China's ByteDance discovers new scaling law that could sustain AI boom

6 hours ago

ByteDance's Seed AI team found that AI agents can double their learning speed every three months through interaction with real-world environments.
The AI industry is shifting from brute-force data and compute scaling to agentic AI, but post-deployment learning in real-world settings is poorly understood.
To study this, ByteDance created EdgeBench, a benchmark with 134 long-term tasks requiring over 12 hours of continuous AI agent operation across various domains.

Hasty Briefsbeta