「经济学人」Can an AI be an inventor

Discovery process探索进程

Can an AI be an inventor? The British Supreme Court considers the arguments人工智能可以当发明家吗？英国高院正在考虑。

Judicial processes involve slow, deliberative thinking. Artificial intelligence (AI) involves mind-blowing computational speed. On March 2nd, in a hearing at the Supreme Court in London, these two worlds came together to raise big questions for intellectual property in the age of machines.

1. 司法程序缓慢，需要细细考量。人工智能（AI）则计算速度惊人。三月二日，伦敦的最高法院举行了听证会，司法界和计算机碰到一起，提出一个机器时代关于知识产权的问题。

In 2019 Britain’s Intellectual Property Office denied Stephen Thaler, an American computer scientist, patents for two inventions—a beverage container and a search-and-rescue device. The applications were not denied because the inventions lacked novelty, but because the named inventor was not a human being.

2. 在2019年，英国知识产权局拒绝美国计算机科学家斯蒂芬·泰勒的两项发明专利申请，一个是饮料容器，另外一个是搜救装备。专利申请失败的原因不是因为发明缺乏新意，而是因为署名的发明人不是人类。

AI is often used as a tool to help researchers find new inventions. Applicants normally put the name of the person (or people) who worked most closely with the machine as the inventor. Until now this has not been a problem because AI has always required human assistance. But Dr Thaler asserts that his “creativity machine”, dabus (short for “Device for the Autonomous Bootstrapping of Unified Sentience”), had come up with the two inventions without any intervention and that it would be dishonest to claim that he was the inventor. He is making the same argument in other jurisdictions. But this is the first time it has reached a country’s highest court.

3. 人工智能常用于研究工具，帮助研究人员创造新发明。申请人通常是与机器关系最密切的人员。而现在，这还不存在什么问题，因为人工智能往往需要人类协助。不过，泰勒博士坚持认为他的“机器发明家” -DABUS(统感自动化引导机器的缩写)独立完成了两项发明，如果他自称为发明人，是在撒谎。在其他的司法审理中，他一直坚持这个观点。可这是第一次，上升到国家最高法庭这个层面。

The immediate stakes may not seem that high. The British government conducted a consultation in 2021 on patents and AI; despite Dr Thaler’s assertions, most respondents said AI is not sufficiently advanced to invent without human involvement. And a ruling in his favour in Britain alone would have limited commercial repercussions. “There is very little point in only the uk protecting AI inventions,” says Matt Hervey of Gowling wlg, a law firm. “If a company has a viable choice to keep its invention a secret and exploit it, it won’t disclose the secret in a patent application to get a monopoly in only one country.”

4. 眼下风险看来不大。英国政府在2021年就专利和人工智能进行的咨询协商，尽管泰勒博士坚持，大多数人仍认为人工智能还没有进化到独立完成发明工作。如果仅在英国做出有利于泰勒的裁定，其商业影响也很有限。“仅在英国保护人工智能的发明，意义不大，” WLG律所的律师马特·赫维（Matt Hervey）表示。“如果有公司秘密选择并利用AI的专利，也不会在专利申请中披露，只为在一个国家获得垄断。”

But the technology is moving fast and the questions raised by the case are profound. There is ambiguity in British law over who, or what, can claim credit for an invention. The Patents Act of 1977 was written long before its drafters could imagine robots creating anything and refers to the inventor as the “deviser”. If the Supreme Court does interpret the law to include machines, that could undermine the societal bargain on which patents are based. Inventors are encouraged to innovate on the understanding that if they disclose their inventions in a patent, they will be granted monopoly rights to recover their investment. An AI is different: robots do not require any incentive to innovate.

5. 不过，技术发展迅速，这个案子提出的问题就很深刻了。英国法律对于发明权利享受人存在模糊界定。1977年专利法案起草时，法案起草人想不到机器人可以创造发明东西，可以作为发明人。如果高院对专利法司法解释包括机器，那么将破坏专利法案存在的社会交易基础。专利是为了鼓励发明创造，如果发明人在其专利中披露了其发明，将拥有独享权利，补偿其发明投入。人工智能则不同，机器人不需要任何刺激鼓励其发明创造。

If the court decides that only a human inventor can hold a patent, its view on how rights can be legitimately transferred from an AI to humans (particularly if it is not clear how the AI works) will be closely watched. Although a machine has not yet created a pharmaceutical product on its own, Avi Goldfarb, an economist researching AI and health care at the University of Toronto, worries that if such inventions are not patentable there will be little incentive for firms to create machines that do drug research. The court is expected to deliver its verdict in a few months. It will not be the last word.

6. 如果法庭裁定，只有人类发明家才拥有专利，如何将发明权利合法地从人工智能转让人类（特别是我们不清楚AI工作机理的情况下）需要密切关注。戈德·法布（Avi Goldfarb））是一名在多伦多大学研究人工智能和医疗的经济学家。他担心，虽然机器不会自己发明医疗产品，但是如果这种发明无法获得专利，生产药品研究机器的公司将失去动力。预计法庭还需要几个月才能宣布判决。这也不是最终结果。

This article appeared in the Britain section of the print edition under the headline "Discovery process" (Mar 9th 2023)