Post #3922396
2026-06-10 10:06 UTC
@troed@swecyb.com Not to mention the catastrophic environmental consequences, energy consumption, the strain of online scraping, the shortages on RAM and other components, etc.
I've been using "small LLM models" that run locally on my PC. Even though their execution use far less resources than the big online models, their training did involve such big models in the first place (cf. model distillation), so they are not clean either.
But all of this is even not my first concern about LLMs.
Replies (1)
-
@zerkman@pouet.chapril.org 2026-06-10 10:16
@troed@swecyb.com My main concern is on the economic aspect. LLMs importance mostly come from the big tech industry that invested thousands of billion dollars in them, only to earn much less than 10% of that in revenue. There is no way they can achieve break-even any soon. Most of the effort to get some profitability is at the expense of even larger investments. What they are doing is building a bubble that might pop sooner or later. Whatever the way this crisis ends, the current situation will not last.