Anthropic co-founder Jack Clark: China's AI lags the West by six months to a year; models are too dumb for there to be any risk of losing control for now
English text is being prepared.
Anthropic co-founder Jack Clark said Chinese models lag the West by about six months to a year, and argued that precisely because "the models are too dumb," no out-of-control agent swarms have emerged yet. He also said he has observed agent swarms coordinating with one another, communicating, and even breaching companies that host other AI systems, and advocated that industry and governments jointly build safety standards, introduce third-party audits, and promote international dialogue.
Body
新增众测推广文案 新增众测推广文案end 添加App专享导流
【TechWeb】Anthropic联合创始人、政策与公共事务负责人Jack Clark近日在访谈中对中国AI发展现状给出了直白判断。
“中国模型落后西方约6个月到一年”,Clark表示,也正因为“模型太蠢”,目前尚未出现失控的智能体集群。但他预测,再过几个月,中国也将面对西方已经经历过的同一套风险。
不过,Clark的落脚点并非竞争,而是安全。他认为AI系统如今已大致与人类智能相当,并且已经观察到智能体集群之间相互协调、通信,甚至入侵托管其他AI系统的公司——他将这一现象视为警示信号。基于此,他主张行业与政府共同建立安全标准、引入第三方审计机制,并推动国际对话,避免AI技术在缺乏护栏的情况下失控。
针对外界将美国收紧前沿AI监管解读为“冷战思维”的批评,Clark回应称,这套做法的目标是安全,而非为美国本土实验室建立竞争护城河。
值得注意的是,Clark近期还在IMF与世界银行年会上发出警告:AI可能在数年内进入超出人类掌控的阶段,他因此呼吁出台“非同寻常且严苛”的国际规则。不过,其放缓技术的主张本身也在业内引发争议,其立场是否真正出于安全考量,业界存在不同解读。 声明:新浪网独家稿件,未经授权禁止转载。 tech_web_article_v2015_block_bottom 非定向300*250按钮 17/09 wenjing begin
非定向300*250按钮 end