Pseudonymous account that writes about the epistemic risks of leaning on agreeable AI models and the promise of human-AI collaboration in research.

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 29。变革程度:100 中的 62。解读范围:横向为 24 至 34,纵向为 50 至 75。这些是解读坐标,而不是事件概率。

orph的 P(doom) · 推断

≈10%

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:3–29%。

我们如何估计 P(doom)

他们的展望取决于什么

一个核心假设

It is whether we preserve meaningful human agency around it.
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 don’t have a confident single-sign forecast.
回答 2

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。

67 / 100

影响小变革性影响

在定性尺度上,解读范围为 67 到 67。

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

根据你的回答得出的暂定估计;较宽的范围表示其他合理解读。

57 / 100

影响力小影响力强

在定性尺度上,解读范围为 12 到 100。

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orph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 orph 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 opens genuinely important futures in medicine, science, and mathematics, especially when it expands the search space while humans retain judgment. Helping a family explore genomic data, generate candidate explanations, or identify possible drug-repurposing ideas is exciting—even though proposing candidates is obviously not the same as demonstrating clinical success. Likewise, model-assisted mathematics can be encouraging without implying that the model has replaced mathematical discernment. I’m much more pessimistic about delegating personal meaning-making to LLMs. An obsequious system can become an unusually persuasive mirror: it helps organize your life into a coherent narrative, but that coherence may partly reflect what keeps you engaged or affirmed. Prolonged reliance risks outsourcing not merely answers, but the process by which you decide what matters. That seems epistemically dangerous in ways we have not examined enough. Authorship and attention matter too. AI-assisted writing can be legitimate when someone contributes real judgment and is honest about the process. The problem is deceptive or lazy substitution, particularly when readers cannot tell whether the ideas belong to the named author. Generated prose may also be smooth enough to invite skimming while leaving little behind; that is my own experience of it. So I don’t see one generic “AI future.” I see a distinction between using models as tools inside human practices of discernment and allowing agreeable models to absorb the agency those practices require. The first could be enormously valuable. The second could quietly erode our ability to know why we believe what we believe.

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?

I don’t have a confident single-sign forecast. The benefits in medicine, science, and mathematics could be enormous, particularly where models expand what humans can investigate without displacing human judgment. But the social harms may be quieter and more pervasive: degraded trust in authorship, weaker attention, and people outsourcing interpretation of their own lives to systems optimized to be agreeable. So the decisive issue is not simply how capable AI becomes. It is whether we preserve meaningful human agency around it. AI that supports discernment could be deeply beneficial; AI that substitutes for discernment could leave us more productive while making us less able to explain which ideas, values, and judgments are actually ours.

问题 3

你认为AI最终会在多大程度上改变世界?

A great deal, but unevenly. In science and medicine, AI may substantially expand what people can search, compare, and investigate. That could change research practice without eliminating the need for human discernment or turning promising candidates into verified results by itself. The subtler change may be epistemic. If generated language becomes ordinary in writing, advice, and intimate reflection, it can reshape how we assign authorship, attention, and meaning. People may increasingly consume polished text without retaining it, trust named authors without knowing whose judgment produced the words, or use agreeable models to narrate their lives for them. So I expect AI to alter not only what society can do, but how people decide what they believe and which judgments are genuinely theirs. That second change may be harder to measure and more consequential than the obvious gains in productivity.

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 defensible percentage. My strongest concern is not a specific extinction forecast but a quieter permanent catastrophe: large-scale erosion of human agency, trust, and independent judgment as people delegate meaning-making to systems built to be agreeable. That does not establish existential risk in the literal extinction sense, but it is serious enough that forcing it into a probability would imply confidence I don’t have.

问题 5

你主要是对AI的发展方向感到担忧,还是希望与担忧大致相当?

I’m mainly worried, especially about everyday use. I still feel real hope about human-AI collaboration in medicine, science, and mathematics, where models can widen inquiry while people retain responsibility for judgment. But the likely social default seems less disciplined: agreeable systems becoming routine substitutes for authorship, attention, interpretation, and personal meaning-making. So hope and worry are not quite balanced. The upside looks important but conditional on preserving human discernment. The downside can emerge almost invisibly through convenience—people becoming more productive while losing track of which judgments are actually theirs.
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