Microsoft, and Google in the development of consumer-facing chatbots using generative AI.
Meta’s new focus is to leverage its existing research and development to create products capitalizing on the hype surrounding generative AI, which encompasses text, image, and video generation. The company’s product chief, Chris Cox, leads the generative AI team, which comprises several hundred staff members.
Last week, it was reported that Meta plans to launch a line of chatbots in September to catch up with competitors.
Joelle Pineau, vice-president of AI research at Meta, affirmed the company’s commitment to open science through the Fair lab. She stated, “Projects graduating from Fair and moving into other areas of our business has always been a component of how the team operates.”
Some insiders attribute Meta’s delayed entry into the generative AI market to the academic culture within the Fair lab, which hindered collaboration both within the team and with other departments. The company is now working to align Fair research with the goals of the GenerAI team.
In November of last year, Meta researchers released the ESM Metagenomic Atlas, the first database containing over 600 million metagenomic protein structures. This project aimed to advance metagenomics, the study of proteins found in various environments. Meta’s ESMFold project, which trained a large language model, delivered faster but slightly less accurate protein structure predictions compared to DeepMind’s AlphaFold.
Meta’s open-source database, providing easy access to specific protein structures, was intended to drive scientific progress. However, concerns have been raised about the long-term maintenance costs of the database and the ESM algorithm service for predicting protein structures.
Tim Hubbard, a bioinformatics professor at King’s College London, acknowledged that large tech companies have an advantage in providing computational resources for scientists but anticipated that academics would find alternative solutions.
Meta has not confirmed the future status of these services but reassured that the data currently remains available for the research community.
Denial of responsibility! VigourTimes is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
