this post was submitted on 10 Mar 2026
46 points (100.0% liked)
Neo-Luddites
58 readers
12 users here now
founded 4 months ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
That is, to my understanding, not at all what is happening. LLMs are mainly getting better because they're being fed with more data and having more layers. This directly makes their environmental impact worse. They will also reach computational limits, hopefully sooner rather than later. So there is currently not much indication that generalist AI will keep getting better, nor that it will become less wasteful. Purpose built smaller models could be both, but they'd far more closely resemble any other software tool than LLMs, both in use and social impact.
AI got a lot better in recent years because of transformers and the multi-head attention mechanism. Some more major breakthroughs will be made with a bit more time that will greatly boost accuracy and efficiency. Technology progresses exponentially so advances will get more and more common. One example of how LLMs are currently getting more efficient is DeepSeek's improvement of sparse attention. There are trillions of dollars being thrown at these problems and they're racing as fast as they can because there's an arms race as well as heavy market competition.