There Is No Posttraining Data Commons — and That Changes Who Controls AI
Pretraining enjoyed a vast public commons of web pages, code, and papers, but posttraining consumes tasks, not documents — and no such commons exists. Even software, the easiest case, yields tasks only through costly manufacturing: SWE-bench kept 2,294 of 90,000 pull requests, and most later audits found widespread defects. The author argues the open community must systematically inscribe human work into RL tasks before a few companies monopolize the ability to direct AI.
Pretraining consumes documents, while posttraining consumes tasks.
- david_shi
What gets me excited about this project is its legibility for philanthropy, like how the Internet Archive is the destination for random recordings found in Grandma's attic.
Having a trusted steward to handle the anonymization and eval design process could kickstart a world where retiring tax accountants, or even active and generous ones, are willing to donate their data to the posttraining commons at a scale that's actually much larger than the "data industry".
- dlwh
Hi, I wrote this a few weeks ago. Didn't expect to see this here, etc., etc.
Happy to answer any questions. We have a discord https://marin.community/
- mrkn1
A very academic way of saying We have all the data in the world, except the data we actually need. You ask a company how something works and they say:
"Oh, theres a process."
Great. Where is it documented?
"in a spreadsheet."
Which spreadsheet?
"Well, theres the old spreadsheet, new spreadsheet, and the spreadsheet that Karen maintains."
Where's Karen?
"She left in 2021."