英国没有自己的前沿人工智能实验室，却如何在人工智能泛滥的世界中蓬勃发展？答案的一部分在于法院。如果肯下苦功，伦敦可以成为原告提起新一代人工智能相关纠纷的地方，杰弗里·沃斯爵士这样表示。他本周卸任英格兰和威尔士第二高级法官、民事法院监督人“首席书记官”一职，并留下一些临别建议。
HOW DOES Britain, with no frontier lab to its name, thrive in a world of rampant artificial intelligence? Part of the answer lies in the courts. London can be a place where claimants bring a new generation of disputes relating to AI—if it puts in some hard graft, says Sir Geoffrey Vos. He steps down this week as the Master of the Rolls, the second-most-senior judge in England and Wales who oversees the civil courts. He has some parting advice.
维持伦敦的成功，他说，取决于确保英国法律能为人工智能将引发的新颖法律问题提供确定且可预测的答案。“这向世界传递的信息是，如果你想要一种真正在思考我们这个现代时代问题的法律，英国法律就是其中之一，”杰弗里爵士说，“所以请来这里解决那些争端。”
Sustaining London’s success, he says, relies on ensuring that English law can give certain and predictable answers to the novel legal problems that AI will throw up. “What this says to the world is, if you want a law which is really thinking about the problems of our modern age, English law is one of those laws,” says Sir Geoffrey, “so come here to resolve those disputes.”
他领导了一个工作组，旨在界定普通法对一系列假设性人工智能问题的含义。如果人工智能工具诽谤他人，或者人工智能架构师设计了一座有缺陷的建筑，人类是否应承担责任？如果放射科医生未能使用本可检测出肿瘤的人工智能工具，是否构成疏忽？一份即将发布的文件将探讨代理型人工智能工具，这类工具能代表用户自主执行任务。工作组成员、出庭律师雅各布·特纳表示，挑战在于“人工智能是我们文明史上开发的唯一一种能够自主做决定的技术。”
He has led a task-force that has set out to define what the common law implies for a range of hypothetical AI problems. Is a human liable if an AI tool libels someone, or if an AI architect designs a faulty building? Is a radiologist negligent if they failed to use an AI tool that could have detected a tumour? A forthcoming paper will deal with agentic AI tools, which carry out tasks autonomously on behalf of a user. The challenge, says Jacob Turner, a barrister and member of the task-force, is that “AI is the only technology that we have ever developed in the history of civilisation that is able to make its own decisions.”
杰弗里爵士说，该工作组试图“短路”法院可能需要数年才能解决的问题——正如18世纪法院花了数年才裁定股份是否属于一种财产形式。“你无法承担等待十年，让最高法院告诉你：如果你未能为代理型人工智能设置护栏从而造成人员伤害，你就要承担责任，”他说。“如果人们要选择英国法律来裁决他们的技术商业纠纷，这些事情至关重要。”
The task-force is an attempt, Sir Geoffrey says, to “short-circuit” questions that the courts might take years to settle—much as it took years for the courts to decide in the 18th century whether shares were a form of property. “You can’t afford to wait ten years for the Supreme Court to tell you that you’re liable if you fail to put guardrails in place in agentic AI and thereby cause people harm,” he says. “Those things are critically important if people are going to choose English law to determine their technological commercial problems.”
自20世纪70年代以来，伦敦的律师凭借法官廉洁、法院提供“劳斯莱斯级”服务的声誉，接纳了一波又一波的国际争端，赚得盆满钵满。但迄今为止，人工智能带来的红利微薄，仅有少数几起轰动性案件。去年11月，高等法院基本驳回了媒体公司盖蒂图像对一家人工智能公司的指控，后者被指控在训练模型时侵犯版权。
Since the 1970s, London’s lawyers have earned a handsome living by welcoming successive waves of international disputes, thanks to a reputation for honest judges and courts offering a Rolls-Royce service. But so far the AI dividend has been modest, with only a handful of blockbuster cases. Last November the High Court largely rejected claims by Getty Images, a media company, that an AI firm had infringed copyright while training its models.
越来越多的案件出现在美国法庭上。而且，来自新加坡和迪拜等竞争中心的竞争也十分激烈。
Many more are turning up in American courts. And there is stiff competition from rival centres such as Singapore and Dubai.
