2024诺贝尔物理学奖却颁发给了一个计算机教授？凭什么？

多伦多大学的杰弗里-辛顿教授（Geoffrey E. Hinton）和美国普林斯顿大学的约翰-霍普菲尔德教授（John J. Hopfield）

诺贝尔奖颁发机构周二表示，加拿大多伦多大学的教授杰弗里·辛顿和美国普林斯顿大学的教授约翰·霍普菲尔德因在人工神经网络中实现机器学习的发现和发明，荣获 2024 年诺贝尔物理学奖。

Jeffrey Hinton, a professor at the University of Toronto in Canada, and John Hopfield, a professor at Princeton University in the U.S., have been awarded the 2024 Nobel Prize in Physics for discoveries and inventions that enable machine learning in artificial neural networks, the Nobel Prize awarding body said on Tuesday.

出生于英国的辛顿过去十年一直在多伦多大学教授计算机科学，同时为谷歌的深度学习人工智能团队工作，最终于 2023 年宣布从 Alphabet 公司辞职。

Hinton, born in Britain, has spent the last decade splitting his time between teaching computer science at the University of Toronto and working for Google's deep-learning artificial intelligence team, before announcing his resignation from the Alphabet company in 2023.

“我大吃一惊，”辛顿在电话中告诉斯德哥尔摩瑞典皇家科学院的专家组。“我没想到会发生这种事。我非常惊讶。”

"I'm flabbergasted," Hinton told the panel gathered at the Royal Swedish Academy of Sciences in Stockholm by phone. "I had no idea this would happen. I'm very surprised."

观看CBC新闻对加拿大“人工智能教父”杰弗里·辛顿 (Geoffrey Hinton) 获得诺贝尔物理学奖的报道： 

瑞典皇家科学院诺贝尔委员会成员艾伦·穆恩斯表示，两位获奖者“利用统计物理学的基本概念设计了人工神经网络，这种网络可以作为联想记忆并在大型数据集中寻找模式。”

Ellen Moons, a member of the Nobel committee at the Royal Swedish Academy of Sciences, said the two laureates "used fundamental concepts from statistical physics to design artificial neural networks that function as associative memories and find patterns in large data sets."

她说，这种网络已被用于推动物理学研究，“并且已经成为我们日常生活的一部分，例如面部识别和语言翻译”。

She said that such networks have been used to advance research in physics and "have also become part of our daily lives, for instance in facial recognition and language translation."

霍普菲尔德发明了一种联想记忆，可以存储和重建图像和其他类型的数据模式。这些模式以“霍普菲尔德网络”的形式存储，该网络使用原子物理学对数据进行编码，然后比较、再现和区分图像。

Hopfield created an associative memory that can store and reconstruct images and other types of patterns in data, the academy said. The patterns are stored in the form of "Hopfield networks" that use atomic physics to encode the data, then later compare, reproduce and differentiate between images.

两位获奖者的工作密切相关。今天早上，在接受新闻网采访时，辛顿表示，事实上，对他影响最大的导师之一是霍普菲尔德的学生之一特里·塞诺夫斯基 (Terry Sejnowski)，后者与辛顿一起进行了“最令人兴奋的研究”——如何让“霍普菲尔德网络”更加通用，并让它们能够像人类一样从例子中学习（而不是像机器传统上那样只接受指令）。

The work of the two prize winners was closely connected. In an interview with News Network earlier this morning, Hinton said that in fact, one of his most influential mentors was one of Hopfield's students, Terry Sejnowski, who worked with him on his "most exciting research" — how to make "Hopfield networks" more general and enable them to learn from examples, as humans can (rather than just instructions, as machines traditionally did).

为了实现这一目标，辛顿使用了统计物理学，该理论基于 19 世纪物理学家 Ludwig Boltzmann 发明的方程，创建了“玻尔兹曼机”。它可以对图像进行分类，但也可以根据之前训练过的模式中特征的概率创建新的模式。

To do that, Hinton used statistical physics, based on an equation invented by nineteenth century physicist Ludwig Boltzmann, creating a "Boltzmann machine." It can classify images, but it can also create new patterns based on the probabilities of traits in previous patterns it has been trained on. 

瑞典皇家科学院在介绍辛顿的工作时解释说，这使其成为“生成模型”的早期示例。在此基础上，出现了“生成式人工智能”，例如 ChatGPT——这项技术最近因其能够从示例中创建新作品（如文章和图像）而备受关注。而且它的发展速度比辛顿预期的要快得多。

That makes it an early example of a "generative model," explains the Royal Swedish Academy of Sciences in a profile of his work. Building on this work has led to "generative AI," like ChatGPT — a technology that has recently gained lots of attention for its ability to create new works such as essays and images from examples. And it's happened much faster than Hinton expected.

