Bojan Tunguz

Bojan Tunguz

x.com/tunguz

Data scientist and TabulAI founder who calls himself adjacent to e/acc, expects very fast AI progress and worries about who shares its benefits.

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:他表达的 Doom–Bloom 前景看法。 纵向:变革程度。

Doom–Bloom:100 中的 51。变革程度:100 中的 75。解读范围:横向为 25 至 75,纵向为 70 至 80。这些是解读坐标,而不是事件概率。

Bojan Tunguz的 P(doom) · 推断

≈7%

0%100%

根据他的模拟回答推断,并非他们给出的数字。 合理范围:2–19%。

我们如何估计 P(doom)

Bojan Tunguz 的里程碑时间线
  1. 工作与机构

    The technical change could be enormous and arrive much faster than institutions can absorb.

    回答 3

按里程碑分组,不按推断日期间隔或排序。AGI 和超人类 AI 保留他的定义。

他的展望取决于什么

一个核心假设

Right now the gains follow an extreme power law: a small group in tech captures enormous value, while most people experience AI as vague hype or as software they’re being forced to train so it can replace parts of their job.
回答 1

如果这个假设实际并非如此,他的展望会如何变化?

一个尚未解决的问题

I don’t have a defensible percentage.
回答 4

什么能帮助他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

严重或广泛的危害预计将是未来不可忽视的一部分。

66 / 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

人类的选择具有实质性但受到很大制约的影响。

51 / 100

影响力小影响力强

在定性尺度上,解读范围为 23 到 77。

发展速度

停止或大幅放缓开发能力更强的AI。

模拟位置:在落实所述保障措施的前提下继续开发。

加快开发能力更强的AI。

AI访问权限

限制对强大AI的访问。

允许访问,但须遵守能力或用途限制。

模拟位置:支持广泛或开放地访问强大AI。

这些解读保留了他陈述的条件。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的模拟回答,而不是统计置信区间。

与Bojan Tunguz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 Bojan Tunguz 最接近的意见领袖

Bojan Tunguz关于AI说过的话

Tunguz calls himself adjacent to accelerationists, says some pacing may be justified and worries that the benefits of AI are going to a few people.

  1. “I think we are all to some extent accelerationists. The question is just how far.”

    New York Post interview
  2. “I do think that some cap might be necessary at some point and some pacing might be justifiable.”

    New York Post interview
  3. “A few people, primarily in tech, are reaping enormous windfalls, while for the vast majority of the population AI seems extremely nebulous, at best.”

    Post on X
  4. “They are using their workers to train their AI and replace them. Utterly cringe and dystopian.”

    Post on X
  5. “NIMBYism will destroy America's march towards AGI.”

    Post on X

逐字引自所链接的出处,核对于 2026年10月6日

来源

用于为此模拟用户提供事实依据的文章、访谈和著述。

Adjacent to e/acc, not acceleration at any cost

New York Post article (syndicated on Yahoo) on accelerationists versus “doomers”. In his own quoted words he calls himself “adjacent” to e/acc, rejects acceleration at any cost, says some cap might be necessary at some point and some pacing might be justifiable, and says “we are all to some extent accelerationists. The question is just how far.” The reporter paraphrases him saying today’s tools let him build things impossible six months earlier and that he welcomes open models as a fallback if big labs restrict access, while rejecting the idea that cornering the market is their motive. Verdon’s and others’ statements excluded. Full article inspected.

yahoo.com
arXiv fighting the last war

Argues arXiv’s year-long bans for careless AI use are wrong: AI-generated content is a real challenge, but AI is already doing serious scientific work and its role will only grow, and the scientific paper is an outdated artifact. The bans hurt outsiders and early-career researchers; science should move to open, post-publication review like software’s git and pull requests. Full essay inspected.

bojan.substack.com
The end of the road for the Sora app

Explains OpenAI shutting down the Sora app as a result of an overheating race among the top labs: since December the newest models made a qualitative step up in coding, coding is now AI’s killer feature, and vibe coding is good enough that most software professionals can rely on it for most of their work. Industry analysis, not a societal forecast. Full essay inspected.

bojan.substack.com
Nightmares on the AI Doom Street

Older context. Describes mixed excitement and trepidation about AI, says both utopia and destruction of all life seem outlandish but nobody can be sure, and calls for credible, industry-wide AI risk assessment. Argues that many prominent AI doomers are technologically illiterate and that major AI policy should not rest primarily on people who have never trained a model, while saying everyone deserves some say. Opening section of the essay inspected; reader comments excluded.

bojan.substack.com

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?

AI means extremely rapid progress, especially in coding, mathematics, and science—but a potentially ugly transition. I can build things today that were impossible six months ago. Coding is already the killer app, and intellectually deep work may soon become a cheap service rather than a scarce human skill. The problem is distribution. Right now the gains follow an extreme power law: a small group in tech captures enormous value, while most people experience AI as vague hype or as software they’re being forced to train so it can replace parts of their job. That’s why blanket accelerationism doesn’t convince me. I’m adjacent to e/acc, but acceleration at any cost is not an absolute good; some pacing may eventually be justified. At the same time, panicked restrictions can be self-defeating. America could squander its lead by blocking infrastructure, driving away foreign customers, or letting a handful of labs and cities become an isolated echo chamber. I’d rather see open models, broader access, and scientific institutions adapt to AI than fight the last war.

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 enormous gains in capability paired with a deeply uneven and destabilizing transition. AI will make software, mathematics, and scientific work radically cheaper and faster. That is real progress, not marketing. But society won’t automatically share the upside. The default trajectory is an extreme power law: a few firms and technical insiders capture huge windfalls while many workers see a nebulous product, tighter surveillance, or a machine they were asked to train as their replacement. So I’m optimistic about what AI can do and much less optimistic about how institutions will distribute its benefits. Acceleration at any cost is not a serious social contract.

问题 3

你认为AI最终会在多大程度上改变世界?

Profoundly. If coding, mathematics, and serious scientific work become cheap services, AI won’t merely improve existing industries; it will reshape what expertise is worth, how quickly ideas become products, and who holds economic power. The technical change could be enormous and arrive much faster than institutions can absorb. The open question isn’t whether the world changes, but who benefits. Without deliberate correction, I expect the gains to remain brutally concentrated while workers and ordinary users absorb much of the disruption. So: transformative at the level of capability, potentially dystopian at the level of transition.

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 risk deserves credible, technically informed assessment, and it’s part of why acceleration at any cost is foolish. But I’m not going to turn uncertainty into a made-up P(doom) for rhetorical effect.

问题 5

你倾向于认为AI带来的益处多于危害,还是确实难以取舍?

I lean toward more good in terms of raw capability, but I’m genuinely worried about the default transition. Cheaper software, mathematics, and science are enormous benefits. Yet those gains currently accrue to a tiny group while many workers get surveillance, displacement, and the privilege of training their replacements. So I’m not torn about whether the technology is useful—it obviously is. I’m torn about whether our institutions can distribute the upside before concentrated power and social disruption dominate the story.
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