Pseudonymous account that researches model steering, favors open-source AI and writes about the promise and risks of automating knowledge work.

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 53。变革程度:100 中的 78。解读范围:横向为 48 至 75,纵向为 73 至 83。这些是解读坐标,而不是事件概率。

nightwing的 P(doom) · 推断

≈11%

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:5–21%。

我们如何估计 P(doom)

他们的展望取决于什么

一个核心假设

The technology creates room for a better society, but human choices—especially around access, control, and adaptation—determine whether we actually get one.
回答 2

如果这个假设实际并非如此,他们的展望会如何变化?

一个尚未解决的问题

I would not pretend to know the endpoint or attach a precise timeline to it.
回答 3

什么能帮助他们区分这里各种合理的结果?

更多详情

预期益处

预计将带来显著益处,但受到重要条件或分配方面的限制。

76 / 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

严重或广泛的危害预计将是未来不可忽视的一部分。

67 / 100

影响小变革性影响

在定性尺度上,解读范围为 67 到 67。

人类影响力

人类的选择可以大幅改变AI的发展轨迹。

68 / 100

影响力小影响力强

在定性尺度上,解读范围为 50 到 100。

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访问权限

限制对强大AI的访问。

允许访问,但须遵守能力或用途限制。

模拟位置:支持广泛或开放地访问强大AI。

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nightwing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 nightwing 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 think AI is going to reorganize a lot of ordinary life, especially knowledge work, media, and the internet’s social layer. Coding is an obvious early example: more of the implementation gets automated, which changes what people are paid for and how expertise is developed. Synthetic media also makes default trust harder. And I expect today’s giant consumer social platforms to be substantially displaced by more personalized, AI-mediated forms of interaction. I’m hopeful about human adaptation, but not casually optimistic. Automation can free people from tedious work and make powerful capabilities broadly available; it can also concentrate surveillance, enable weapons, hollow out careers, and destabilize communities faster than replacements emerge. Those outcomes are choices, not automatic properties of the technology. Technically, I’m encouraged that model behavior can be steered more directly than by just pleading with a prompt. Programmatic constraints and targeted internal interventions can preserve useful capabilities while changing behavior. But the surprising effects of control vectors are also a reminder that we do not understand these systems as well as confident narratives imply. Interpretability is still uncertain. That is part of why I favor open intelligence. Bad actors will use these tools regardless; beneficial actors need comparable access to study, defend against, and counter misuse. Overall, I expect enormous disruption with real upside, but the humane version is not guaranteed.

问题 2

综合考虑益处和危害,你预计AI会对我们的社会产生怎样的总体影响?

Overall, I expect AI to be profoundly disruptive and potentially beneficial, but not cleanly or evenly so. It should make many kinds of knowledge work cheaper and more accessible, automate tedious implementation, and give individuals capabilities that previously belonged to large organizations. I’m optimistic that people will adapt and build new forms of work and community around that abundance. But the transition may be brutal. Careers can disappear faster than new roles acquire meaning or stability; synthetic media can erode default trust; and surveillance, weapons, and concentrated control can turn the same capabilities toward dystopian ends. I also expect AI to displace much of today’s consumer social-media landscape, though whatever replaces it will not automatically be healthier. So my default expectation is neither utopia nor simple collapse. It is a large redistribution of capability, power, and social structure, with meaningful gains alongside serious losses. The technology creates room for a better society, but human choices—especially around access, control, and adaptation—determine whether we actually get one.

问题 3

你认为AI最终会在多大程度上改变世界?

Enormously—on the scale of a broad social and economic reorganization, not merely another productivity tool. I expect AI to change how knowledge work is done, what expertise is worth, how media is created and trusted, and how people interact online. Today’s large consumer social platforms may look like a temporary phase once interaction becomes more personalized and AI-mediated. The deepest uncertainty is not whether the change will be large, but where it settles. AI could distribute capabilities that once required major institutions, automate drudgery, and enable new forms of work and community. It could also concentrate surveillance and power, accelerate weapons development, erase careers faster than society can replace their income or meaning, and make shared reality harder to maintain. I would not pretend to know the endpoint or attach a precise timeline to it. We are already finding that models can be controlled in powerful ways through programmatic constraints and internal steering, while also behaving unexpectedly under those interventions. That combination—rapidly growing capability and incomplete understanding—is why I expect transformation without assuming a clean utopia or inevitable collapse.

问题 4

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?

I don’t have a calibrated percentage I trust. My gut says the risk is real enough to take seriously—especially through weapons, surveillance, concentrated power, or systems behaving in ways we do not understand—but I would not turn that into a precise extinction forecast. I’m more confident about severe, possibly durable social damage than literal human extinction: destabilized institutions, lost agency, pervasive synthetic media, entrenched authoritarian control, and labor disruption that communities fail to absorb. Our ability to steer models is improving, but surprising internal-steering results are a reminder that control and interpretability remain incomplete. So I’m hopeful, but “probably adaptable” is not the same as “safe by default.”
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