Target He took an important step in the career of artificial intelligence. Technological giant He released a family of models that can revolutionize scientific developmentThe field to achieve this, Meta used one of the most extensive data sets, when the liba created in the field of computing chemistry.
In the publication of its website, the AI Research Meta announced the launch of the universal models for atoms (UMA). These language models They studied with data compilation, which includes 500 million atomic structures in 3DThe META field also has common molecules of 2025, a set of data for promoting molecular opening.
“Today we are launching several models, reference points and innovative data sets that transform the method of turning to forecasting molecular properties, linguistic processing and neurobiology“Meter mentioned.”These achievements are the result of purposeful scientific and academic progress and are a significant step towards advanced artificial intelligence“
Meta showed that open molecules are needed 2025 more than 6,000 million hours of calculations for 100 million quantum calculations Interactions between atoms and molecules. A new data set can be used to predict the properties of molecules and materials in scenarios, such as the formation of chemical ties. Meta claims that This database provides unprecedented accuracy in the design of the atomic scaleField
In addition to the data set, Meta has released UMA, a model trained in more than 30,000 million atoms, which allows us to understand molecular behavior. Both will be the key to accelerating the design of materials on an atomic scale.
New target models will help to detect new materials and medicines
According to the fundamental research group of AI (Fair), models use a unique architecture that allows Increase the container without speedAfter the assessment, the researchers found the field that UMAS can function in a similar or better way than other specialized ones.
“We are pleased to see as the USA OMOL25 and UMA community to help advance the research of molecular science and materials, which, ultimately, can lead to new achievements in all industries“
The launch of the family of models and data recruitment can help solve some of the most difficult problems facing humanityThe field among examples is the development of materials to reduce the influence of climate change or new molecules that accelerate industrial progress. UMA models are capable predict movement and behavior on an atomic scalewhich leads to a reduction in detection and development processes.
Open Molecules 2025 data set and UMA models family Available through a hugThe repository field is publicly available, although data is required, such as a full name, date of birth or organization, in order to gain access to them for free.
Source: Hiper Textual
