惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

L
LangChain Blog
博客园 - 司徒正美
美团技术团队
Martin Fowler
Martin Fowler
雷峰网
雷峰网
aimingoo的专栏
aimingoo的专栏
博客园 - 三生石上(FineUI控件)
Vercel News
Vercel News
酷 壳 – CoolShell
酷 壳 – CoolShell
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
爱范儿
爱范儿
U
Unit 42
Y
Y Combinator Blog
月光博客
月光博客
Hugging Face - Blog
Hugging Face - Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
有赞技术团队
有赞技术团队
GbyAI
GbyAI
H
Help Net Security
量子位
Last Week in AI
Last Week in AI
博客园_首页
腾讯CDC
小众软件
小众软件

Hacker News: Ask HN

The New Window Delete ChatGPT Atlas Spyware Tell HN: Qwen Free Tier Is Discontinued Ask HN: SeedLegals Partnerships in London, worth it? Ask HN: How to highlight talent from untraditional backgrounds? Ask HN: We dont need a programming language now? Durable Object alarm loop: $34k in 8 days, zero users, no platform warning What if Time at the subatomic level has multiple arrows? How to add MidnightBSD Key to UEFI Secure Boot DBX? (Revoked and Forbidden Keys) Ask HN: What's your experience working at xAI as an AI tutor? Any engineers here with experience of clinical data standards? Ask HN: Who is using OpenClaw? Agent Skills for Software Test Automation Ask HN: Who needs contributors? Claude Code is thinking too much Ask HN: What Is the Big-O Order of a Jigsaw Puzzle? Ask HN: Stepping into a new role as a Senior, mentoring dos and dont's? Founder from Zurich heading to SF and Austin for the first time Hacker News No Manual Screenshots: I Built a Scalable Screenshot API Using Cloud Playwright Ask HN: Thought experiment: AGI giving us answers we don't like? Ask HN: I quit my job over weaponized robots to start my own venture 1% Vacancy, 81% Preleased: Where Midmarket Compute Deploys in 2026 Ask HN: Preferred pricing model for sound effects libraries? Copy of the email I sent to my undergraduate professors on Nov 30, 2025 Model API Performance | Hacker News Ask HN: Are open-weight LLMs the new offline encyclopedias? Valgrind 3.27 RC1 is out Claude Code OAuth down for >12 hours Ask HN: What's Better?–Tauri or Electron?
ASK HN: AI models are built on all of us, should their we...
rhuber · 2026-06-18 · via Hacker News: Ask HN

Every frontier model is a compression of humanity's collective output. The training corpus is our books, our code, our forum arguments at 2am, our Wikipedia edits, our Stack Overflow answers, the photographs we posted, the songs we wrote about. No lab created this knowledge. They distilled what we already made, together, over centuries.

We already have a legal framework for exactly this situation: the patent. Society's bargain with inventors is simple. You get a temporary, exclusive right to profit from your contribution, and in exchange the knowledge eventually becomes public domain so everyone can build on it. The monopoly is the incentive; the expiry is the price. We decided long ago that no one should own an idea forever, even one they genuinely originated.

A trained model is a stronger case for this bargain, not a weaker one. The inventor of a drug at least synthesized a novel molecule. A model is overwhelmingly a derivative work, its capabilities being our capabilities, re-encoded. Yet current practice grants the labs a perpetual, closed monopoly over that re-encoding, with none of the public-domain backstop we demand of every other field.

So here's the proposal: a model's weights must be published openly after a fixed term, say one year from release.

One year is an eternity in this field. It's more than enough to capture the competitive advantage of being first, to recoup the enormous training cost, to build the products and the brand on top of a frontier model. Labs would still race to lead, because leading for twelve months is hugely valuable. But the moat would be temporal, not permanent, and that's the whole point. A time-limited exclusivity rewards genuine innovation without letting a handful of companies fence off the commons indefinitely.

The downstream effects are exactly what we'd want. Yesterday's frontier becomes today's public infrastructure. Researchers get real artifacts to study for safety and bias instead of black boxes. Startups and the open-source community build on a steadily advancing public baseline. Smaller players and the public sector stop being permanently a generation behind. And the labs keep their incentive to push forward, because the only way to stay ahead is to keep inventing, not to sit on a model trained on everyone's work and rent it back to them forever.

Patents got the trade-off right: protect the inventor long enough to reward the work, then return the knowledge to the people it ultimately came from. Models, trained on all of us, deserve the same deal, on a clock that matches how fast this field actually moves.