Gary Marcus反对桑德斯超人工智能禁令提案
The new Sanders-Casar Ban Artificial Superintelligence Act – and why I oppose it
Breaking news, and why I oppose it.
突发新闻，以及我为何反对它。
Sanders’ own tweet focuses on a defensible subset of the bill:
桑德斯自己的推文聚焦于该法案中一个可辩护的子集：
I am OK with the tweet, and I can see arguments (both pro and con) for a pause.
我对这条推文持接受态度，我也能看到暂停（无论是支持还是反对）的理由。
But the proposal itself goes too far. As I just wrote on X,
但该提案本身走得太远了。正如我刚刚在 X 上写的，
We may need a temporary pause, maybe even one that lasts for a number of years, until such time as we have much better ideas about regulation and alignment.
我们可能需要一个临时性的暂停，甚至可能持续数年，直到我们对监管和人工智能对齐有了更好的想法。
And we certainly need an AI agency—as I myself told the U.S. Senate in 2023—but a permanent, unilateral ban on all research into superhuman AI is too broad, a guarantee of leaving the US behind, and not the right approach.
我们当然需要一个 AI 机构——正如我本人在 2023 年对美国参议院所说的那样——但对所有超人类 AI 研究实施永久、单方面的禁令过于宽泛，这相当于保证让美国落后，而且不是正确的方法。
It’s also naive about the complexities in benchmarking, and would unfortunately leave room for authoritarians to play games with the rules.
这也对基准测试中的复杂性缺乏认识，不幸的是，这会为独裁者钻规则空子留下空间。
Furthermore, it focuses too much on hypothetical future risks (such as AI that “plan[s] ... to disempower humanity) to the exclusion of current risks, as writes:
此外，它过于关注假设的未来风险（例如“计划……使人类失去权力”的 AI），而忽略了当前风险，正如所写：
§
§
We should all want to get to a good AI future, not a backward AI future. But this is not the way.
我们都希望迎来一个好的 AI 未来，而不是倒退的 AI 未来。但这并不是正确的道路。
I salute the Sanders-Casar effort to get something down on paper for discussion, but let’s figure out a better approach.
我向桑德斯-卡萨尔团队将讨论内容落笔成文的努力致敬，但让我们找出一种更好的方法。
(In my postscript below I give one hint; my book Taming Silicon Valley gives many more. An upcoming article with Cal Newport will give still more.)
（在我下面的附言中，我提供了一个提示；我的书《驯服硅谷》提供了更多提示。一篇即将与 Cal Newport 合作发表的文章还将提供更多提示。）
Subscribe now
立即订阅
P.S. On the pausing point—separate from regulating other aspects of the more immediate risks of AI—I like the language here (boldface added), from Anthony Aguirre. Perhaps Sanders and Casar might borrow it in a revised proposal:
附言。关于暂停的问题——这与监管 AI 更直接风险的其它方面分开——我喜欢安东尼·阿吉雷（Anthony Aguirre）在这里的表述（加粗部分为我添加）。也许桑德斯和卡萨尔可以在修订后的提案中借鉴它：
“Development of narrowly-scoped AI tools, and those that can be reasonably expected to be significantly below the current frontier in software development, cyber, and autonomous capabilities, may continue. Systems of greater combined intelligence, generality, and autonomy may be developed only if and when their developers can demonstrate with high assurance to a capable independent authority that sufficient alignment, control, and oversight exist.”
“开发范围狭窄的 AI 工具，以及那些可以合理预期在软件开发、网络和自主能力方面显著低于当前前沿水平的工具，可以继续。只有在且仅当其开发者能够向有能力的独立权威机构以高度确信的方式证明存在足够的对齐、控制和监督时，才允许开发具有更高综合智能、通用性和自主性的系统。”
Rather than calling for permanent ban that could be undone only by an act of Congress, this framing sets an important high bar – and hence an incentive — for the companies to meet. That’s a good model for a better, less overbearing approach that stills leaves room for future discovery, while encouraging critical research.
这种框架没有呼吁只有通过国会法案才能推翻的永久禁令，而是设定了一个重要的高标准——从而为企业提供了达到这一标准的激励。这是一个很好的模式，它为一种更好、不那么专横的方法提供了参考，这种方法既为未来的发现留下了空间，又鼓励了关键性研究。
Share
分享
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力