Intellectual hub of the topic
Hume AI has open-sourced TADA, a speech model that performs frame-by-frame alignment of text and audio, making speech synthesis fast and predictable.
Runway has released its 'Characters' tool, which allows you to maintain a character's appearance across different scenes, ensuring they remain recognizable.
Indian company Sarvam AI has open-sourced two large language models – 30B and 105B – with a focus on supporting the languages of India.
This article examines the accuracy of AI transcription for pharmaceutical names, identifies which models perform best, and explains the importance of this for medicine.
Voice AI agents can already do a lot, but they are still far from achieving full autonomy. Let's explore what elements are missing for the next step in their development.
Inception Labs has released Mercury 2, a new generation of diffusion language models that generate text in a fundamentally different way than the AI assistants we are accustomed to.
Lab
Computer Science
Researchers trained a language model on synthetic languages and found that AI learns some grammatical patterns intuitively, while others it seems to miss entirely.
We explore how small language models learn to distinguish between important and incidental personal information in queries to preserve privacy without losing meaning.
Anthropic has released Claude Sonnet 4.6, an updated model featuring improved context understanding, more accurate responses, and honest behavior in complex situations.
Don’t miss a single experiment!
Subscribe to our Telegram channel —
we regularly post announcements of new books, articles, and interviews.