@aks @f4grx @thomholwerda @kde

The tool doesn't matter, the output does.

Handling LLMs is such a spectrum, from vibe coding without reading or understanding anything to being 100% identical to what the author would have written by hand. Banning the latter would be like banning certain text editors.

We have been getting human written slop for years! We deal with this the same way.

@d_ed

Yes, but unlike text editors, LLMs and the companies behind them also:
- assist in waging war using the same tech
- drive up costs for marginalized and often poor communities by building lots of data centers
- throw out all their climate goals to pursue AI
- lie constantly about the capability of their tools to hype them up, causing lots of damages
- I could keep going...

point is, unlike a text-editor, LLMs have real implications to their use that reach beyond the software.

@aks @thomholwerda @kde @f4grx
@d_ed @aks @f4grx @thomholwerda @kde It's a fundamental mistake to imagine a product as a static thing completely separate from the process that produced it. And the tool shapes the hands of the person who uses it.

@d_ed @aks @f4grx @thomholwerda @kde

The tool doesn't matter, the output does.

this feels dangerously close to the "let's stay apolitical and just focus on the code" argument that some people make to justify not banning bigoted and otherwise toxic contributors from foss communities

@aylamz @aks @f4grx @thomholwerda @kde and if the output mattered you wouldn’t be using that tool ^^
@d_ed @aks @thomholwerda @kde the tool 100% matters in case of llms!
@d_ed Can claude be selfhosted? I think one restriction could be that #SAS models are forbidden.