On 23/07/26 at 11:15 +0900, Simon Richter wrote: > And, speaking of sustainability, the most important point: our AI policy is > also a statement on whether we believe the efficiency[2] gains are worth the > resource usage. We cannot remain neutral on this topic. > > [...] > > Such a position would also include a statement that we believe the resource > usage in land, water, energy and material to be adequate for the value it > provides to free software and its users, including those in the areas where > new datacenters are being built. > > We cannot avoid making such a statement, because the resource usage is well > known at this point
Are you sure that the resource usage is that well known at this point? What I see is announcements mainly target at investors, with: * Confusion between general datacenter usage (not necessarily AI-related) and AI-related usage * Confusion between training and inference * Announcements about future datacenters that may or may not realize At the same time, there are still open questions about: * How much models can be optimized (example: Google claims that Gemini 3.6 Flash, released two days ago, reduces output token usage by 17% compared to Gemini 3.5 Flash, released two months ago) * Whether it's possible to select an optimal model for each task to minimize costs * What are the efficiency gains in the specific context of Debian development * If those efficiency gains allow us to write better software faster, what is the impact of this better software on the world at large? (if AI helps us make dpkg 20% more resources efficient, is it a net gain or a net loss for the world? what if if AI helps up provide a newer upstream version that is 20% more resources efficient?) Currently we are mostly gathering data points. This is very useful, but this is different from drawing a general conclusion. I fear that if Debian made a statement on resources usage vs efficiency of AI today, there's a high risk that we will feel a bit ridiculous about what we stated in 6 months. Lucas
