Mateusz Malinowski

Today Roberto and Marianna walk through PhysicalRealismBench-U — how the benchmark works, how they built it, and what VLMs still get wrong about basic physics. 🔗 link.reka.ai/labs/physical-… #PhysicalAI #AI #RekaLabs #ComputerVision #VLM #PhysicalRealismBench

World models are increasingly central to how agents learn and plan. Today we're releasing WorldModelGym, a benchmark built around a single question: if an agent uses a world model to choose among actions, does it pick the right one? We call this decision-based fidelity. 100+ tracks across Atari, Meta-World, DeepMind Control, and classic control. One frozen policy. Reality scores it. Read the full post → reka.ai/labs/research/…

Most evaluations of world models ask, "Does this look right?" 🌎 We built WorldModelGym to ask a different question: if an agent actually uses this model to make decisions, does it still choose well? 🏋️🏃‍♂️🏁🦾 A simulation can look physically plausible and still lead an agent completely astray. 📉⚠️ Stay tuned. More tomorrow at reka.ai/labs. 👀

The result

Input image:

I wondered if I can use Spatial Prompting, without an input photo, to Marey by @moonvalley ? I tried it, and the results surprised me! 🧵

Welcome to a world where you don’t just generate videos - you direct them. All with legally safe Marey!

Introducing Marey by Moonvalley, the world’s first fully licensed AI video model built for professional production. Marey transforms filmmaking today. Its unprecedented creative controls enable you to realize expansive visions, execute complex VFX sequences, and maintain complete creative authority throughout your projects. This is generative filmmaking for those who need to control every frame. We can't wait to see what you make. bit.ly/MeetMarey

TED has long been a source of inspiration for me, so it is especially rewarding to see our text-to-video model, named Marey, empowering creative artists to build that opening.

What does 40 years of TED Talks look like? For the opening of #TED2025, we worked with @Asteria_FilmCo to find out. This custom film uses a dynamic blend of motion graphics, gaussian splats and generative AI elements to pay tribute to the history of innovation on the TED stage:

While the USA and U.K. consider training AI models on copyrighted material, potentially bypassing artists' rights, we should develop a system that equitably benefits both creative professionals and AI development, ensuring that everyone's interests are protected and valued.

#ai, #ArtificialIntelligence , #Media, #Hollywood , #MachineLearning , #vision

Marey, the first clean text-to-video model, has finally landed 🌙 One thing is clear - the future of filmmaking is set to transform in extraordinary and beautiful ways where everyone receives fair share. Now it is my turn, I'm creating my first movie using Marey!

Stay Tuned

We're sharing progress on our video-to-audio (V2A) generative technology. 🎥 It can add sound to silent clips that match the acoustics of the scene, accompany on-screen action, and more. Here are 4 examples - turn your sound on. 🧵🔊 dpmd.ai/v2a

An impressive new kind of generative foundation model! Trained completely unsupervised, it generates endless *controllable world models* that are (1) controllable via interpretable discrete actions (2) generated based on image prompts including drawings! So proud of the Genie team for this work (my contribution was minor)

Student researcher position applications are open at Google Deepmind! I'm hosting a SR in the intersection of bias and generative models. If you're an interested PhD student please reach out! google.com/about/careers/…

@sociotiose @sharongoldman Examples? Most hypotheses are accepted / rejected based on simplicity, impact, novelty, and how they work on the existing benchmarks. Here are some reviews: openreview.net/group?id=ICLR.… How do you think the process should look like?

🧵NEW in today's AI Beat: According to @sociotiose, today's AI is not about science - it's about alchemy, rooted in magical metaphors. I thought about that as I read today's story by @oliverwhang21 in the @nytimes, 'How to Tell if your AI is Conscious" /1 venturebeat.com/business/today…