When talking about the energy consumed by LLMs don't be fooled by arguments focusing solely on the direct power consumption of these models, because they are externalizing a lot of it. Browsers are now doing proof-of-work calculations to access many websites, because websites need to protect themselves from AI scrapers. That takes power! Let that sink in: every computer, tablet or phone on earth is now consuming more power *every time it accesses a webpage* because of "AI".
Discussion
Loading...
@gabrielesvelto YES. This is having a daily impact on the useability of the ENTIRE world wide web, as we repeatedly have to prove we are human, or see sites go dark due to the constant DDOS attacks that webscraping LLM bots bring.
@gabrielesvelto @karl and most numbers floating around are about inference, not training (which is vastly more power hungry). and they never factor the carbon/energy footprint of manufacturing the hardware.
@gabrielesvelto don't forget the bandwidth. That all uses energy as well
@gabrielesvelto Hello 👋 how are you doing?, How would you describe your writing style?
@gabrielesvelto This is such a critical point. The indirect energy costs of AI are staggering. It's one reason I believe in local-first AI processing — tools that run in your browser use YOUR device's power, not burning datacenter energy for every keystroke. The cloud-everything model is unsustainable.
@gabrielesvelto Even earlier than that, there are the energy and resources consumed to build the chips and servers those LLMs run on.
@gabrielesvelto Good point. The pursuit of surveillance at ALL costs (and our expense!)
@gabrielesvelto scrapers are not merely a consequence of AI.