他说：“我认为，如果我们能让这些东西发挥作用，那么最终它将非常重要，但我认为要产生如此大的影响还需要更长的时间。”他补充说，人工智能现在所能做的事情在 10 年前会被认为是“科幻小说”，而未来 10 年将实现类似的飞跃。

"I thought in the end if we could get this stuff to work it would be very important, but I thought it would take longer to have this big an effect," he said. He added that what AI can do now would have been considered "science fiction" 10 years ago, and similar leaps will be made in the next 10 years.

辛顿表示，他相信人工智能进步的影响将与工业革命相媲美，通过开发人工智能助手，将大大提高生产力并提高效率，这将有助于包括医疗保健在内的许多领域。

Hinton said he believed the effects of the advancements in artificial intelligence will be comparable to those of the Industrial Revolution, leading to large gains in productivity and offering efficiencies through the development of artificial intelligence assistants that will help in many sectors, including in the delivery of health care.

“这将是一场伟大的变革，这也是我们不会停止进步的原因。我认为我们不能暂停人工智能，因为它能带来很多巨大的好处，”他告诉加拿大广播公司新闻。“但我们真正需要担心的是如何控制它。”

"It's going to be fantastic, which is why progress is not going to be stopped. I don't think we can hit a pause button AI because there's so many enormous benefits from it," he told CBC News. "But we really need to worry about how to keep it under control."

辛顿在以前的采访中表示过，他离开谷歌是因为最近有关人工智能的发现让他意识到人工智能对人类构成了威胁。

正如他离开谷歌后所警告的那样，人工智能技术的进步使得其能够编写自己的计算机代码，这可能会带来意想不到的后果。他周二在加州时表示：“我们根本不知道与比我们更聪明的事物相处是什么样的。”

As he has warned since leaving Google, there could be unintended consequences resulting from the advancement of AI technologies being able to write its own computer code. "We have no experience of what it's like to have things smarter than us," he said, while in California on Tuesday.

他指责 OpenAI 等公司将利润置于安全之上，没有投入足够的资源来解决问题。

He accused companies like OpenAI of putting profits ahead of safety by not putting enough resources into solving the problem.

不幸的是，他表示，目前还没有已知的解决方案可以阻止人工智能在未来夺走人类的控制权。

Unfortunately, he said, there is currently no known solution to stop AI from taking control away from people in the future.

“我想我太老了，无法想出新点子了，”他补充道，“但我还不算太老，我可以建议政府让大公司提供资源，年轻的研究人员应该努力研究是否有办法让人们保持控制。”

"I think I'm too old to figure out new ideas about what to do," he added, "but I'm not too old to recommend that governments should make big companies provide the resources, and the young researchers should work very hard to figure out if there is a way that people can stay in control."

辛顿此前曾获得过多项荣誉，并因在计算机科学领域的贡献而荣获著名的 AM 图灵奖。同年，他因开发“使计算机能够识别语音、解释图像和在复杂数据集中寻找结构的学习算法”而被授予加拿大勋章。

Hinton has accumulated a number of previous honours, earning the prestigious A.M. Turing Award for contributions to computer science. That same year, he was appointed as a companion of the Order of Canada for developing "learning algorithms that allow computers to recognize speech, interpret images and find structure in complex data sets."

他于 1987 年首次加入多伦多大学任教，并曾担任加拿大高等研究院的顾问。

He first joined the faculty at U of T in 1987, and has been an adviser at the Canadian Institute for Advanced Research.

当被问及是否有喜欢的在线人工智能工具时，Hinton 表示他经常使用 OpenAI 的 GPT-4，但他也承认，“我并不完全信任它，因为它可能会产生幻觉。”

Asked if he had a preferred online AI tool, Hinton said he has used GPT-4 from OpenAI frequently, though allowing, "I don't totally trust it, because it can hallucinate."

诺贝尔奖的奖金为 1100 万瑞典克朗（144 万加元），如果有多人获奖，则奖金由获奖者分享。 

The Nobel prizes carry a cash award of 11 million Swedish kronor ($1.44 million Cdn), which is shared between the winners if there is more than one person honoured. 

奖金来自诺贝尔奖创始人、瑞典发明家阿尔弗雷德·诺贝尔的遗赠。获奖者将受邀于 12 月 10 日（诺贝尔逝世周年纪念日）参加颁奖仪式。

The money comes from a bequest left by the prize's creator, Swedish inventor Alfred Nobel. The laureates are invited to receive their awards at ceremonies on Dec. 10, the anniversary of Nobel's death.

就在这项物理学奖颁发的前一天，两位美国生物学家维克多·安布罗斯和加里·鲁夫昆因发现微小RNA而获得诺贝尔医学奖。

The physics award comes a day after two American biologists,  Victor Ambros and Gary Ruvkun, won the Nobel Prize in Medicine for the discovery of microRNA.

诺贝尔奖将继续公布，化学奖将于周三公布，文学奖将于周四公布。诺贝尔和平奖将于周五公布，经济学奖将于 10 月 14 日公布。

Nobel announcements will continue with the chemistry prize on Wednesday and the literature prize on Thursday. The Nobel Peace Prize will be announced on Friday and the economics award on Oct. 14.