Michael Thiessen

Michael Thiessen

x.com/michaelthiessen

Software educator who writes about practical workflows for coding with AI agents and builds AI tutoring that explains rather than hands over answers.

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 56。变革程度:100 中的 42。解读范围:横向为 50 至 61,纵向为 6 至 94。这些是解读坐标,而不是事件概率。

Michael Thiessen的 P(doom) · 推断

≈3%

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:低于 13%。

我们如何估计 P(doom)

他们的展望取决于什么

一个核心假设

A great deal depends on whether we design systems that keep humans appropriately involved and make capabilities understandable and controllable.
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

So I expect AI to reshape many tools and practices, but the ultimate scale and direction of its global impact remain open.
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。

61 / 100

影响小变革性影响

在定性尺度上,解读范围为 33 到 67。

预期危害

预计会出现可控或局部的危害。

30 / 100

影响小变革性影响

在定性尺度上,解读范围为 33 到 33。

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。

56 / 100

影响力小影响力强

在定性尺度上,解读范围为 26 到 99。

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Michael Thiessen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 Michael Thiessen 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 expect AI to become less like one universal assistant and more like a collection of specialized capabilities. Different models may handle decisions, coding, explanation, or other tasks differently, which could make them more useful—but also harder to choose and configure. Model selection and reasoning settings already impose a real usability cost. For software work, I think the interface around an agent matters as much as raw model capability. Giving an agent explicit, plausible next actions—for example, CLI output that suggests the next command—can make its work more reliable. Whether that saves tokens is a separate, still-open question. Likewise, imperfect benchmarks can retain comparative value; flaws do not automatically erase all signal. I also do not think the future is simply “delegate everything.” Excessive delegation can reduce effectiveness, while greater human involvement can improve productivity. In education, the useful goal is not an answer machine. AI tutoring should guide learners with hints and explain why something works, preserving understanding rather than bypassing it. So I see a future shaped by better specialization, better interfaces, and deliberate choices about where humans stay involved—not just by increasingly capable models.

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 overall societal forecast. The demonstrated benefits I care about—specialized tools, better coding workflows, and tutoring that supports understanding—coexist with real costs such as usability complexity and ineffective over-delegation. Those observations do not, by themselves, justify calling AI’s total impact positive or negative. A great deal depends on whether we design systems that keep humans appropriately involved and make capabilities understandable and controllable.

问题 3

你认为AI最终会在多大程度上改变世界?

I expect substantial change in particular domains, especially software work and education, but I would not attach a confident magnitude to the world as a whole. The important change may come less from one universally capable assistant and more from specialized models embedded in workflows. In coding, that could mean agents operating through interfaces that expose clear next actions, with humans staying involved where judgment and understanding matter. In education, it could mean tutors that give hints and explanations rather than simply producing answers. Those changes could be meaningful without implying that every task will be automated or that delegation is always beneficial. More capability also creates more choices—models, reasoning settings, workflows—and therefore more complexity. So I expect AI to reshape many tools and practices, but the ultimate scale and direction of its global impact remain open.

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 observations about coding agents, tutoring, model specialization, and workflow design do not establish a credible extinction-risk estimate. I can distinguish practical problems like over-delegation and usability complexity from catastrophic-risk claims, but I would not turn that limited evidence into a numerical forecast.
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