OutofAi
443 posts
🚨 We’re hiring: Data Analyst 🚨 Join us @BritishProgress to source, clean and analyse UK policy data and turn it into clear charts and sharp written analysis that drives our research programme. We’re looking for someone quantitatively rigorous, AI-native in their workflow, and genuinely interested in public policy. £33,000 - £38,000 | London (80 Strand) | 4 days in office Applications reviewed on a rolling basis: 👉 britishprogress.org/opportunities/…
The LTX-2 release (from 2 days ago) is the VEO/Sora moment for open source video generation 😍 - First open model with synchronized audio+video in one pass - 19B params, runs on consumer GPUs (~24GB+) - Lip-synced dialogue, ambient sound, music - Native 4K at 50 FPS, up to 20 sec ⬇️ Try the demo on Hugging Face (it's super fast)
⏩ We analysed Britain’s 100 fastest-growing companies and discovered that 54% have a foreign-born founder – the highest percentage on record since we began tracking it in 2019. In other words, over half of some of our most amazing businesses exist because of immigrants.
"Whatever you care about, whatever your vision for this country, it will be much easier to achieve it if we make the UK rich again." @matthewclifford at LFG: Make or Break 🚀
Exploring beyond RGB? Join us at ICRA 2025 for an exciting workshop on Thermal Infrared in Robotics (TIRO)! 🔥🤖 Share your latest thermal imaging projects in robotics or take part in our competition. 🔥 Workshop webpage sites.google.com/view/tiro25/
In student-teacher #ReinforcementLearning, the student may fail to imitate the teacher if the teacher’s behavior isn’t inferable from the student’s partial observability🤖. Our #ICLR2025 paper tackles this by training the teacher to adapt to the student’s limitations. Links ⬇️
The Australian Open doesn’t have full broadcast rights for all matches. So, its YouTube livestream uses AI to generate Nintendo Wii Tennis cartoon avatars that mimics the action on a 2-minute delay. As a result, this animated clip of Daniel Medvedev smashing his tennis racket on the net might be the greatest generative AI output to date.
Video models != world models "We find that across a range of current models (Sora, Runway, Pika, Lumiere, Stable Video Diffusion, and VideoPoet), physical understanding is severely limited, and unrelated to visual realism"
