[Hayvok]

> I think we need better regulation of these companies to prevent them from doing actual damage rather than trusting them to self-regulate against hypothetical ones.

What is the actual damage you are concerned about and determined to regulate? Simply saying "environment, workers, and the economy" is so broad I can't imagine what an effective regulation would look like. How would you even word the regulation?

[agentultra]

First, I can't possibly have all of the answers.

However the evidence that these companies are doing real damage now is all around us.

I already gave the example of Microsoft using billions of litres of fresh water to cool data centres during a drought.

In the case of labour the SAG-AFTRA strike, a contributing factor was the use of AI in the industry.

There are some estimates that the carbon emissions of these training efforts dwarf the airline industry and will grow to consume, like crypto, more energy than small countries soon [0].

Not sure that we need to be protected against hypothetical, super-intelligent, self-aware AGI systems that are, if even possible, decades away when people are using what we have today to lay off labourers by training models on their work and replacing them.

[0] https://www.ll.mit.edu/news/ai-models-are-devouring-energy-t...

Update: removed unnecessary bit

[Hayvok]

In your water example, perhaps that liter of water used to cool the datacenter is offering software to a hospital that offers life saving treatments. How will you measure the trade-off of a liter of water used one way vs. another?

Likewise, how can we distinguish between a ton of carbon emitted in the datacenter vs. a ton of carbon emitted by an airplane? Again, you might train an AI and emit one ton of carbon and that AI a save a million lives. Contrast that ton of carbon emitted with any number of frivolous airline flights by rich talking heads.

It may sound like I'm being deliberately difficult/obtuse, but this is exactly why regulation is so difficult to do well, especially in such a rapidly innovating space.

[agentultra]

We don't really need to argue hypotheticals to make progress. We know that airlines move people around and someone taking a trip on that airplane might be a doctor who could end up being at the right place at the right time to save the world!

That doesn't mean we give up and don't regulate the airline industry.

We can put caps on how much water data centres are allowed to consume in a given period in order to protect vital ecosystems and ensure enough fresh water for other uses.

We can write labour laws that protect workers from employers training models on their employees' work and then laying them off.

There's a lot we can do that isn't being done, "because innovation."