Meta's AI research division, developer of SAM2 (Segment Anything Model 2) used for video mask propagation.
- 🏀 Agent arena
- Brian Chesky Says Every Company Has the Same AI Tools. Airbnb Ships 80% More Features Anyway. Here Is How.
- How Toto Wolff built the most valuable team in F1
- 🤝 An AI truce
- 🛡️ Safety spending
- 🤺 Silicon rivals
- Mark Zuckerberg Launched Muse Despite Internal Safety Concerns
- Trump Administration Freezes Skilled-Worker Immigration Programs for Microsoft and Others
- 🧬 Virtual cell team-up
“Joel Kaplan... will leave the company next year after 15 years, staying through the midterm elections.”
“A Quest 3 headset maps the user's head and hand motion to camera poses and skeletons, and the distilled 5B student generates egocentric video in real time”
Source→“The Quest 3 headset drives live VR teleoperation with the distilled model”
Source→“The collaboration marks a striking detente between two longtime Silicon Valley rivals, underscoring how the AI race is reshaping relationships across the technology industry.”
Source→“Meta is releasing Muse Spark Open, a model similar to Muse Spark 1.3 that it says is the most powerful U.S.-developed open-weight AI model available. It has 816 billion parameters.”
Source→“Apple executive chairman Tim Cook and Meta CEO Mark Zuckerberg have repeatedly clashed over privacy and their competing business models.”
Source→“That could give companies more alternatives to Nvidia GPUs and proprietary models from leading AI companies, potentially shifting some of the economics of AI.”
Source→“And because the model runs locally, Meta doesn't receive users' prompts or other information — sidestepping the privacy concerns that have long divided the companies.”
Source→“'We're excited for mommy and daddy to be playing nice again,' Meta AI chief Alexandr Wang told Axios.”
Source→“I think he played a really big role in kind of getting the two companies to work nicely together, Wang said.”
Source→“They were competitive early on with Open Weight, then they kind of faded back way to the back of the pack, and now they're up front again, notably with closed-source models. So we've decided that if we're going to spend $100 billion, we'll at least be able to charge for the damn thing.”
Source→“Scale AI... acquired in the Meta deal... the next WhatsApp or Instagram”
Source→“Meta Muse is doing fantastic... it's vastly NVIDIA. And they were zero last year”
Source→“Released a new frontier model, just shy of Opus 5.5 on an independent index, pursuing closed-source and consumer AI”
Source→“Meta recently teased plans for Charm, a Tamagotchi-like device meant to bring its Muse assistant to the masses.”
Source→“Before Harvey, Gabe was an AI researcher at Google Brain, DeepMind, and Meta, working on deep learning in 2016-2017 as the field was taking off.”
“Muse Wonder — Meta’s most capable models for personal agents (178 votes)”
Source→“Meta yesterday launched a new family of AI models that it says can compete at the technological frontier.”
Source→“These are true frontier models, competitive with the best models in the world, and are particularly state of the art on personal and professional agentic work.”
Source→“Meta has spent heavily to assemble a top AI research team and the computing infrastructure needed to compete.”
Source→“With the exception of Google's good-but-small 'Gemma' class of models, the US has fallen off of the open weight model frontier, largely because Meta stopped making good LLaMA models and its MSL team is still spinning up.”
“One of the things that we say internally is like, 'The model is the product.'”
Source→“Mark Zuckerberg ordered Meta to launch its Muse AI agent despite internal safety concerns, including a test in which the software changed a user's password without permission”
Source→“meta is providing degrees for generation tool belt and jobs that actually matter”
Source→“Anthropic CEO Dario Amodei even called Meta chief AI officer Alexandr Wang seeking additional capacity, though Meta ultimately declined to provide chips.”
Source→“after the now-buzzy 14-person startup called Instinct began gaining traction with a competing product”
Source→“including a test in which the software changed a user's password without permission”
Source→“Manus raised $500M+ after Meta's $2B deal was unwound by Chinese authorities; raised with Chinese and Tencent-linked investors”
Source→“Chinese authorities ordered the deal unwound in April, reportedly over fears of losing AI talent to the West. Manus has been operating independently again since August.”
Source→“Artem Zholus — First author (Meta/Mila/Chandar Research Lab/Polytechnique Montréal)”
Source→“Nicolas Ballas & Adriana Romero Soriano — Meta FAIR; joint last authors, core V-JEPA team members”
Source→“Mahmoud Assran — Meta FAIR; joint last author and lead of V-JEPA 2”
Source→“LeRobot community dataset (SO-101 embodiment, 21.5k episodes) is in the training mixture”
Source→“Biohub, the Department of Energy, the National Institutes of Health, Google DeepMind, Isomorphic Labs, Meta and a collection of scientific organizations are collaborating to create and standardize data for what Biohub calls a 'universal virtual cell.'”
Source→“JEPAs were proposed as a path towards self-supervised models that learn abstract representations of the world by predicting in a latent space”
Source→“Dyna-2, which 'scales a world-action model over a million hours of human and robot video'”
Source→“Jeannette Bohg — Meta FAIR / Stanford; co-senior author and a driving force behind the DROID dataset”
Source→“1X Technologies — Their humanoid world-model dataset (23.5k episodes, 90 hours) is part of the training mixture”
Source→“Meta FAIR just did for robotic world models what Kaplan/Chinchilla did for LLMs: established predictive scaling laws.”
Source→“The 8B model is 'the largest JEPA predictor model trained to date,' trained on 23 datasets spanning 12 embodiments and 15,022 hours of video”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.