@[email protected]Once you realize that a
#LLM is just a lossy compression of its textual sources (improperly named "training data"), you see how using a
#CodingAssistant provided as a service is just a way to share your own source code with your competitors.
Just like your competitors, you will see its contextualized output and will say "how smart!" while paying to work as a data annotator for
#Anthropic,
#OpenAI,
#Google and so on, providing new code and feedbacks they will compress in their matrices.
At the end of the day, companies using these
#AI tools totally deserve the outcomes as they will lose any edge over the market, with all newcomers producing cheap clones of their products.
Otoh, the
#BigTech stealing their code don't deserve it and will leverage their position to squeeze the market.
Sometime I wonder: maybe we should find ways to decompress (and thus exfiltrate) proprietary software into the commons as long as gready managers still push these tools on programmers. After all they are stealing from the
#FreeSoftware commons, violating
#copyleft and misattributing works under permissive licenses alike.
However it would legitimize such companies, and this would harm the
#commons and society much more than a few niche software clone could ever benefit them.
@[email protected]