undefined | Better HN

0 pointsastromaniak1y ago0 comments

> At least in this case, they didn't call it "open source", which it isn't.

There are no open source AI models, strictly speaking. Because they are not reproducible even will all the data and training setup. Randomness is embedded in training. It could be done without, but performance suffers. So, only for learning and some academic research.

0 comments

38362936481y ago

So a program built with an non-determinisitc compiler cannot be open source? Or even just a C program that will change based on the order each object file is built? What about build time stamps?

You can absolutely open source an LLM, it's just that noone has. And you can absouletely bundle the random seem with the training data, it's just that, again, noone has.

And runtime randomness is hardly an issue. We write programs with RNG in them all the time and noone takes issue with that

j / k navigate · click thread line to collapse

0 pointsastromaniak1y ago0 comments

> At least in this case, they didn't call it "open source", which it isn't.

0 comments

38362936481y ago

So a program built with an non-determinisitc compiler cannot be open source? Or even just a C program that will change based on the order each object file is built? What about build time stamps?

You can absolutely open source an LLM, it's just that noone has. And you can absouletely bundle the random seem with the training data, it's just that, again, noone has.

And runtime randomness is hardly an issue. We write programs with RNG in them all the time and noone takes issue with that

j / k navigate · click thread line to collapse