Post #4747797 · Elektrine

2026-09-18 11:15 UTC

Before descending into my interpretation of the different aspects, I should put my cards on the table. They are not stone tablets, just the standards I intend to use: Preserving the ecological systems that sustain human life is my first priority. No promised capability makes their destruction acceptable. We must coûte que coûte avoid economic dependencies that leave us without real alternatives. Changes to society’s basic socioeconomic structure must be publicly and democratically moderated. Companies cannot control that process alone. Technology, including AI, should serve human beings, not the other way around. These are not hard scientific criteria. They are personal value judgments. I do not think that element can be removed from evaluation, however sternly one looks at the spreadsheet. Making those values explicit seems more honest than pretending they are absent. I will use them to examine resources, ownership, institutions, incentives, and effects. 8/34

Replies (1)

@masek@infosec.exchange 2026-09-18 11:15

By these standards, training is currently the part of AI I find most open to criticism. Not because training necessarily consumes most electricity over a model’s entire life. At massive scale, inference can consume more. My concern is that training decisions trigger the race for larger clusters, more power, more cooling, more chips, and more data before we know what the next increase in scale will actually buy us. The infrastructure bonanza is real. US data centers are projected to consume 9.5–15.3% of national electricity by 2030. The burden is not only carbon: it includes water, land, minerals, hardware production, e-waste, grid pressure, and costs concentrated in host communities. Training and anticipated inference both drive this buildout. One isolated training run is not my point. The problem is an expansion cycle in which the ecological commitment arrives first and proof of additional value is asked to catch up later. 9/34

Open ##4748054