Loading activity…
@sedielem
Research Scientist at Google DeepMind (WaveNet, Nano Banana, Gemini Omni). I tweet about ML, music, generative models (personal account).
AI
The researchers report 1.19 FID on ImageNet 256×256 using 12X less compute than Drifting-B.
AI
The researchers say their work explores design choices for continuous diffusion language models, including how to embed tokens.
AI
The H-CDLM announcement claims denoising hierarchical token representations brings better generations for less compute.
AI
A post says looping at a given noise level also reduces reliance on CFG.
AI
A post announcing the preprint says the method reached 1.91 FID on ImageNet-256 at higher classifier-free guidance scales.
AI
A blogger says their new post explores how you could invent every transformer layer by thinking about memory.
AI
A researcher calls the slicing clever but warns that the chosen model's biases remain.
AI
One post says Dominik's result looked more convincing than expected; another praises novel ideas rather than just scaling.
AI
Creator Diogo Almeida claims Jev is 20–200x faster and 40–400x cheaper, with free output tokens. Investor DCVC says it led TypeSafe’s $40 million seed round.
AI
A September 12, 2026 hiring announcement describes work on world models with the people who built Veo, Genie and Gemini Omni. A follow-up names Mountain View and San Francisco.
AI
Dieleman says he spoke with Fundamental CSO Marta Garnelo, a friend and former colleague, about diffusion models, music generation, blogging and Kaggle.
AI
AI researcher replies to original creators praising the Inceptionism infinite zoom videos.
AI
Google DeepMind researcher posts historical perspective on the approach's recent resurgence.
