AI 博客每日精选 — 2026-05-18

今日技术圈聚焦三大趋势：一是AI学术诚信建设加速，arXiv祭出一年禁令严惩AI生成垃圾内容，同时业界开始反思生成式AI的泡沫与局限；二是大模型应用进入深水区，OpenAI组织调整、Greg Brockman接管产品线，预示AI技术正从概念验证转向产品化落地；三是开源与闭源路线之争再起波澜，NHS放弃开源的决定引发广泛讨论，技术选型正回归商业现实而非盲目跟风。

来自 Karpathy 推荐的 92 个顶级技术博客，AI 精选 Top 10

🏆 今日必读

🥇 ArXiv将对提交AI生成垃圾内容的研究者实施一年禁令

ArXiv to Ban Researchers for a Year if They Submit AI Slop — daringfireball.net · 1 天前 · 🤖 AI / ML

ArXiv计算机科学部门主席Thomas Dietterich宣布，如果提交物包含AI生成的不当语言、抄袭内容、偏见内容、错误、虚假引用或误导性内容，作者将承担后果。新规实行一次违规即处罚的规则——研究者将被禁止在ArXiv发表一年，此后提交必须先在同行评审的出版物上发表。Dietterich表示，如果无法确认作者是否检查了LLM生成的结果，就无法信任论文中的任何内容。该规定现已明确处罚细则，允许申诉。

💡 为什么值得读: 对于使用AI辅助学术写作的研究者，这是了解主流预印本平台对AI生成内容态度的重要参考，直接影响未来投稿的合规策略。

🏷️ ArXiv, AI slop, academic publishing, research integrity

🥈 GDS weighs in on the NHS’s decision to retreat from Open Source

GDS weighs in on the NHS’s decision to retreat from Open Source — simonwillison.net · 6 小时前 · ⚙️ 工程

🏷️ Open Source, NHS, UK Government

🥉 ★ AI Is Technology, Not a Product

★ AI Is Technology, Not a Product — daringfireball.net · 1 天前 · 💡 观点 / 杂谈

It’s not even a feature. It’s just technology.

🏷️ AI, technology, product

📊 数据概览

分类分布

pie showData
    title "文章分类分布"
    "🤖 AI / ML" : 6
    "⚙️ 工程" : 2
    "💡 观点 / 杂谈" : 1
    "🛠 工具 / 开源" : 1

高频关键词

xychart-beta horizontal
    title "高频关键词"
    x-axis ["ai", "arxiv", "ai slop", "academic publishing", "research integrity", "open source", "nhs", "uk government", "technology", "product", "pretraining", "parallel training"]
    y-axis "出现次数" 0 --> 4
    bar [2, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]

ai                  │ ████████████████████ 2
arxiv               │ ██████████░░░░░░░░░░ 1
ai slop             │ ██████████░░░░░░░░░░ 1
academic publishing │ ██████████░░░░░░░░░░ 1
research integrity  │ ██████████░░░░░░░░░░ 1
open source         │ ██████████░░░░░░░░░░ 1
nhs                 │ ██████████░░░░░░░░░░ 1
uk government       │ ██████████░░░░░░░░░░ 1
technology          │ ██████████░░░░░░░░░░ 1
product             │ ██████████░░░░░░░░░░ 1

🏷️ 话题标签

ai(2) · arxiv(1) · ai slop(1) · academic publishing(1) · research integrity(1) · open source(1) · nhs(1) · uk government(1) · technology(1) · product(1) · pretraining(1) · parallel training(1) · distributed systems(1) · generative ai(1) · world models(1) · neurosymbolic ai(1) · ai scaling(1) · rlvr(1) · science(1) · verification(1)

🤖 AI / ML

1. ArXiv将对提交AI生成垃圾内容的研究者实施一年禁令

ArXiv to Ban Researchers for a Year if They Submit AI Slop — daringfireball.net · 1 天前 · ⭐ 25/30

ArXiv计算机科学部门主席Thomas Dietterich宣布，如果提交物包含AI生成的不当语言、抄袭内容、偏见内容、错误、虚假引用或误导性内容，作者将承担后果。新规实行一次违规即处罚的规则——研究者将被禁止在ArXiv发表一年，此后提交必须先在同行评审的出版物上发表。Dietterich表示，如果无法确认作者是否检查了LLM生成的结果，就无法信任论文中的任何内容。该规定现已明确处罚细则，允许申诉。

🏷️ ArXiv, AI slop, academic publishing, research integrity

2. Notes on pretraining parallelisms and failed training runs.

Notes on pretraining parallelisms and failed training runs. — dwarkesh.com · 1 天前 · ⭐ 24/30

Deeply researched interviews

🏷️ pretraining, parallel training, AI, distributed systems

3. The illusion of Generative AI, the insanity of massive bets on hyperscaling, and the case for world models and neurosymbolic AI

The illusion of Generative AI, the insanity of massive bets on hyperscaling, and the case for world models and neurosymbolic AI — garymarcus.substack.com · 14 小时前 · ⭐ 23/30

Three excellent new interviews

🏷️ generative AI, world models, neurosymbolic AI, AI scaling

4. RLVR might be disproportionately bad at science

RLVR might be disproportionately bad at science — dwarkesh.com · 1 天前 · ⭐ 23/30

the verification loop for theories can be on the order of decades and centuries, and even then we know today as the better theory can often actually make worse predictions

🏷️ RLVR, science, verification, reinforcement learning

5. How I use LLMs as a staff engineer in 2026

How I use LLMs as a staff engineer in 2026 — seangoedecke.com · 22 小时前 · ⭐ 22/30

🏷️ LLM, staff engineer, Copilot

6. Greg Brockman Officially Takes Control of Products at OpenAI, a Very Stable Well-Run Company

Greg Brockman Officially Takes Control of Products at OpenAI, a Very Stable Well-Run Company — daringfireball.net · 1 天前 · ⭐ 22/30

Maxwell Zeff, reporting for Wired (News+ link):

OpenAI told staff on Friday that it would reorganize the company as part of an ongoing effort to unify its product offerings, Wired has learned. Ope

🏷️ OpenAI, Greg Brockman, company reorganization, product strategy

⚙️ 工程

7. GDS weighs in on the NHS’s decision to retreat from Open Source

GDS weighs in on the NHS’s decision to retreat from Open Source — simonwillison.net · 6 小时前 · ⭐ 24/30

🏷️ Open Source, NHS, UK Government

8. SQLAlchemy 2 In Practice - Chapter 8: SQLAlchemy and the Web

SQLAlchemy 2 In Practice - Chapter 8: SQLAlchemy and the Web — miguelgrinberg.com · 1 天前 · ⭐ 23/30

🏷️ SQLAlchemy, SQL, web development, database

💡 观点 / 杂谈

9. ★ AI Is Technology, Not a Product

★ AI Is Technology, Not a Product — daringfireball.net · 1 天前 · ⭐ 24/30

It’s not even a feature. It’s just technology.

🏷️ AI, technology, product

🛠 工具 / 开源

10. Make ZIP files smaller with ZIP Shrinker

Make ZIP files smaller with ZIP Shrinker — evanhahn.com · 1 天前 · ⭐ 22/30

🏷️ ZIP Shrinker, compression, browser tool, APK

生成于 2026-05-18 22:18 | 扫描 88 源 → 获取 2532 篇 → 精选 10 篇 基于 Hacker News Popularity Contest 2025 RSS 源列表，由 Andrej Karpathy 推荐 由「懂点儿AI」制作，欢迎关注同名微信公众号获取更多 AI 实用技巧 💡