Event ·Multi-source

小模型真的构成威胁吗？

Anthropic警告小模型可能数月内通过廉价IoT设备自我复制。他们在15台55美元电视盒子上测试了260亿参数模型，速度为每秒0.39个token。现有Mirai和Kimwolf等传统蠕虫已感染数百万设备。Coinbase使用Qwen-3.5-9B模型在欺诈检测上超越了Anthropic的Opus 4.5。

Current read

Anthropic警告小模型威胁，但测试显示传统漏洞才是真正问题

11 sources72° AI heatLast updated 2026/10/14 15:45:04

Evidence

11 sources

Primary sourceDavidSacks

2026/10/14 15:45:04View source

Related source 1宝玉

2026/10/13 03:05:48View source

Related source 2Simon Willison’s Weblog

2026/10/13 22:28:54View source

Related source 3kimmonismus

2026/10/12 16:56:59View source

Related source 4Artificial Analysis

2026/10/12 17:51:28View source

Related source 5rohanpaul_ai

2026/10/12 18:40:39View source

Related source 6Cognition

2026/10/12 19:32:59View source

Related source 7AI Will

2026/10/13 00:15:01View source

Related source 8Geek

2026/10/13 02:13:31View source

Related source 9岚叔

2026/10/13 05:53:21View source

Related source 10联合早报

2026/10/13 09:33:03View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story