3 ms·
The thing they are pointing at and which is the thing people want is the output of the training engine, not the inputs. This is like someone saying they have an
by saurik 2y ago
The thing they are pointing at and which is the thing people want is the output of the training engine, not the inputs. This is like someone saying they have an open source kernel, but they only release a compiler and a binary... the kernel code is never released, but the kernel is the only reason anyone even wants the compiler. (For avoidance of anyone being somehow confused: the training code is a compiler which takes training data and outputs model weights.)
- _heimdall 2y agoThe output of the training engine, I.E. the model itself, isn't source code at all though. The best approximation would be considering it obfuscated code, and even then it's a stretch since it is more similar to compressed data. It sounds like Meta doesn't share source for the training logic. That would be necessary for it to really be open source, you need to be able to recreate and modify the codebase but that has nothing to do with the training data or the trained model.
- saurik 2y agoI didn't claim the output is source code, any more than the kernel is. Are you sure you don't simply agree with me?