Julia Galef

Julia Galef

x.com/juliagalef

Author of “The Scout Mindset” who writes about reasoning well and changing one’s mind, and has explored why people disagree about advanced AI.

AI将如何改变世界?

文明层面的变革渐进式变化DoomBloom
未确定范围的中心解读范围

横向:他们表达的 Doom–Bloom 前景看法。 纵向:变革程度。

Doom–Bloom:100 中的 50。变革程度:100 中的 50。解读范围:横向为 45 至 55,纵向为 0 至 100。这些是解读坐标,而不是事件概率。

Julia Galef的 P(doom)

尚未估计

他们的模拟回答中关于灾难性风险的信息不足,无法进行估计。

我们如何估计 P(doom)

他们的展望取决于什么

一个尚未解决的问题

I don’t have a confident overall forecast.
回答 1

什么能帮助他们区分这里各种合理的结果?

更多详情

人类影响力

你尚未对此确定立场。该点标示的是未确定范围的中心,而非温和立场。

50 / 100

影响力小影响力强

在定性尺度上,解读范围为 0 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Julia Galef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

在其表达的少数立场上,与 Julia Galef 的模拟世界观最接近的意见领袖

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 don’t have a confident overall forecast. The key questions are still how difficult AGI is, how much recent progress should update our expectations, and what models drive disagreement about the consequences of systems that surpass human capabilities. Before jumping to optimism, doom, or a policy conclusion, I’d want to identify the real cruxes. Are people disagreeing about scaling, bottlenecks in reasoning, the feasibility of controlling highly capable systems, or the institutions governing deployment? What observations would change each side’s mind? “AI means progress” and “AI means catastrophe” can both function more like team slogans than forecasts unless we unpack the mechanisms behind them. So my strongest view is methodological: we should treat confidence as something to earn through investigation, calibration, and serious engagement with disagreement. The future could depend enormously on AI, but that importance makes careful reasoning more necessary, not less.

问题 2

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 numerical estimate. Giving a gut percentage would suggest more calibration than I can justify. I’d first want to resolve—or at least better understand—the main cruxes: how likely broadly superhuman systems are, what capabilities imply about control, and whether institutions can respond before failures become irreversible. The magnitude of the possible harm makes the question important, but it doesn’t by itself determine the probability.

问题 3

如果有的话,你预计AI会造成哪些重大危害?

I don’t have a well-supported forecast of which major harms AI will actually cause. Plausible concerns range from misuse and institutional failures to loss of control over systems that become broadly more capable than humans, but listing scenarios is not the same as assigning them high probability. I’d want to separate several questions: What capabilities are likely to emerge? Who will have access to them? Which failures are reversible, and which could become permanent? How much warning would we get, and could institutions respond in time? The most useful debate would identify where people’s models diverge and what evidence would update them, rather than treating either reassurance or catastrophe as the default conclusion.

问题 4

如果有的话,你预计人们会从AI中获得哪些重大益处?

AI could plausibly accelerate scientific discovery, improve decision-making, and make valuable expertise more widely available. But those are potential benefits, not a confident forecast about what people will actually receive. I’d want to distinguish capability from distribution and outcomes. A system may become excellent at research, education, or medicine without its gains being broadly shared—or without institutions using it well. The key questions are which tasks AI can reliably perform, how humans and organizations adapt, who controls access, and what evidence would show that apparent progress is translating into durable human welfare. I’m optimistic about the range of possible benefits, but I don’t think that alone justifies a confident overall prediction.
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