参议员伯尼·桑德斯和众议员格雷格·卡萨尔提出了一项法案，旨在永久禁止美国的人工超级智能。《禁止人工超级智能法案》还将暂停最先进人工智能的研发工作，并设立人工智能部。
Senator Bernie Sanders and Representative Greg Casar have introduced a bill to permanently ban artificial superintelligence in the US. The Ban Artificial Superintelligence Act would also pause work on the most advanced AI. And it would create a Department of Artificial Intelligence.
这两位立法者于周三提出了该法案。他们首先将其提供给美联社，由美联社的乔伊·卡佩莱蒂进行了报道。桑德斯是来自佛蒙特州的独立参议员。来自德克萨斯州的民主党人卡萨尔担任国会进步核心小组主席。
The two lawmakers introduced the bill on Wednesday. They gave it first to the Associated Press, whose Joey Cappelletti reported it. Sanders is an independent from Vermont. Casar, a Texas Democrat, chairs the Congressional Progressive Caucus.
“当你正冲向悬崖时，你不能只是松开油门，你必须猛踩刹车。当人类的未来岌岌可危时，我们不能让寥寥无几的大型科技公司首席执行官自行制定规则，”桑德斯在一份声明中说。
“When you are racing towards a cliff, you don’t just ease up on the gas pedal. You hit the brakes. When the future of humanity is at stake, we cannot let a handful of Big Tech CEOs write their own rules,” Sanders said in a statement.
超级智能禁令涵盖的内容 法案文本禁止构建或部署人工超级智能。它从两个方面对其进行定义。一是在大多数领域或任务中超越人类认知表现的系统。二是能够计划并实施毁灭人类或剥夺人类权力的系统。
What the superintelligence ban covers The bill text bans building or deploying artificial superintelligence. It defines that in two ways. One is a system that beats human cognitive performance in most domains or tasks. The other is a system able to plan and carry out the destruction or disempowerment of humanity.
该法案还将暂停先进人工智能系统的研发工作。分节摘要将这些系统定义为使用10^25次或更多运算进行训练的系统。暂停将立即生效，并将在新部门完全配备人员并制定明确规则后结束。
The bill would also pause work on advanced AI systems. The section-by-section summary defines these as systems trained on 10^25 or more operations. The pause would start at once. It would end when the new department is fully staffed and has set clear rules.
此后，任何构建或销售先进人工智能的公司都需要获得该部门的特许经营权。摘要将系统改变自身功能的能力列为需要警惕的警示信号。Semafor的尼古拉斯·吴报道称，该法案禁止“递归自我改进”。
After that, any company that builds or sells advanced AI would need a charter from the department. The summary lists a system’s ability to change its own functions as a warning sign to watch for. Semafor’s Nicholas Wu reported that the bill bans “recursive self-improvement”.
监禁刑期与“企业死刑” 违反禁令或暂停令者最高可面临20年监禁。该法案的单页摘要将其与非法制造核武器的处罚进行了比较。
Prison terms and a “corporate death penalty” People who break the ban or the pause could face up to 20 years in prison. The bill’s one-pager compares that with the penalties for unlawfully building nuclear weapons.
企业将面临该单页文件所称的“企业死刑”。违规公司将被吊销特许经营权，并须将其知识产权和资产移交联邦政府。该法案保护向该部门举报超级智能风险的员工。
Companies would face what the one-pager calls the corporate death penalty. A company that breaks the rules would lose its charter. It would also hand its intellectual property and assets to the federal government. The bill protects employees who report superintelligence risks to the department.
在国际上，美国将寻求达成国际协议并实施AI算力出口管制。目标是防止超级智能在任何地方被构建。
Abroad, the US would seek international agreements and export controls on AI computing. The aim is to stop superintelligence being built anywhere.
卡萨尔表示：“我们的法案禁止开发人工超级智能，并推动达成国际协议，确保没有人、在任何地方构建出人类无法控制的过于强大的AI。”
“Our bill bans the development of artificial superintelligence and pushes for international agreements so that no one, anywhere, builds AI too powerful for humans to control,” Casar said.
实验室内部的支持 几位AI公司员工在一份共享给美联社的声明中支持该法案。其中一位是OpenAI安全系统团队的研究员胡安·费利佩·塞龙·乌里韦。他表示，超级智能“要么会发展得极好，要么会发展得极坏”。
Support from inside the labs Several employees at AI companies backed the bill in a statement shared with the AP. One is Juan Felipe Cerón Uribe, a researcher on OpenAI’s Safety Systems team. He said superintelligence “could either go extremely right or extremely wrong”.
谷歌DeepMind软件工程师斯万特·肖尔茨以个人名义表示：“禁止超级智能开发在可预见的未来将是一个积极的变化。”
“A ban on superintelligence development would be a positive change for the foreseeable future,” said Swante Scholz, a Google DeepMind software engineer speaking for himself.
桑德斯告诉美联社，避免最危险的AI需要美中合作，就像冷战时期的核军控一样。上周，他与史蒂夫·班农共同呼吁AI放缓发展。本月初，他主持了一场关于AI风险的参议院简报会。桑德斯对美联社说：“不用天才也能说出‘慢下来’这句话。”
Sanders told the AP that avoiding the most dangerous AI would require cooperation between the US and China, as nuclear arms control did in the Cold War. Last week he called for an AI slowdown alongside Steve Bannon. Earlier this month he hosted a Senate briefing on AI risk.
国会前景 美联社报道，该法案在共和党控制的国会中前景渺茫。已有三项其他AI法案在那里等待，尚无投票日期。
“It doesn’t take a genius to say, ‘slow it down’,” Sanders told the AP.
特朗普总统则推向相反方向。周二他在联合国表示，美国现在将把AI称为“超级智能”。他说美国将鼓励这项技术，而非加以限制。美联社报道，美国放缓而中国继续前进是反对广泛限制的主要论据之一。
The odds in Congress The bill faces long odds in the Republican-controlled Congress, the AP reported. Three other AI bills are already waiting there with no date for a vote.
“我认为特朗普对这一问题的实质知之甚少，”桑德斯告诉美联社。“我认为他主要关心的是放缓带来的经济影响。”
President Donald Trump has pushed the other way. On Tuesday he told the UN that the US would now call AI “super intelligence”. He said the US would encourage the technology, not rein it in. A US slowdown while China keeps going is one of the main arguments against broad limits, the AP reported.
据Semafor报道，桑德斯周三对记者表示，国会必须在复会后尽快采取行动。他还希望2028年总统候选人现在就着手解决这一问题。
“I think Trump has very little understanding of what this issue is about,” Sanders told the AP. “I think he’s mainly concerned about the economic implications of a slowdown.” Sanders told reporters on Wednesday that Congress must act as soon as it returns, Semafor reported. He also wants 2028 presidential candidates to take up the issue now.