undefined | Better HN

0 pointsSOLAR_FIELDS1y ago0 comments

Fair. In reading our thread again I realized we are probably talking past each other. Really this discussion should be about the philosophical aspect of what constitutes open source in the context of LLM’s. Which is a tough and more interesting topic.

Really the conversation should be reframed to be something along the lines of “is it even ethical for these companies to offer their LLM as anything other than open source”? The answer, if you look into what they do, is “probably not”. Arguing about the technicalities of whether they follow the letter of whatever rule or regulation is probably a waste of time. Because it is completely obvious to anyone who understands how this works that these models are built and sold off the backs of centuries of open source work not licensed or intended to be used for profit.

0 comments

lolinder1y ago

> Really the conversation should be reframed to be something along the lines of “is it even ethical for these companies to offer their LLM as anything other than open source”? The answer, if you look into what they do, is “probably not”.

Agreed, but I'm personally of the opinion that this is true for all intellectual endeavors. Intellectual property is the great sin of our generation, and I hope we eventually learn better.

And I think you've hit at the heart of the matter: the push for open source training data has never been about the definition of open source, it's always been a cover for complaints about where the data was sourced from. Which is also why we won't see it any time soon—not until the lawsuits wind their way through the courts, and even then only if the results are favorable towards training.

9dev1y ago

It was a joy to follow along this discussion. Thank both of you.

j / k navigate · click thread line to collapse