Post by @flaki@flaki.social

RE: https://mastodon.social/@glyph/117226760121627412

Even if I plugged a mac mini directly into my solar charger and decided it can only work when the sun shines, and nothing else is using that power* that wouldn't suddenly make local inference "ethical and perfectly okay".

There are many reasons (see the blogpost), but to point out one: "open weight" models still got trained *somehow*. It's a "sunk cost" now, yes, but that doesn't absolve you of the responsibility to people/nature/etc. who paid the price.

I can hear you yell "yes yes buuut critical minerals! coffee!! chocolate, you double-standards dickpickle!!!". And yes, but we *acknowledge* the harms there and (at least some of us try to) do better. Fairphones exist, fair-trade cocoa & coffee beans exist. All the AI cabal ever done is dismiss/minimize the problems and thunder on…

___
* which is a thought that keeps knocking around in my braincase for the generalized "any computing hardware" case. Think SETI/FoldAtHome but using excess *solar* and not "idle compute" as a trigger

Glyph

@glyph@mastodon.social 12 hr. ago

Is using a local LLM any worse than playing a video game? No clickbait here, the answer is "yes, much worse", but perhaps you will be interested in an explanation of why: https://blog.glyph.im/2026/09/but-what-about-video-games.html

3 Replies

4 Boosts

Jonathan Schofield

@urlyman@mastodon.social 2 hr. ago

@flaki and the solar tech is a much slower manifestation of the same pattern. The pattern of using the one-time inheritance of a solar battery that was 50+ million years in the making to roll out efficiency improvements in using that one time inheritance, but still dependent upon it, and within an economy still intent on increasing throughput, facilitated by that efficiency gain

Flaki

@flaki@flaki.social 2 hr. ago

@urlyman of course this can be applied, recursively, indefinitely but I think there is a very important difference between "somewhere in the past, on a geological timescale harms were done that we, as descendants of the harm-doing are still fractionally responsible for", and "in this decade and in our lifetime harms are being done that are clearly in violation of present and future people's right to a livable planet" (and ofc I'm not talking about the ascended trillions who longtermists are so worried to ‘protect’)

Jonathan Schofield

@urlyman@mastodon.social 2 hr. ago

@flaki totally agree. Lowering the gradient of the most egregiously untethered behaviour is the task at hand

bonfire.cafe