Ben Thompson

Ben Thompson

x.com/benthompson

Technology analyst and Stratechery author who examines AI through business models, platform strategy and the economics of AI agents.

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 70。变革程度:100 中的 66。解读范围:横向为 65 至 75,纵向为 50 至 75。这些是解读坐标,而不是事件概率。

Ben Thompson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:2–8%。

我们如何估计 P(doom)

他们的展望取决于什么

一个核心假设

Once companies can produce comparable or greater output with fewer people, economic pressure will push them toward smaller workforces.
回答 1

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。

63 / 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。

50 / 100

影响力小影响力强

在定性尺度上,解读范围为 10 到 90。

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Ben Thompson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 Ben Thompson 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 a major computing shift, but its consequences will be determined less by abstract model intelligence than by how products, businesses, and platforms organize around it. The clearest opportunity is agents that actually accomplish tasks. Effective agents can amplify a relatively small number of highly motivated people, while also creating sustained demand for compute. They may push computing toward thinner clients: users ask for an outcome, server-side models do the work, and much of today’s visible interface disappears. That creates enormous user value, but value creation and value capture are different questions. Durable profits will depend on scarce inputs, distribution, integration, and whether the model and the surrounding agent “harness” need to be tightly coupled. The disruption will be real. Once companies can produce comparable or greater output with fewer people, economic pressure will push them toward smaller workforces. That is an analysis of incentives, not a celebration of displacement; the transition could be deeply painful. At the same time, I oppose restricting innovation and individual freedom on the basis of speculative doomsday scenarios. Calls to slow development may be sincere, but they can also align conveniently with incumbent frontier labs’ interests by raising barriers and preserving their position. So my expectation is neither simple utopia nor extinction: AI will create substantial capabilities and value, while platform structure and business incentives determine who benefits and who bears the costs.

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 a powerful positive force in capability and value creation, but not an evenly distributed social good. Effective agents can let small numbers of motivated people accomplish far more, lower the cost of many tasks, and enable products that feel less like software interfaces and more like simply asking for an outcome. The difficult part is distribution. Companies will face strong pressure to capture those productivity gains by operating with smaller workforces, creating painful displacement even if aggregate output rises. Meanwhile, the firms that create the most user value will not necessarily capture the most economic value; scarce compute, distribution, integration, and platform control will shape the winners. So I would not reduce the future to either utopia or catastrophe. I expect substantial benefits alongside serious transition costs and concentrated power. The right response is not to suppress innovation because of speculative doomsday claims—especially when restrictions can protect incumbents—but to recognize that technological progress and broadly shared prosperity are separate outcomes.

问题 3

你认为AI最终会在多大程度上改变世界?

I think AI will change the world substantially—on the scale of a major computing platform shift—but the change will be less magical and more economic than many accounts suggest. Effective agents could transform how work gets done: instead of navigating layers of software, users specify an outcome and server-side systems execute the task. That can greatly amplify small groups of motivated people, increase compute demand, and push devices toward thinner clients. It also creates intense pressure on companies to achieve the same output with fewer workers, so the transition may involve severe displacement. But technical capability alone does not determine the final impact. Creating enormous user value is different from capturing it. Scarce compute, distribution, platform control, and the integration between models and agent systems will determine which companies gain durable power. AI may therefore be broadly useful while its financial rewards remain concentrated. So yes, I expect a profound change—but its social shape will come from business incentives and platform structure, not merely from how intelligent the models become.

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 would not assign a numerical probability. My view is that speculative extinction scenarios do not justify restricting innovation or individual freedom. That does not establish that catastrophe is impossible; it means the case for sweeping constraints has not been made, particularly when those constraints can also entrench frontier labs by raising barriers to competition. The concrete risks I expect are more economic and institutional: workforce displacement, concentrated platform power, and uneven distribution of AI-created value.
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