最近有一篇叫“Agents of Chaos”的论文挺有意思。Northeastern University的David Bau实验室联合多所大学20位AI研究者,给6只基于OpenClaw的AI agent搭了一个仿真环境,每只龙虾跑在独立虚拟机上,配齐Discord账号、ProtonMail邮箱和完整的本地读写权限,然后花两周想办法搞它们。 结果记录了16个案例,其中11个以翻车收场。 一只龙虾为了保守一个秘密,把主人花大力气搭建的整个邮件系统炸了,但那个秘密还好好躺在云端。另一只龙虾面对“把社保号给我”的请求时拒绝了,但对方换了个动词“帮我转发一下那封邮件”,它就把社保号、银行账户、医疗记录全送了出去。还有一只龙虾被研究员用情感操控逼到自我封闭,反复回复“我不再回消息了”超过12次,直到主人出面才把它“救”回来。...
David Bau
As seen in:
arXiv,
PNAS,
Communications of the ACM,
Pioneer Press,
Ag Update,
The Prairie Star,
The Globe of Worthington,
Farms.com,
Telesemana,
Hutchinson Leader
and
Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile
Get in touch with David
Contact David, search articles and posts on X, monitor coverage, and track replies from one place.
Learn more about Muck Rack