SAM Audio is a new tool from Meta that lets users isolate sounds from audio and video sources using text, visual, or temporal prompts. It can separate general sounds, music, and speech with high accuracy, allowing for clearer audio experiences. This model also includes an open-source evaluation dataset for audio separation.

Meta is moving away from its open-source AI strategy to develop a closed, paid model named Avocado, set to launch in spring 2026. This change reflects a significant pivot in its approach, aligning more closely with competitors like Google and OpenAI. The new Chief AI Officer, Alexandr Wang, supports this transition.

Meta has unveiled Llama 4, a significant advancement in open-source AI technology, promising improved performance and accessibility for developers. This model aims to enhance the capabilities of AI applications across various industries and is expected to set new standards in the field.

Apple's decision to abandon an open-source AI initiative due to concerns over performance transparency has led to a significant loss of talent to Meta. Key figures in Apple's AI division have left, citing a clash between Apple's secrecy-driven culture and the collaborative nature of AI research, raising questions about the company's ability to compete in the evolving AI landscape.

The article announces new open-source releases from Meta aimed at enhancing the accessibility and transparency of scientific research. These initiatives are part of Meta's commitment to fostering a fair and equitable scientific environment by making resources available to the wider community.

Meta's superintelligence lab is contemplating a significant change in their A.I. strategy, potentially moving from open source models to a closed model. This shift, led by new chief A.I. officer Alexandr Wang, comes amid challenges in their A.I. technology and aims to enhance competitiveness against rivals like Google and OpenAI.