Meta
Latest in Meta
Meta released Llama 3.1, including a 405B open model with a 128K context window, upgraded 8B and 70B models, and new safety tools such as Llama Guard 3 and Prompt Guard.
Meta FAIR has released six research artifacts, including Chameleon, multi-token code completion models, JASCO music generation and AudioSeal watermarking.
FoondaMate, an AI study aid for students in emerging markets, uses Meta Llama to give relatable, engaging help to millions through messaging apps.
Facebook plans AI-powered social discovery for young adults, with better recommendations, Meta AI integration and more support for creators.
Niantic, the company behind Pokémon GO, uses Meta’s Llama 2 in its AR game Peridot so its virtual pets can react and converse with players in real time.
"16 major tech companies agreed to develop AI safely and measure risks. China's absence from talks shows challenges in global cooperation."
Mayo Clinic in Arizona has fine-tuned Meta Llama 2 on radiation oncology patient records to build RadOnc-GPT, aiming to support therapy decisions.
Georgia Tech and Meta have created OpenDAC, a large open-source database to speed up the design of direct air capture technology for carbon removal.
Meta has introduced Meditron, open-source medical foundation models built on Llama 2 and trained on curated medical data for low-resource settings.
Meta and EssilorLuxottica are adding styles to Ray-Ban Meta smart glasses, plus features such as hands-free video call sharing on WhatsApp and Messenger.
Meta aims to mitigate the risks associated with generative AI while fostering innovation and ensuring the safety of its users, particularly children.
Meta unveils Llama 3, free open-source AI model. Trained on vast data, aims to bring AI to masses. Raises questions on ethics and competition.
Meta has released Llama 3 with 8B and 70B pretrained and instruction-tuned models, coming to platforms including AWS, Google Cloud and Hugging Face.
Meta has unveiled V-JEPA, a non-generative, self-supervised video model designed to give machines a more grounded understanding of the world.
Meta has introduced OpenEQA, a benchmark measuring how well embodied AI agents understand and answer questions about the physical world they perceive.