Source: Cailian Press

Editor: Zhou Ziyi

Image Source: Generated by Wujie AI

According to sources cited by the media on Monday (August 7th), social media giant Meta has laid off a research team that had previously developed an artificial intelligence (AI) method capable of predicting protein structures. This indicates that the company is abandoning pure scientific projects in favor of developing AI products that are more commercially viable.

Previously, Meta had employed approximately 12 scientists for a project called ESMFold, which trained a large language model to process a large amount of biological data to predict protein structures and used AI to create the first database containing over 600 million protein structures. This progress had once received praise from medical professionals involved in developing new drugs and treatment methods.

According to three individuals familiar with Meta's restructuring plan, the ESMFold team was disbanded this spring as part of the company's large-scale layoffs. However, there had been no previous reports about this.

The sources added that Meta is still hiring thousands of AI scientists and engineers, compared to the smaller size of the ESMFold team. Nevertheless, the decision to cancel the project indicates that Meta hopes to abandon "blue sky research" (engaging in basic scientific research without considering the possibility of practical application in the short term) and instead support AI projects that can generate revenue.

Yaniv Shmueli, a former research scientist and engineering manager at Meta AI who was involved in ESMFold, stated, "Meta is trying to adjust its research strategy to understand more about how to create advanced intelligence, making it a business for Meta rather than just some novelty projects."

ESMFold Team

Meta established the Fair (Facebook AI Research) lab in 2013, hiring top scholars in the field and dedicating itself to research in this area.

In November of last year, researchers at Fair published a paper in the journal "Science," detailing the research achievements of ESMFold: a database containing 6.17 billion macrogenomic protein structures created by machine learning, known as the ESM macrogenomic map. Macrogenomics is the study of little-known proteins from environmental samples around the world, including microbes in soil, oceans, and the human body.

The ESMFold project first trained a large language model to learn evolutionary patterns and directly generate accurate structure predictions from the DNA sequences of proteins.

Meta also created an open-source database, allowing scientists to easily retrieve specific protein structures relevant to their work, and expressed hope that this work would "drive further scientific progress."

Meta's project was considered a competitor to DeepMind's protein folding prediction technology, AlphaFold, which was once considered a scientific breakthrough in 2020 for its accuracy comparable to laboratory methods. However, ESMFold's language model described structures 60 times faster than AlphaFold, despite being less accurate.

Tim Hubbard, a professor of bioinformatics at King's College London, suggested that large tech companies may have an advantage in rapidly and massively deploying computing resources and providing costly services to scientists.

However, the long-term challenge lies in maintaining the substantial costs of running algorithm services and databases. Meta has not confirmed whether it will continue to provide this service in the future, but the data is currently available for use by the research community. Hubbard expects the academic community to find ways to continue this work.

Full-scale Advance in the AI Field

Meta is one of the earliest large tech groups to invest in AI. Since the establishment of the Fair lab, it has published many papers and gained recognition from the scientific community for its advancements in AI.

However, at present, the company has begun to lag behind competitors such as OpenAI, Microsoft, and Google, all of which have consumer-facing generative AI (AIGC) chatbots.

2023 was referred to as the "year of efficiency" by CEO Mark Zuckerberg, and Meta has undergone a large-scale restructuring in recent months, adjusting its management structure and laying off approximately 20,000 employees.

The new focus of Meta will be to leverage its long-term research and development in the AI field to create products centered around AIGC, a technology that can generate text paragraphs with a human-like tone, as well as images and videos.

According to two sources, in February of this year, Meta established an AIGC team led by product executive Chris Cox, which currently has hundreds of employees, including those transferred from the Fair lab. It is reported that Meta is currently attempting to reconfigure Fair's research to align with the goals of the GenAI team.

Last week, reports indicated that Meta plans to launch a series of chatbots with different personality styles as early as September this year in an effort to surpass its competitors.

Joelle Pineau, Vice President of AI Research at Meta, stated, "Meta remains committed to exploratory research based on open science, and the transfer of other projects from the Fair lab to our business has always been part of the team's operation."

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。