一系列由身处 AI 技术研发一线的人士出镜的视频，重复了外界"末日派"近来发出的部分警告。
"在我看来，人类灭绝的概率大约就是一枚硬币的正反面。"前 OpenAI 与 Google DeepMind 员工 Geoffrey Irving 在一场新访谈中这样说。这是十几场 AI 研究者访谈中的一场，受访者包括 OpenAI、Google 与 Anthropic 的现任和前任员工。自称致力于研究 AI 能力与动机的非营利组织 Palisade Research 收集并在 frominside.ai 上线了这些访谈。
多位研究者与 Irving 一起警告 AI 可能导致人类灭绝。Google DeepMind 研究科学家 Neel Nanda 表示，"至少有 10% 的概率它会造成人类灭绝，这个数值高得离谱。"我们近来做关于"AI 安全"共同体的报道时曾引用过 Nanda 与 Irving 的表态，也提到这个词的含义究竟指什么、以及人们该为此做些什么，都极难界定——即便是这批视频，也没能给出一个统一的解决方案。
前 OpenAI 研究者 Daniel Kokotajlo 说得更直接：超级智能 AI"基本上会有神一般的力量"，他补充道，"不幸的是我们完全不知道该如何控制它们，所以很可能谁也控制不了。这件事听起来有多危险，实际上就有多危险，绝不能让它发生。"
每段视频都可完整观看，该网站同时把不同人谈及相似话题的片段做了聚合整理，包括"这一切是否只是炒作或营销"，以及"为什么明知这么危险还要去做"。
本文来源：https://www.theverge.com/ai-artificial-intelligence/1002238/openai-google-anthropic-ai-researchers-safety-interviews
若有版权问题，请来信告知：13951431913#139.com 或在右方提交评论。
A series of videos featuring people on the front line of AI technology research and development echoed some of the warnings recently issued by outside "doomers".
"In my view, the probability of human extinction is about a coin flip," former OpenAI and Google DeepMind employee Geoffrey Irving said in a new interview. This was one of more than a dozen interviews with AI researchers, including current and former employees of OpenAI, Google and Anthropic. Palisade Research, a nonprofit that describes itself as dedicated to researching AI capabilities and motivations, collected and posted these interviews on frominside.ai.
Several researchers have joined Irving in warning that AI could lead to human extinction. Google DeepMind research scientist Neel Nanda said, "There is at least a 10% probability that it will cause human extinction, and that number is absurdly high." In our recent coverage of the "AI safety" community, we had quoted statements by Nanda and Irving and noted that it is extremely difficult to define what the term really refers to or what people should do about it—even these videos did not provide a unified solution.
Former OpenAI researcher Daniel Kokotajlo put it more directly, saying that superintelligent AI "will basically have godlike powers," and he added, "Unfortunately, we have absolutely no idea how to control them, so very likely no one will be able to control them. This is as dangerous as it sounds, and it must absolutely not be allowed to happen."
Each video can be watched in full, and the site also aggregates clips in which different people discuss similar topics, including "whether all this is just hype or marketing" and "why people do it despite knowing it is so dangerous."