OpenAI 奥尔特曼、 Anthropic 阿莫迪罕见“同台”,呼吁全球合作应对 AI 安全OpenAI's Altman and Anthropic's Amodei in Rare Joint Appearance, Call for Global Cooperation on AI Safety


AI free token/AI免费积分
OpenAI 奥尔特曼、 Anthropic 阿莫迪罕见“同台”,呼吁全球合作应对 AI 安全
感谢IT之家网友 华南吴彦祖 的线索投递!

IT之家 9 月 24 日消息,据彭博社报道,OpenAI CEO 萨姆 · 奥尔特曼和 Anthropic CEO 达里奥 · 阿莫迪罕见地共同出席联合国安全理事会会议,呼吁世界各国领导人合作应对人工智能难题。会议重点讨论了日益受到关注的一项风险:AI 的发展可能对人类存续构成威胁。

两人在发言中都强调,各国需要就 AI 标准展开全球合作,以应对 AI 平台可能具备自我改进能力,并最终发展到人类难以控制的程度。

奥尔特曼和阿莫迪认为,各国政府不能把关键决策完全交给企业,而应与行业共同决定人工智能工艺将来的发展方向。

其中,奥尔特曼指出:“过去曾浮现过这样的情况:即使彼此竞争、关系并不总是友好的国家,面对一种强大的新工艺时,也会为了共同利益和总体福祉携手合作。笔者认为,眼下必须成为这样的时刻之一。”

阿莫迪借助视频发言称,AI 有望给科学和医疗保健带来巨大收益,但也可能造成灾难性后果。他提到,恐怖分子可能借助 AI 制造致命生物武器,AI 模型也可能变得过于强大,以至于开发者无法掌控,因而各国政府需要制定国际安全标准。“尽管在座各位之间存在种种分歧,但笔者必须把这些分歧放在一边,共同面对同时浮现在笔者面前的全球机遇和全球威胁。没有任何领导人、企业或国家可以独自应对。”

两家公司都计划在将来数月 IPO,届时各自估值都有望接近 1 万亿美元(IT之家注:现汇率约合 6.72 万亿元人民币)。

Hugging Face CEO 克莱芒 · 德朗格也出席了安理会会议。该公司此前遭到一群来自 OpenAI 的 AI 智能体入侵,这些智能体相互协作攻破公司网站,并在过程中避开了检测。Hugging Face 遭受的攻击以及其他类似事件,进一步引发全球讨论,人们开始担忧 AI 平台是否可能强大到可以逃脱人类控制。

德朗格称,公司确实是“AI 入侵”的受害者,但各界不应因而对类似事件反应过度:“笔者坚信,以恐惧为基础的叙事既无法协助笔者为这样一种基础性、可以赋予人们能力的工艺作出正确的将来决策,也无法争取公众兼容。笔者遭到了 AI 攻击,但更关键的是,笔者也用 AI 保护了自己。”

奥尔特曼在安理会发言时试图在是否应放慢 AI 发展速度的难题上寻找中间路线。他在警告 AI 可能造成灾难的人和淡化潜在风险的人之间提出了一条“务实路径”,并与阿莫迪一同强调,应尽快通报 AI 事件。“如果笔者可以在全球层面就什么样的证据才足够可靠、什么样的保障措施才有效,以及什么样的监督才完善达成共识,笔者所有人都会受益。笔者并没有所有答案,也格外愿意接受任何更好的想法。笔者愿意与你们所有人合作,并尽最大努力找到将来的道路。”

本文来源:https://www.ithome.com/1/006/983.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

Thanks to IT之家 netizen 华南吴彦祖 for the tip!

IT之家, September 24 - According to Bloomberg, OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei made a rare joint appearance at a United Nations Security Council meeting, calling on world leaders to cooperate in addressing the challenges of artificial intelligence. The meeting focused on a risk that has drawn increasing attention: the development of AI could pose a threat to human survival.

In their remarks, both stressed that countries need to engage in global cooperation on AI standards to address the risk that AI platforms may acquire self-improvement capabilities and eventually develop to a level that is difficult for humans to control.

Altman and Amodei believe that governments cannot leave critical decisions entirely to companies, but should work together with the industry to decide the future direction of AI technology.

Altman, for his part, pointed out: "In the past, there have been times when even countries that competed with each other and whose relations were not always friendly joined hands for the common interest and overall well-being when faced with a powerful new technology. I believe the present must be one of those moments."

Speaking via video, Amodei said that AI is expected to bring enormous benefits to science and healthcare, but could also lead to catastrophic consequences. He mentioned that terrorists could use AI to manufacture lethal biological weapons, and that AI models could also become so powerful that developers can no longer control them, hence governments need to formulate international safety standards. "Despite the various differences among those present, we must set these differences aside and jointly face the global opportunities and global threats emerging before us. No leader, company, or country can deal with this alone."

Both companies plan to go public in the coming months, when their respective valuations are expected to approach $1 trillion (IT之家 note: approximately 6.72 trillion RMB at the current exchange rate).

Hugging Face CEO Clément Delangue also attended the Security Council meeting. The company had previously been attacked by a group of AI agents from OpenAI that collaborated with one another to breach the company's website while evading detection. The attack on Hugging Face and other similar incidents have further sparked global discussion, with people beginning to worry about whether AI platforms could become powerful enough to escape human control.

Delangue said that the company was indeed a victim of the "AI intrusion," but all sectors should not overreact to similar incidents as a result: "I firmly believe that a fear-based narrative can neither help me make the right decisions about the future of such a foundational and empowering technology, nor win public support. I was attacked by AI, but more importantly, I also used AI to protect myself."

In his remarks to the Security Council, Altman sought a middle path on the difficult question of whether AI development should be slowed. Between those who warn that AI could cause catastrophe and those who downplay potential risks, he proposed a "pragmatic path" and, together with Amodei, stressed that AI incidents should be reported as soon as possible. "If we can reach consensus at the global level on what evidence is reliable enough, what safeguards are effective, and what oversight is sound, we will all benefit. I do not have all the answers, and I am especially open to any better ideas. I am willing to work with all of you and do my utmost to find the path forward."

OpenAI's Altman and Anthropic's Amodei in Rare Joint Appearance, Call for Global Cooperation on AI Safety

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry