High-quality translation tools don’t exist today for hundreds of low-resource languages, preventing people from accessing information and creating content online in their preferred or native languages. Meta AI researchers are working to change this through the No Language Left Behind (NLLB) project. The open source models created through the project can deliver highly-accurate translations directly between 200 languages.
With Wikipedia serving as one of the most widely-used sources of information today, it’s important the biographical information it provides is accurate and without gender bias or inequity. Meta AI researchers are making Wikipedia more representational with an open-source AI model that automatically creates biographical articles about important public figures, along with a novel dataset to evaluate model performance on real biographies of women from historically marginalized groups.
Through a joint research initiative with NYU Langone Health and Meta AI, researchers developed a model that could accelerate MRI scans without degrading the quality of the images. And with a faster scanning process via this new AI model, we hope to improve the patient experience, help expand access to the technology, open up new use cases and help advance AI research in other domains.
With the effects of climate change in mind, Meta AI and Carnegie Mellon University’s Department of Chemical Engineering have teamed up to create the Open Catalyst Project. Its researchers use AI to accelerate quantum mechanical simulations by 1,000x to find more efficient and scalable solutions for storing and using renewable energy.
Through our ongoing wav2vec research project, we've developed a way to build speech recognition systems that require little to no transcribed data. This need for less labeled data enables greater speech recognition system accessibility to less represented languages.