“Torrenting from a corporate laptop doesn’t feel right”: Meta emails unsealed – Ars Technica https://arstechnica.com/tech-policy/2025/02/meta-torrented-over-81-7tb-of-pirated-books-to-train-ai-authors-say/
New open source release from Meta FAIR: Audiobox Aesthetics was trained on 562 hours of audio aesthetic data annotated by professional raters across four dimensions to create a model that enables the automatic evaluation of aesthetics for speech, music and sound. https://x.com/AIatMeta/status/1889418249466683449
Meta presents: Pippo : High-Resolution Multi-View Humans from a Single Image Generates 1K resolution, multi-view, studio-quality images from a single photo in a one forward pass https://x.com/arankomatsuzaki/status/1889515688647373113
Pippo: High-Resolution Multi-View Humans from a Single Image https://yashkant.github.io/pippo/
Our Automated Compliance Hardening (ACH) tool hardens platforms against regressions with LLM-based test generation. More on how this works and why it matters at Meta ➡️ https://x.com/AIatMeta/status/1890137619608268871
We optimized Sonar across two critical dimensions that strongly correlate with user satisfaction — answer factuality and readability. Our results show Sonar outperforms Llama 3.3 70B Instruct and other frontier models in key areas. https://x.com/perplexity_ai/status/1889392624399761869
Meta says it may stop development of AI systems it deems too risky | TechCrunch https://techcrunch.com/2025/02/03/meta-says-it-may-stop-development-of-ai-systems-it-deems-too-risky/
MobileLLM – a facebook Collection https://huggingface.co/collections/facebook/mobilellm-6722be18cb86c20ebe113e95
Using AI to decode language from the brain and advance our understanding of human communication https://ai.meta.com/blog/brain-ai-research-human-communication/
Our Approach to Frontier AI | Meta https://about.fb.com/news/2025/02/meta-approach-frontier-ai/
Cerebras Launches World’s Fastest DeepSeek R1 Llama-70B Inference – Cerebras https://cerebras.ai/blog/cerebras-launches-worlds-fastest-deepseek-r1-llama-70b-inference
Meta introduces VideoJAM This AI makes video animations smoother and more realistic by improving how motion is generated. 10 wild examples: 1. Fingers press into a shimmering slime ball. https://x.com/minchoi/status/1887526485835018284
🤔 How’s DeepSeek-R1 usage stacking up against Llama 3 on @huggingface 3 weeks after launch? Just pulled 30-day numbers: DeepSeek: 1K+ derivatives created by the community (10M downloads) 8 original models (5M downloads) Llama 3: 9K+ derivatives created by the community (30M https://x.com/fdaudens/status/1888615746789367915
