Should useful corrections to AI count as paid work?
A discussion on reddit.com explores whether detailed corrections provided to AI models should be considered paid work. The core argument is that a thorough correction, unlike a simple thumbs down, represents significant knowledge that may have taken years to acquire. If a company utilizes such corrections to train future AI models, the question arises whether the individual providing the correction should be compensated, or if the utility of the AI tool itself is sufficient recompense. This query is part of research into public views on AI training data.
- Published
- Sep 7, 2026, 05:44
- Source type
- Dev community
- Tier
- Community
- Source status
- Healthy
Times shown in UTC
More details
A thumbs down on a bad answer and a detailed correction from someone who knows the subject aren't the same contribution.
If a company uses that correction to train a future model, it's getting something that might have taken the person years to learn. Should that person get paid, or is using the tool enough in return?
Obviously paying for every message would invite a lot of junk. But treating all of it as ordinary feedback seems off too.
Where would you draw the line between using a product and doing useful work for the company behind it?
For context, I work on an AI project and am asking as part of research into people's views on training data.