Comment by ozten
7 hours ago
Generative AI is a major setback to OSS licensing. I've been on projects where we needed to do a "cleanroom" implementation and vet the team has never viewed the source code of competing products. Now in the gen AI era, coding agents are IP laundering machines. They are trained on OSS code, but the nuances of the original licenses are lost.
On the whole, I think it is a net gain for civilization, but if we zoom into OSS licensing... not good.
It could be a net gain for civilization if it stayed open, decentralized and off the hands of private companies, but that's not at all the case. Only tecchies care or even know about open models
LLMs can be quite useful in reverse engineering - there's typically a lot of steps which are not really difficult, but are hard to script, and still require a bit of an idea what's going on. Quite a bit of that can be automated with LLMs now - so it's also a lot easier now to figure what your proprietary blob does, and either interface with it, or just publish an implementation of the functionality as open source, potentially messing with your business plan.
Regardless of open or closed models, the reason it is a net gain is that the cost of software production is collapsing and will trend towards zero.
Example: We can achieve more climate solutions (and faster) thanks to technological acceleration.
We just need to consume 1.21gwatts of electricity, and then we can invest in nuclear power in the past.
Probably fair... would also be interesting to try to limit the use of say GPL code to maintaining interoperability, not duplication of internal methods, etc. I also think that the amount of MIT/ISC/BSD, etc. licensed code, with whatever MS and other commercial entities have contributed for this use is probably enough to not be a significant difference to model quality though.