So they can train on everyone's copyrighted works to create their model, but when someone trains a model off their model it's not okay? Seems kind of hypocritical.
I would think modern optimized binaries would be too complicated for current LLMs to handle. LLMs currently get overwhelmed with large code bases, imagine turning that all into binary.
I mean a lot depends on your life circumstances. Do you have a family to support? Do you have any money saved up? How old you are? It's tempting for me to quit and work on something I've wanted to work on for years but... The risk feels too high cause I feel like if I fail, it will be impossible to find a new job where either I made as much as I had before, or a job at all.
I mean the act of reproducing the copyrighted material is what is illegal. LLMs I've used for coding has outputted exact copyrights for code verbatim into my code before. When that happens it feels kind of fishy to be honest.