（双语文稿）如何确保人工智能系统（AI）安全

Major AI developers and top government officials have agreed on measures that aim to ensure that artificial intelligence (AI) systems are safely built and deployed.
主要人工智能开发人员和政府高级官员已就旨在确保人工智能(AI)系统安全构建和部署的措施达成一致。
The announcement came at the end of a two-day meeting in Seoul, South Korea. The gathering continues discussions from a similar event held last November in Britain. The goal of both meetings was to bring AI developers together with government representatives to establish goals and guidelines related to safe AI development.
在韩国首尔举行的为期两天的会议结束时宣布了这一消息。这次会议延续了去年11月在英国举行的类似会议的讨论。这两次会议的目标都是将人工智能开发人员与政府代表聚集在一起，制定与安全人工智能开发相关的目标和指导方针。
Many governments and international organizations across the world are seeking to create rules for AI technology development. Officials and activists have raised concerns that AI systems could present risks. The concerns include AI possibly attacking online systems controlling infrastructure. Another concern is that AI systems might help spread misinformation or create "deepfakes" to mislead voters in elections.
世界各国政府和国际组织都在寻求制定人工智能技术发展的规则。官员和活动人士对人工智能系统可能带来风险表示担忧。这些担忧包括人工智能可能会攻击控制基础设施的在线系统。另一个担忧是，人工智能系统可能会帮助传播错误信息或制造“深度造假”，在选举中误导选民。
Legislative attempts to create rules for AI development, especially in the United States, have not been successful. Last October, President Joe Biden issued an executive order calling on companies working on AI projects to develop guidelines for protecting users of the technology.
为人工智能发展制定规则的立法尝试并不成功，尤其是在美国。去年10月，乔•拜登总统发布了一项行政命令，呼吁从事人工智能项目的公司制定保护该技术用户的指导方针。
At the latest meeting, officials from 16 different technology companies signed a voluntary agreement on Tuesday. In it, they promised to develop AI technology in ways that limit possible harm to the public. Among the companies were American-based Google, IBM, Amazon, Microsoft and OpenAI. Others included China's Alibaba and Tencent, as well as South Korea's Samsung Electronics and G42 from the United Arab Emirates.
在最近一次会议上，来自16家不同科技公司的官员周二签署了一项自愿协议。在这份文件中，他们承诺开发人工智能技术的过程中要限制可能对公众造成伤害。这些公司包括美国的谷歌、IBM、亚马逊、微软和OpenAI。其他包括中国的阿里巴巴和腾讯，以及韩国的三星电子和阿拉伯联合酋长国的G42。
The British government co-led the summit with South Korea. It said in a statement that the leaders in attendance sought to establish "a common understanding of AI safety and align their AI research."
英国政府与韩国共同主持了此次峰会。它在一份声明中表示，与会领导人试图建立“对人工智能安全的共识，并协调他们的人工智能研究”。
One of the main parts of the latest agreement calls for creating a network of publicly supported safety institutes designed to guide research and testing of AI technologies. Some of these institutes have already been set up by Britain, the U.S. and Singapore.
最新协议的主要内容之一是建立一个由公共支持的安全研究所组成的网络，旨在指导人工智能技术的研究和测试。英国、美国和新加坡已经建立了一些这样的机构。
United Nations Secretary-General Antonio Guterres told the meeting through a video link that AI technology has continued to quickly develop since the group's last meeting in Britain. Such development had led to "life-changing technological advances." But Guterres added that the progress had also created "life-threatening new risks."
联合国秘书长安东尼奥•古特雷斯通过视频连线在会议上表示，自该组织上次在英国召开会议以来，人工智能技术继续快速发展。这样的发展导致了“改变生活的技术进步”。但古特雷斯补充说，这一进展也造成了“危及生命的新风险”。
"We cannot sleepwalk into a dystopian future," he added. "Where the power of AI is controlled by a few people — or worse, by algorithms beyond human understanding," he said.
“我们不能梦游般地走向一个反乌托邦的未来，”他补充说。他说:“人工智能的力量被少数人控制，或者更糟的是，被人类无法理解的算法控制。”
Earlier this year, the U.N. General Assembly approved its first resolution on AI. The measure received support from all 193 U.N. member states. It called for international efforts to ensure that new technologies help all nations, respect human rights, and stay "safe, secure and trustworthy."
今年早些时候，联合国大会通过了第一项关于人工智能的决议。这项措施得到了联合国所有193个成员国的支持。它呼吁国际社会共同努力，确保新技术帮助所有国家，尊重人权，并保持“安全、可靠和值得信赖”。
Attendees agreed on the need for AI companies to exchange ideas on their progress and development methods. The group also agreed to provide information on safety.
与会者一致认为，人工智能公司有必要就其进展和发展方法交换意见。该组织还同意提供安全信息。
South Korea's science and information minister, Lee Jong-Ho, spoke to the gathering before it ended on Wednesday. He said worldwide AI development had reached a point where "cooperation is not an option; it is a necessity."
韩国科学与信息部长李钟镐(Lee Jong-Ho)在周三会议结束前发表了讲话。他表示，全球人工智能发展已经到了新的阶段：“合作不是一种选择;这是必须的。”
AI experts welcomed the steps to start creating rules, or regulations, to keep the technology safe. But some noted that such rules need to be enforced to be effective.
人工智能专家对开始制定规则或法规以确保技术安全的措施表示欢迎。但一些人指出，这些规定需要强制执行才能发挥效力。
Francine Bennett is director of the AI-centered Ada Lovelace Institute in Britain. She told Reuters news agency, "We need to move past voluntary...the people affected should be setting the rules via governments."
弗朗辛•贝内特(Francine Bennett)是英国以人工智能为中心的Ada Lovelace研究所的主任。她对路透社表示:“我们需要超越自愿……受影响的人应该通过政府制定规则。”
AI companies should be able to prove their products meet established regulations before they even reach the market, said Max Tegmark. He is president of the Future of Life Institute, a nonprofit bringing attention to the risks of AI. Tegmark said companies should equate safety with profit and attempt to limit possible harms during the development of systems.
马克斯•泰格马克说，人工智能公司应该能够在产品进入市场之前证明他们的产品符合既定法规。他是未来生命研究所（Future of Life Institute）的主席，该研究所是一家非营利组织，旨在关注人工智能的风险。泰格马克表示，公司应该将安全和利润并重，并要限制系统开发过程中可能造成的危害。
South Korean minister Lee said a big problem is that any laws aimed at making AI systems safer are far behind continuous development and public releases. He told Reuters, "But for safe use by the public, there needs to be flexible laws and regulations in place."
韩国部长李表示，一个重大问题是，任何旨在使人工智能系统更安全的法律都远远落后于人工智能的持续发展和公开发布。他告诉路透社，“但为了公众安全使用，需要有灵活的法律法规。”
I'm Bryan Lynn