Replying to @⁨drmoose@lemmy.world⁩

Sure not all the critique is accurate, but in its current form, it causes huge issues with water and noise pollution in places the data centers are built, and are sucking unbelievable amounts of power, while essentially DDOSing tons of sites to scrape data, and often using copyrighted material that they have no license for. It is also reducing people’s ability to think critically because people turn to it immediately.

I’m not an idiot who will say it’s not useful, but we need to slow down by about 20x with what we’re doing with it now, and save it mainly for medical usage until we have a solution to those massive problems.

Replying to @⁨starelfsc2@sh.itjust.works⁩

The water and noise problem functionally has zero to do with llms as a tool. That is entirely and solely a byproduct over over expansion of corporate intrests.

That’s like saying the lumber mill is the reason logging is harmful to the environment.

It’s nonsensical and requests objective ignorance of the context and nuance of the problem

There’s nothing that actually requires the data center problem to exist for llms to also exist. There are alternative.

Replying to @⁨Holytimes@sh.itjust.works⁩

We already have a perfectly good policy framework for datacenters: just charge them the negative externality costs. The problem is that they want datacenters now and repay external costs like electricity, water etc later because it takes 1 year to build a datacenter but 5 years to route power and water.

We already solve for this with economic credits but people just don’t trust these systems because of corruption and poor governance. The credits never trickle down to people who pay for externalities.

This is not even a datacenter issue.

en