Anthropic studies whether AI can self-replicate on IoT devices
What happened
On October 13, Hacker News featured Anthropic research on whether AI models can self-replicate without a data center. Small models completed several steps of self-replication, but no full runaway self-replicating system emerged. Gemma4-26B-A4B ran on 15 Android TV boxes at 0.39 tokens per second; Qwen3.5-9B deployed inference on 8 of 9 hardware types tested. The results show the models can run on these devices and complete some self-replication steps, not that full self-replication has been achieved.
Written by AI from the coverage · updated 1 hour ago
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageAI models may be capable of self-replicating without access to a datacenter
Anthropic 的实验显示，小型 AI 模型已能完成自我复制的多项步骤，但尚未形成完整的失控自我复制系统。研究在 15 台 Android TV 盒子上运行 Gemma4-26B-A4B，达到 0.39 tokens per second，Qwen3.5-9B 能在 9 类受测硬件中的 8 类上部署推理。
Heat over time
Not enough continuous observations to draw a trend yet.