男友太凶猛1v1高h,大地资源在线资源免费观看 ,人妻少妇精品视频二区,极度sm残忍bdsm变态

Global EditionASIA 中文雙語Fran?ais
Business
Home / Business / Finance

Special attention needed to ensure AI safety, US professor says

By Mike Gu in Kong Hong | chinadaily.com.cn | Updated: 2025-01-14 19:00
Share
Share - WeChat
US computer science professor Stuart Russell talks to the media at the 2025 Asia Financial Forum (AFF) in Hong Kong on Tuesday. MIKE GU / CHINA DAILY

Stuart Russell, a distinguished professor of computer science at the University of California, Berkeley, emphasized the need for special attention to the safety of artificial intelligence (AI) during its development, when participating in a group interview at the 2025 Asia Financial Forum (AFF) held in Hong Kong.

For safety reason, AI systems need to have behavioral red lines, Russell said. "The problem with general-purpose AI is that it can go wrong in so many ways that you can't easily write down what it means to be safe. What you can do is write down some things that you definitely don't want the systems to do. These are the behavioral red lines," he explained to the reporters why building behavioral red lines for AI is important.

"We definitely don't want AI systems to replicate themselves without permission. We definitely don't want them to break into other computer systems. We definitely don't want them to advise terrorists on how to build biological weapons," Russell said.

He added that it is hoped that AI development will always be under human control, rather than becoming uncontrollable.

This is why it is crucial to generate behavioral red lines at the early stages of AI development, especially with the help of governments, Russell said.

"So, you can make a list of things that you definitely don't want to do. It is quite reasonable for governments to say that before you can put a system out there, you need to show us that it's not going to do these things," he said.

Russell pointed out that AI gives rise to new forms of cybercrime. Currently, criminals are using AI to craft targeted emails by analyzing social media profiles and accessing personal emails, he said. This allows AI to generate messages that reference ongoing conversations, impersonating someone else, he added.

Russell, however, stated that AI also boosts the defense of crimes. "On the other side, we have AI defenses. I'm part of a team in various universities in California working together to use AI as a defense to screen emails against fishing attacks, to look at the activities of algorithms operating within the network, and to see which ones are possibly engaging in various activities", he said.

When asked about AI competition between countries, Russell said, "I think, in general, competition is healthy". However, he emphasized that excessive competition in AI should be approached with caution, as it could jeopardize AI safety. "Safety failures damage the entire industry. For example, if one airline doesn't pay enough attention to safety and airplanes start crashing, that damages the whole industry," he said.

AI cooperation, based on safety, is both allowable and economically sensible, Russell said. "In collaboration with several AI researchers from the West and China, we've been running a series of dialogues on AI Safety, specifically to encourage cooperation on safety. Those have been quite successful. The behavioral red lines I mentioned earlier are a result of those discussions," he said.

Regarding AI cooperation between China and the United States, Russell stated that both countries now place a strong emphasis on ensuring AI safety.

"I think there's at least as much interest in that direction in China as there is in the US. Several senior Chinese politicians have talked about AI safety and are aware of the risks to humanity from uncontrolled AI systems. So, I really hope that we can cooperate on this dimension," he said.

"The US and China have agreed not to allow AI to control the launch of nuclear weapons, which I think is sensible," he added.

mikegu@chinadailyhk.com

Top
BACK TO THE TOP
English
Copyright 1995 - . All rights reserved. The content (including but not limited to text, photo, multimedia information, etc) published in this site belongs to China Daily Information Co (CDIC). Without written authorization from CDIC, such content shall not be republished or used in any form. Note: Browsers with 1024*768 or higher resolution are suggested for this site.
License for publishing multimedia online 0108263

Registration Number: 130349
FOLLOW US
CLOSE
 
主站蜘蛛池模板: 黑河市| 濮阳县| 平山县| 泸定县| 岳池县| 剑川县| 遂川县| 平陆县| 独山县| 鱼台县| 徐汇区| 富民县| 文安县| 邵阳市| 安国市| 金堂县| 阳城县| 枣阳市| 布尔津县| 葫芦岛市| 宝兴县| 尼木县| 定襄县| 都江堰市| 额济纳旗| 犍为县| 清涧县| 大洼县| 舞钢市| 察雅县| 张家川| 丽水市| 贺州市| 滕州市| 泰兴市| 隆德县| 邵阳县| 石嘴山市| 蕉岭县| 静安区| 梁河县|