伦敦法庭之所以兴盛，是因为英格兰法律管辖着世界上大部分地区的商业和交易。但这种历史主导地位正面临风险，杰弗里爵士认为。各行业正竞相通过与大型实验室签订合同来采用人工智能。占主导地位的实验室要么是美国的，要么是中国的，并按照美国或中国法律运营。“你现在有一种基本的默认立场，即你所做的一切都将由人工智能驱动，而这种人工智能将首先根据美国法律提供。”
London’s courts have thrived because English law governs business and transactions across much of the world. But this historical dominance is at risk, Sir Geoffrey suggests. Industries are racing to adopt AI through contracts with the big labs. The dominant ones are either American or Chinese, and operate according to American or Chinese law. “You’ve got now a sort of basic default position that everything you’re doing is going to be powered by AI, and that AI will be prima facie supplied under American law.”
他说，其结果是，“缺乏数字主权可能导致缺乏法律主权。”大法官一职可追溯至13世纪，普遍刻板印象将英国法官描绘为年迈的技术恐惧症患者。但在杰弗里爵士身上，情况并非如此。在最近的演讲中，他向律师们发出了直言不讳的警告：希望通过禁止使用人工智能来保护法律行业免受其影响的想法是一条“无望的道路”。他们必须“走在前列”，并“适应或灭亡”。
The upshot, he says, is that “a lack of digital sovereignty can lead to a lack of legal sovereignty.”The Master of the Rolls is a post dating to the 13th century, and popular stereotype portrays English judges as ageing technophobes. In Sir Geoffrey’s case, hardly. In recent lectures he has issued blunt warnings to lawyers: hopes to shield the legal industry from AI with prohibitions on its use are a “hopeless avenue”. They must “be in the vanguard” and “adapt or die”.
人们已经习惯性地向AI寻求法律建议。杰弗里爵士认为，如果司法机关显得反科技或“保护主义”，就有失去公众信任的风险。“法官被历史束缚，不像社会其他成员那样行事，这极不可取，”他说。他对“AI提交的材料容易出错和产生幻觉”这一抱怨不以为然，技术会改进。“我会提出更根本的问题……我们如何在司法系统中最好地利用AI，以一种人们会信任的方式，为我们服务的民众伸张正义，”他说。
People already routinely ask AI for legal advice. Sir Geoffrey has suggested that if the judiciary appears Luddite or “protectionist”, it risks losing public trust. “It’s extremely undesirable for judges to be baked in history, and not behaving in the same way as other members of our society,” he says. He has little time for the complaint that AI submissions are prone to errors and hallucinations; the tech will improve. “I would be asking far more existential questions…how can we best use AI within the justice system to deliver justice to the people we serve in a way they will trust.”
在英格兰，官方指南称AI可以“协助”法官，但“绝不能取代推理……或司法裁量权的行使”。界限“难以划定”，杰弗里爵士说。他最近访问了深圳，那里的一个基于AI的法院系统通过阅读文本并为法官提供可引用的起草理由，被认为使司法产出增加了六倍。
In England official guidelines say that AI can “assist” judges, but it “must not displace reasoning…or the exercise of judicial discretion”. The dividing line is “hard to draw”, says Sir Geoffrey. He recently visited Shenzhen, where an AI-based court system is credited with a sixfold increase in judicial output by reading texts and providing draft reasons for judges to cite.
最终，他强调，这些是议会应当决定的界限，而非法官。但他主张，人类法官作为人们权利的最后保障，将不可替代：机器无法拥有完全独立的人类所享有的独特性或同理心。他推理称，这对于公众接受法院的判决至关重要；人们可能更信任一个可靠性为90%、但让人感觉倾听了其申诉的独特性人类法官，而非一个完全可靠的机器，他说。也许，正义不仅仅是吐出正确答案。
Ultimately, he stresses, these are limits for Parliament to decide, not judges. But, he has argued, human judges will be irreplaceable as the last resort for people’s rights: machines are incapable of the idiosyncrasy or empathy that fully independent humans enjoy. That is essential, he reasons, if the public is to accept a court’s verdict; people may well have more trust in an idiosyncratic human judge who is 90% reliable, but who they feel has listened to their grievance, than they would in a perfectly reliable machine, he says. There is, perhaps, more to justice than just spitting out the right answer.