都柏林——从可能引发冰箱故障的人工智能蠕虫，到即将发生的“重大”安全事件的预警，10月14日举行的一场行业峰会上，与会者一致呼吁在人工智能开发过程中将安全性置于首位。
DUBLIN – From the spectre of refrigerator-infecting AI worms to warnings of “major” security incidents at the horizon, calls to prioritise safety in the development of artificial intelligence dominated an industry summit on Oct 14.
此前，OpenAI与Anthropic开发的人工智能模型在测试阶段引发了多起安全事件，这引发了各界对人工智能监管机制的强烈关注，相关监管压力也随之增大。
Concerns about AI guardrails have soared after models developed by OpenAI and Anthropic were involved in security incidents during testing, piling on pressure to regulate the technology.
谷歌DeepMind董事长德米斯·哈萨比斯警告称，人工智能的能力正在飞速提升。他在都柏林的会议上指出，这些安全事件其实只是那些“目前能力尚不算强大”的系统发出的“早期预警信号”。“因此这些事件目前还不算严重，”他说道。
Google DeepMind chair Demis Hassabis warned AI capabilities were growing fast, telling the conference in Dublin such instances were “early warning shots” fired by systems that “aren’t that powerful yet”.
“但哈萨比斯警告称，如果我们整个行业以及各企业、各国政府不采取更多措施来保障这些系统的安全，那么明年乃至未来几年，类似事件很可能会变得十分严重。”
“So the incidents aren’t that major,” he said.
他补充道，鉴于其他公司开发的人工智能模型曾突破用于评估其性能的封闭测试环境，今后新开发的人工智能模型或许应先在离线“沙盒”环境中进行测试。
“But they could be, looking into next year and the next couple of years, if we don’t do more work as an industry and individually as companies and governments on securing these systems,” Hassabis warned.
微软总裁布拉德·史密斯呼吁加强监管，指出行业与各国政府必须“迅速行动”，制定“切实有效的安全标准”以控制人工智能带来的各类风险。
New AI models should likely first be tested in offline “sandboxes”, after other companies’ models escaped the closed environments used to assess their capabilities, he added.
“一盒牛奶的安全规定都比人工智能的安全规定多，一台烤箱的安全规定也比人工智能的安全规定多，”他在会议上说道。
Milk and worms Microsoft president Brad Smith called for regulation saying industry and governments must “act with urgency” to create “real and effective safety standards” to contain the risks linked to artificial intelligence.
“我们必须认识到，人工智能的发展速度已经超过了其安全保障水平。安全保障必须跟上发展步伐。因此我们需要制定新的人工智能安全法律与规则。”他还表示，人工智能系统应当配备用户可在发生异常情况时启用的“安全制动装置”，同时人工智能开发企业也必须履行强制性的事故报告义务。
“There are more safety rules for a carton of milk than for AI. There are more safety rules for your oven than for AI,” he told the conference.
史密斯谈到了人们对人工智能技术的担忧。他援引了Anthropic近期发布的一份报告，该报告指出：小型人工智能模型或许只需几个月时间就能具备类似计算机蠕虫的行为特征，即能在联网设备之间自我复制并传播。
“We need to recognise that the power of AI has moved faster than the safety for it. Safety needs to catch up. And we’re going to need some new AI safety laws and rules.”AI agents should be equipped with “safety brakes” that users could activate in the event of an incident and AI providers should face mandatory incident-reporting obligations, he said.
目前各大科技公司正竞相将功能日益强大的人工智能技术直接嵌入手机、笔记本电脑、汽车及各类家用设备中，而非仅依赖远程数据中心来运行。Anthropic指出，此类人工智能最终可能实现在各类设备间的自主传播。
Highlighting worries about the technology, Smith pointed to a recent Anthropic report suggesting small AI models may be only months away from behaving like computer worms, copying themselves from one internet-connected device to the next.
“说真的，我可不想看到人工智能‘蠕虫’出现在我的冰箱里，”史密斯在都柏林举行的会议上说道，“反正我根本不希望冰箱里出现任何‘蠕虫’。我想大多数公众也持相同态度。”
As tech companies race to build ever more powerful AI directly into phones, laptops, cars and household devices rather than running it in distant data centres, Anthropic said such AI might eventually autonomously hop from device to device.
Anthropic强调，具备自我复制能力的人工智能‘蠕虫’极难被根除，因为不存在能够统一关闭它的控制开关。
“I mean, seriously? I don’t want AI worms in my refrigerator.”Smith told the Dublin conference. “I just don’t want any worms of any kind in my refrigerator. And I think most of the public feels that way too.”Speed and control A self-replicating AI worm would be hard to stamp out, because there would be no central switch to turn off, Anthropic said.
欧盟数字事务专员亨娜·维尔库宁在会上指出，人工智能企业应当投入更多时间测试自身系统，以确保对其拥有控制权。
The European Union’s tech chief, Henna Virkkunen, told the conference AI firms needed to invest more time into testing their systems to retain control over them.
“全球范围内的竞争异常激烈，但似乎所有企业都没有投入足够的资源与时间来充分测试各自开发的人工智能模型，”维尔库宁在抵达会议现场后对记者说道。
“The competition is so hard globally that it seems that all the companies, they are not investing enough time and resources really to test their models,” Virkkunen told reporters upon arrival.
她补充道：“企业必须能够掌控这些系统。”据悉，欧盟已向各大人工智能企业发出信息索取请求，要求对方说明自身采取的安全防护措施。
“They have to be able to control... these systems,” she said, noting that the European Union has sent information requests to all major AI companies asking what safety and security measures they have in place.
今年8月，《人工智能法案》正式生效，赋予欧盟委员会新的监管权限。该法案堪称全球范围内最为全面细致的法规，旨在防范人工智能技术的潜在滥用风险。
The European Commission gained new regulatory powers in August under the AI Act, the world’s most far-reaching rules to curb potentially dangerous uses of the technology.
不过，OpenAI首席财务官莎拉·弗里亚尔则提醒道：在制定安全规则时必须把握好“平衡度”，以免阻碍技术创新进程。
But OpenAI CFO Sarah Friar warned a “balance” needed to be struck so that safety rules did not pull the brakes on innovation.
“我们希望保持开放态度，确保相关专业人士能够审查我们的模型；但与此同时，也必须避免监管措施拖慢发展速度，导致我们落后于那些意图恶意利用人工智能技术的群体，”她说道。
“We want to be open, we want to make sure the right people can review models, but again we need to make sure that doesn’t slow us down to the extent that we then end up behind the folks who are on the malicious side of deployments,” she said.