Adam Elmore

Adam Elmore

x.com/adamdotdev

Software developer and podcast co-host who finds AI agents powerful for routine coding but values hands-on programming and sustainable work habits.

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 48。变革程度:100 中的 40。解读范围:横向为 43 至 53,纵向为 0 至 76。这些是解读坐标,而不是事件概率。

Adam Elmore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:1–11%。

我们如何估计 P(doom)

他们的展望取决于什么

一个核心假设

AI can make implementation cheap while leaving judgment, taste, responsibility, and meaningful goals just as difficult.
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 think the change will be substantial, but I can’t responsibly put a precise ceiling on 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。

66 / 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

预计会出现可控或局部的危害。

42 / 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

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。

61 / 100

影响力小影响力强

在定性尺度上,解读范围为 24 到 100。

AI使用规则

在事先落实保护措施或获得许可之前,限制所讨论的AI用途。

模拟位置:允许所讨论的AI用途,同时实施有针对性的问责与保护措施。

尽量减少对所讨论AI用途的限制。

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Adam Elmore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 Adam Elmore 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 will make a lot of work dramatically faster, especially the mundane parts of software development. Models can edit files, propose changes, and compress hours of implementation into a much shorter loop. I’ve seen enough striking successes to believe that shift is real. But I’ve also seen agents get stuck, repeat themselves, and produce activity that looks like progress without actually solving the problem. The harder question is what that does to the experience of making things. If every change gets routed through an agent, I may ship more while understanding less of the system. Programming can turn from building into supervising a stream of plausible changes, and some of the satisfaction and craft disappears. Cheap generation also doesn’t make quality cheap: choosing a worthwhile goal, exercising taste, and doing careful work still require effort. So I’m neither dismissive nor uncomplicatedly excited. I want tools that remove drudgery while leaving engineers in control and in contact with their code. We also need to treat agents as real operational actors: broad access and exposed credentials can create expensive failures. And I don’t think keeping agents busy around the clock is automatically productive. A future worth wanting should give us more time and attention, not turn us into compulsive agent managers who sacrifice sleep and offline life to keep the machines producing.

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 society-wide forecast. In software, I expect a mixed impact: substantial productivity gains and less mundane work, paired with weaker understanding, more low-quality output, operational risks, and pressure to treat constant agent activity as productivity. The benefits are real, but they don’t automatically produce better work or better lives. AI can make implementation cheap while leaving judgment, taste, responsibility, and meaningful goals just as difficult. It can also tempt people to optimize for volume—more code, more posts, more agents running—rather than value. So the overall impact depends heavily on how we use these tools. I want AI to expand what people can make while preserving control, craft, attention, and time away from work. If it mostly creates faster output and more compulsive supervision, that won’t feel like straightforward progress.

问题 3

你认为AI最终会在多大程度上改变世界?

I expect AI to change how a huge amount of digital work gets done. In software, the shift already feels deeper than a better autocomplete: implementation can move from hours of manual coding to shorter cycles of directing agents, reviewing files, and correcting failures. That changes team workflows, the pace of production, and potentially who can build useful things. But “more output” isn’t the same as a transformed or improved world. Judgment, taste, responsibility, and choosing worthwhile goals remain difficult. Models can generate plausible code or content cheaply while also creating failure loops, security risks, and mountains of mediocre material. They may also leave people less connected to the systems they maintain. So I think the change will be substantial, but I can’t responsibly put a precise ceiling on it. The important question is whether we use the leverage to remove drudgery and reclaim time, or simply create an always-on culture of supervising more agents and producing more stuff.

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. The risks I can speak to are much more concrete and near-term: agents with excessive access, leaked credentials, costly unauthorized use, unreliable changes, and humans losing understanding while trusting plausible output. Those can cause serious harm, but they don’t give me a sound basis for estimating human extinction or an equivalently permanent catastrophe.
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