AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Meta’s Fundamental AI Research team hosted a two-day Open Source AI for Science Summit in Menlo Park, bringing together researchers from universities, hospitals and industry. Announcements and demonstrations included a larger brain-data target, chemistry tools and models, assistive robotics work, and an open mathematics library; several applications remain under development or planned.

Meta’s Fundamental AI Research team brought researchers from universities, hospitals and industry together for a two-day Open Source AI for Science Summit at the company’s Menlo Park headquarters last month, presenting work in brain research, assistive robotics, chemistry and mathematics. The company also announced an expansion of its Digital Brain Project’s data target and highlighted tools intended to let researchers and educators build on openly shared AI models, software and data.

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

The Digital Brain Project, launched with the Rothschild Hospital Foundation and the University of Montreal, funds researchers to collect and openly share de-identified brain data. Meta said it is raising the project’s target from 15,000 to 25,000 hours. The stated aim is to support research into human intelligence and neurological health. The announcement describes a target, not a report that the additional data has already been collected.

Researchers from the University of Pittsburgh presented an assistive-robotics system built using Meta’s DINOv3 and SAM 3 models. The team plans to demonstrate a robot that can help a person eat and is testing whether users can give it instructions in their own words. The summit report describes this as work in progress, not a system already available for routine use.

Meta also released UMA Nitro, a faster version of its Universal Model for Atoms. In work with Carnegie Mellon University and UC Riverside, UMA models identified multiple crystal forms of drug molecules without additional training data, Meta said. A separate collaboration with the University of Chicago used UMA to simulate enzymes atom by atom. These are research applications; the report does not provide quantitative performance results or establish clinical benefits.

For education, UMA Playground lets users build and visualize molecules and explore chemical reactions. Discovery Education said it is available to high school chemistry teachers with classroom guides. Arizona State University and Carnegie Mellon’s RealChem course plans to incorporate the tool into student activities. Meta says Discovery Education reaches more than four million educators and 45 million students; RealChem is used by nearly 100 institutions.

At a glance
reportWhen: Summit held last month; reported Novemb…
The developmentMeta’s FAIR team reported on projects and partnerships presented at its November 2026 Open Source AI for Science Summit.

Open Tools Reach Labs and Classrooms

The summit’s significance lies less in a single completed breakthrough than in the range of shared infrastructure and partnerships it brought together. Openly accessible data, models and verification tools can let researchers build on existing work rather than start from scratch, while collaborations with universities and hospitals connect AI development to specific scientific and accessibility needs.

The potential effects differ by project. A larger brain-data collection could give researchers more material for studying neurological questions, subject to the project’s data practices and research goals. Atom-scale simulations may help scientists narrow what to investigate experimentally, while classroom software can make molecular processes easier to visualize. These possibilities do not establish that the tools will produce medical treatments, new materials or improved learning outcomes; the summit report presents them as research and education efforts.

For mathematics, computer-checkable proofs offer another model of shared work. If a proof is formalized in Lean, software can verify its logical steps, potentially making it easier for other mathematicians to build on it. The approach depends on researchers choosing meaningful questions and accurately expressing their intended mathematics, rather than handing mathematical judgment over to AI.

Products Worth Considering

As an affiliate, we earn on qualifying purchases.

Projects Presented Across Four Fields

The event was organized by Meta’s Fundamental AI Research team and included hands-on workshops with researchers behind Meta’s models. According to Meta’s account, the two-day program covered healthcare, neuroscience, chemistry, materials science and mathematics. The company’s newsroom report is the source for the announcements and examples described here; it does not provide independent evaluations of the projects.

In mathematics, Meta highlighted Lean, an open-source language that allows computers to check formalized proofs. FAIR researchers also presented ATLAS, an open library that, according to Meta, formalizes 26 open-access textbooks and contains more than 45,000 computer-checked definitions and results. The company said mathematicians guide the work by selecting research questions and ensuring that formal statements reflect their intended meaning.

Across the featured areas, the stated approach is to share tools and collaborate with outside institutions. The summit report describes project goals, demonstrations and planned integrations, but offers limited detail about independent validation, access conditions for each resource or timelines for wider deployment.

“The project funds researchers to collect de-identified brain data and share it openly.”

— Meta’s Fundamental AI Research team, in its summit report

Evidence and Deployment Still Developing

The summit account does not provide independent assessments of the featured models or report detailed benchmarks for UMA Nitro’s speed, accuracy or performance against alternatives. It also does not state when the expanded brain-data target will be reached, how much qualifying data is currently available, or which researchers can access it under what conditions.

The Pittsburgh robot’s demonstration remains planned, and the report gives no deployment schedule or evaluation of how well it may work for users with disabilities. RealChem’s integration is also described as planned, without a start date. More broadly, the report does not establish that simulated molecules or enzymes will lead to medicines, or that classroom use will improve student outcomes. Those results remain to be tested and documented.

Demonstrations and Partnerships Ahead

The next stated milestones are the University of Pittsburgh team’s planned demonstration of its eating-assistance robot and RealChem’s planned integration of UMA Playground into student activities. Meta’s expanded Digital Brain Project target will depend on researchers collecting and sharing the additional de-identified data. The summit report does not give dates for these milestones.

Researchers and educators will be able to judge the projects through further demonstrations, published methods and results, and information about access to the shared resources. Meta’s stated direction is to continue working with outside scientific communities and make models, code and research available for others to build on; the scope and schedule of future releases were not specified in the report.

Key Questions

What was the Open Source AI for Science Summit?

It was a two-day meeting hosted by Meta’s Fundamental AI Research team at the company’s Menlo Park headquarters. Researchers from universities, hospitals and industry discussed open-source AI applications across several scientific fields.

What changed in the Digital Brain Project?

Meta said the project’s target for collecting de-identified brain data is increasing from 15,000 to 25,000 hours. The announcement concerns the target; it does not say the additional data has already been collected.

Is the assistive robot ready for use?

The summit report describes a University of Pittsburgh research system built on DINOv3 and SAM 3. The team plans to demonstrate a robot that can help someone eat; the report does not say it is ready for general use.

Where can chemistry teachers use UMA Playground?

Discovery Education said the tool is available to high school chemistry teachers with classroom guides. RealChem, a college chemistry course associated with Arizona State and Carnegie Mellon, plans to include it in student activities.

What is ATLAS?

ATLAS is an open library for formalized mathematics. Meta says it includes 26 open-access textbooks and more than 45,000 computer-checked definitions and results.

Source: about.fb.com

You May Also Like

The Future Of AI Computing: 8 Graphics Cards To Consider In 2026

Discover the eight leading graphics cards to consider for AI computing in 2026, highlighting performance, features, and future-proofing for professional workloads.

Verizon Communications Surges In Global Coverage

Verizon Communications reports a major increase in its global coverage, with 31 mentions in recent data, signaling a strategic expansion.

10 Best Wireless Earbuds For Your Everyday Routine In 2027

A product roundup compares 10 wireless earbud listings by noise cancellation, workout fit, form factor and the details buyers should verify.

Android 17 QPR2 Is Here, And It’s More Than Just A Pixel Makeover

Google has begun rolling out Android 17 QPR2 to supported Pixels with new customization, Auracast, privacy and multitasking features.