One of the Best TC Mods for Half-Life Got a New Major Update
#diffusion
Thermodynamic Computing Promises Energy-Efficient AI Images
Thermodynamic computing uses physical circuits that changes in response to noise, such random thermal fluctuations in the environment, to perform low-energy computations. A recent spate of experiments, theories and prototype hardware have shown it’s especially good at randomization tasks, and may be equally good at diffusion model tasks (e.g. image generation) in the future.
Thermodynamic Computing Promises Energy-Efficient AI Images
Thermodynamic computing uses physical circuits that changes in response to noise, such random thermal fluctuations in the environment, to perform low-energy computations. A recent spate of experiments, theories and prototype hardware have shown it’s especially good at randomization tasks, and may be equally good at diffusion model tasks (e.g. image generation) in the future.
The rapid diffusion of LLMs tests the capacity of organisations to bear loads
This was an interesting language by (the increasingly sinister) Alex Karp for something I’d been struggling to articulate: AI imposes loads on organisations, so those organisations with the capacity to bear that load have a huge advantage. He suggests “you just can’t obfuscate what can bear the load and what can’t”. It reveals organisations which have been pretending to bear the loads but are actually full of cracks. Which is exactly what I’ve been arguing with UK higher education. The problem is the timing as much as the technology itself.
Nucleus (@EsotericCofe)
간단한 3D 장면을 현실감 있는 렌더로 변환하고 바닥 반사까지 자동으로 추가하는 시연을 공유함. 해당 기술은 확산(diffusion) 기반 렌더링을 게임 엔진에 적용하는 방향으로 발전 중이라는 평가로, 향후 'diffusion game engines' 시대가 가까워졌음을 시사함.
Moonlake (@moonlake)
프롬프트에서 장면까지 이어지는 엔드투엔드 파이프라인을 소개: 자체 엔진으로 달 표면 같은 환경을 생성하고 이를 디퓨전 출력으로 변환해 유연한 미디어 자산으로 리스킨했다고 설명. 이 기능은 웹 앱 기반으로 구축되어 게임 생성과 콘텐츠 제작 워크플로우를 통합하는 사례로 보임.
Interesting read! The GPT‑2 + TTS combo sounds powerful, though diffusion‑based video models (like the ones Widvid uses) often give smoother motion and better coherence. Would love to see a comparison sometime! #GenerativeAI #Diffusion
🚀 Meet Widvid – AI video generation at lightning speed!
🎥 1080p, up to 16s clips in <10s ⏱️
✅ Text‑to‑Video
✅ Image‑to‑Video
✅ Reference‑to‑Video with flawless character consistency
🎬 Cinematic cameras & templates
Try it now 👉 [link] #AI #VideoGeneration #Diffusion #CreativeTech
Bearbeitet:
AI just disappoints
#climax #diffusion #disappointed #emotional #generate #images #local-opensource
https://www.fortschrittsanzeige.de/ai-just-disappoints/
You generate images with Stable Diffusion, and they just don't make you happy. You think "
fly51fly (@fly51fly)
새 논문 'High-accuracy and dimension-free sampling with diffusions'는 차원 수에 의존하지 않는(diffusion) 고정밀 샘플링 기법을 제안합니다. 저자 K. Gatmiry, S. Chen, A. Salim(UC Berkeley·Harvard) 공동저자이며 arXiv에 공개되어 확률적 생성모델의 샘플링 정확도와 확장성 개선에 기여할 가능성이 있습니다.
fly51fly (@fly51fly)
논문 'Reverse Flow Matching'은 온라인 강화학습을 위해 확산(diffusion) 기반과 흐름(flow) 기반 정책을 통합한 통일된 프레임워크를 제안합니다. Z Li, S Tang, N Azizan(표기된 소속 MIT)이 저자로, arXiv에 2026년 공개된 연구로 온라인 RL의 정책 표현 및 학습 안정성 개선에 기여할 가능성이 큽니다.
https://x.com/fly51fly/status/2011555863115575725
#reinforcementlearning #diffusion #flow #onlinelearning #arxiv
Brie Wensleydale (@SlipperyGem)
dx8152의 Gaussian Splash LoRA와 FAL의 다중 카메라(Multi-angle) LoRA를 비교 실험한 트윗입니다. 게시자는 1번 원본, 2번 Sharp Gaussian Splat, 3번 dx8152의 Gaussian Splash LoRA 적용, 4번 FAL의 다중 앵글 LoRA 적용 결과를 나열하고 ComfyUI-Sharp 관련 GitHub 리포를 참고하며 추가 테스트를 예고하고 있습니다.
Z.ai (@Zai_org)
GLM-Image 출시: 오픈소스 이미지 생성 모델로, 오토리그레시브와 디퓨전의 하이브리드 아키텍처를 결합해 강한 글로벌 의미 이해와 고해상도 시각적 디테일을 동시에 추구함. 저자는 전반적 품질에서 주류 디퓨전 모델과 동등하다고 주장하며, 오픈 리서치 측면에서 주목할 만한 마일스톤으로 소개함.
Euro de handball 2026 : 8 bars strasbourgeois où regarder les matchs avec ses potes
#vismaviedéditrice sur le blog du site web #mastobook #mastolivre #foires #internationales #export #diffusion #distribution #creerdesponts #culture
🆕 Unsloth ra hướng dẫn chi tiết cách chạy mô hình diffusion Qwen‑Image (2512 & Edit‑2511) trên máy cá nhân. Hướng dẫn bao gồm: • Sử dụng GGUF, FP8 trong ComfyUI, stable‑diffusion.cpp, Diffusers • Tạo workflow & prompt • Điều chỉnh siêu tham số (sampling, guidance). Nhanh chóng đưa AI tạo ảnh chất lượng lên PC!
#AI #MachineLearning #Diffusion #QwenImage #Unsloth #CôngNghệ #AIVietnam #MôHìnhAI #TạoẢnh
https://www.reddit.com/r/LocalLLaMA/comments/1q7e5jk/unsloth_published_a_guide_on_how_to_run
⏸️ Een mooie technische video over hoe een diffusion model op een nogal contra intuïtieve manier uit een hoop random pixels een beeld genereert dat dingen uit een tekst prompt laat zien, en de knoppen waar men zoal aan draait. Zoals een negatieve prompt "GEEN extra vingers" 🤣
Heldere Illustraties met een klein speelgoed 2d model.
Hầu hết LLM như GPT, Claude, Gemini dùng mô hình tự hồi quy: tạo token từng cái → gây độ trễ, chi phí cao. Mô hình ngôn ngữ diffusion bắt đầu với câu trả lời nhiễu và tinh chỉnh toàn bộ chuỗi trong vài bước song song, giảm latency 5‑10× và chi phí. Dù khó đào tạo và cần hạ tầng mới, nhưng rất hứa hẹn cho các ứng dụng thời gian thực (code autocomplete, trợ lý trong sản phẩm). #LLM #AI #Diffusion #Autoregressive #AIVietnam #TríTuệNhânTạo
I did a basic image search from StartPage yesterday, for "Sisyphus".
Wow.
I guess it's been a while since I tried this. There are lots of famous paintings and drawings around this classic character and his story, but rather than getting a lot of those, 95% of the results were very obvious "AI" slop. Awful, shitty slop. Stuff that jumps off the page saying "No human intellect or feeling was used in the production of this image".
A few big landowners are really letting their flocks shit all over the village commons. Not gonna have much of a commons left at this rate.
#commons #AI #slop #diffusion #image #search #village #VillageCommons
Client Info
Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst