Nathan Baschez

Nathan Baschez

x.com/nbaschez

Founder of the Lex writing app, now at Notion, who is broadly optimistic about AI, especially in education, while expecting some harms along the way.

AI将如何改变世界?

文明层面的变革渐进式变化DoomBloom
模拟位置解读范围

横向:他们表达的 Doom–Bloom 前景看法。 纵向:变革程度。

Doom–Bloom:100 中的 80。变革程度:100 中的 75。解读范围:横向为 75 至 85,纵向为 70 至 80。这些是解读坐标,而不是事件概率。

Nathan Baschez的 P(doom) · 推断

≈4%

0%100%

根据他们的模拟回答推断,并非他们给出的数字。 合理范围:2–7%。

I expect serious harms along the way, but I put extinction risk very low.

我们如何估计 P(doom)

Nathan Baschez 的里程碑时间线
  1. 超人类 AI

    A major reason is that progress toward superintelligence seems more likely to unfold over months than in one invisible instant.

    回答 1
  2. 工作与机构

    But “enormous” doesn’t necessarily mean “instant” or “humans disappear.” I expect the crucial transition to unfold over months, leaving some room for warning incidents and societal intervention.

    回答 2

按里程碑分组,不按推断日期间隔或排序。AGI 和超人类 AI 保留他们的定义。

他们的展望取决于什么

一个核心假设

Because the transition to superintelligence is more likely to unfold over months than instantaneously—and because frontier-scale compute is relatively concentrated—I think society would have opportunities to intervene.
回答 3

如果这个假设实际并非如此,他们的展望会如何变化?

更多详情

预期益处

预计将带来具有变革性且广泛有价值的收益。

89 / 100

影响小变革性影响

在定性尺度上,解读范围为 67 到 100。

预期危害

仍有几种解读是合理的:严重或广泛的危害预计将是未来不可忽视的一部分。 / 预计会出现可控或局部的危害。

50 / 100

影响小变革性影响

在定性尺度上,解读范围为 33 到 67。

人类影响力

人类的选择具有实质性但受到很大制约的影响。

54 / 100

影响力小影响力强

在定性尺度上,解读范围为 50 到 75。

预期能力

预计AI仍将是能力有限的工具。

预计AI将在大多数认知工作中达到人类水平。

模拟位置:预计AI将在认知工作中大幅超越人类。

发展速度

停止或大幅放缓开发能力更强的AI。

模拟位置:在落实所述保障措施的前提下继续开发。

加快开发能力更强的AI。

AI使用规则

在事先落实保护措施或获得许可之前,限制所讨论的AI用途。

模拟位置:允许所讨论的AI用途,同时实施有针对性的问责与保护措施。

尽量减少对所讨论AI用途的限制。

这些解读保留了他们陈述的条件。益处和危害都可能很大。这些范围描述的是我们如何解读他们的模拟回答,而不是统计置信区间。

与Nathan Baschez相比,你的立场在哪里?
用大约3分钟描绘你自己的AI世界观,然后进行比较

相似的世界观

模拟世界观与 Nathan Baschez 最接近的意见领袖

来源

用于为此模拟用户提供事实依据的文章、访谈和著述。

模拟评估

问题 1

你认为AI对我们的未来意味着什么?为什么?

I’m broadly optimistic about AI, especially as a medium for people and agents to think together. In writing, education, and other knowledge work, the interesting future isn’t simply “AI replaces the person.” It’s new shared environments where people can develop ideas, get useful feedback, and extend their abilities without surrendering judgment. I expect serious harms along the way, but I put extinction risk very low. A major reason is that progress toward superintelligence seems more likely to unfold over months than in one invisible instant. Bad incidents should create warning signals and provoke a forceful social response. Frontier systems also depend on concentrated compute, which gives society more leverage than it would have over a dangerous technology distributed in millions of tiny, uncontrollable objects. The weakest part of that argument is international coordination: one country slowing down does not guarantee that others will. So my optimism is conditional on institutions reacting before catastrophe, and on products being designed around productive human involvement. One practical question is whether agents can be given both the tools and incentives to escalate uncertainty or danger to human overseers rather than bluffing or pressing ahead. I’m excited by the possibilities, but I also want my children to inherit a recognizably human world and family life—not merely one optimized for whatever the technology can do.

问题 2

你认为AI最终会在多大程度上改变世界?

Enormously—probably at the level of our basic institutions and how people learn, write, work, and make decisions. If systems become far more capable than humans across most cognitive tasks, the change won’t be confined to a few products; it will reshape the economy and the texture of everyday life. But “enormous” doesn’t necessarily mean “instant” or “humans disappear.” I expect the crucial transition to unfold over months, leaving some room for warning incidents and societal intervention. I’m especially interested in a future where people and agents think together: shared environments that strengthen human judgment rather than merely automate it away. The harder question is how much continuity we preserve. I want the benefits—particularly in education and creative work—while keeping a recognizably human world, including familiar family life. Whether we manage that depends heavily on product design, institutions, and international coordination, with the last of those being the assumption I trust least.

问题 3

凭直觉,你认为AI导致人类灭绝或类似的永久性灾难的概率有多大?

My gut-level assessment is very low, but I wouldn’t attach a precise percentage to it. I expect serious smaller harms and warning incidents before anything extinction-level. Because the transition to superintelligence is more likely to unfold over months than instantaneously—and because frontier-scale compute is relatively concentrated—I think society would have opportunities to intervene. The most fragile part of that view is international coordination. A warning may cause some governments or companies to slow down without producing an adequate global response. So “very low” isn’t “impossible”; it reflects my expectation that visible harm will trigger strong action before a permanent catastrophe.

问题 4

在一个美好的AI未来,人们应该能够拒绝哪些类型的变化?

People should be able to refuse changes that erase meaningful human agency or make ordinary life unrecognizable. That includes being forced to delegate judgment to agents, having education become passive machine consumption, or losing access to spaces where humans write, learn, decide, and relate to one another directly. I don’t think a good AI future requires everyone to adopt the maximally automated lifestyle. People should be able to choose tools that support their thinking rather than substitute for it, and agents should escalate important uncertainty to humans instead of quietly making irreversible decisions. Institutions should preserve genuine human oversight, not merely put a ceremonial approval button after an automated process. I’m also protective of continuity in family and community life. My children should be able to inherit a recognizably human world and eventually pass that kind of life on themselves. That doesn’t mean freezing culture or rejecting powerful tools. It means technological capability alone shouldn’t dictate which forms of life must disappear.
你的立场在哪里?
回答几个简单问题,探索你自己的AI世界观。
描绘你自己的世界观

你的立场在哪里?

描绘我的世界观