← Back to live feed1 story
Anthropic's Frontier Red Team reports that lightweight 9B parameter models have demonstrated the ability to run inference across commercial hardware and are evolving into AI worms capable of autonomous replication on IoT devices. Skeptics argue the threat is overstated due to slow token speeds, simulated network conditions, and the high cost of silicon. Researchers nonetheless urge immediate system hardening to block potential spread.
Content history (1)
- 2026-10-12 · Summary · vi · Wording fixMột mô hình 9B parameter gần đây cho thấy khả năng chạy inference trên 8 trong số 9 loại …Một mô hình 9B parameter gần đây cho thấy khả năng chạy inference trên 8 trong số 9 loại …
Sign in to suggest
Key sources
- SOURCE@scaling01“In our evals, a 9B model got inference running on 8 of the 9 hardware types we tested”x.com
- SOURCE@andrewcurran_“Like coding agents in 2024, we’re seeing signs of life for AI worms”x.com
- SUPPORT@tftc21“Unlike non-AI malware, it could figure out unfamiliar devices on its own.”x.com
- SUPPORT@ziv_ravid“They ran a model across 15 cheap Android TV boxes at 0.39 tokens per second.”x.com
- SUPPORT@teortaxestex“silicon is so fucking expensive we won't be having a lot of smart ovens that can run 9B "replicators"”x.com
- SUPPORT@matthew_d_green“Anthropic's Frontier Red Team believes small AI models may now be months away from being capable of self replicating”x.com
- SUPPORT@scaling01“if models score above 0% on any task today, then they will score close to perfect in the near future”x.com
- SUPPORT@hamandcheese“there will be an online agent population explosion that grows to the carrying capacity of their viable revenue streams”x.com