I can understand that take, but to me the more relevant comparison is the fossil fuel boom.

Like the advent of more powerful creative tools (camera, printing press, etc.), fossil fuels allowed us to do what we were already doing but faster.

Unlike a camera, though… Coal, oil, and gen AI all have to pull raw material from somewhere in order to operate, and produce undesired byproducts as a result of their operation.

In the early days of fossil fuel, it must been impossible to even conceive the thought that there might be limits to how much we could safely extract raw materials or dump hazardous residual crud.

From one person’s perspective, the world seems so impossibly large. But it turns out, there are limits, and we were already well on our way to exceeding them by the time we realized our predicament.

I think we’re sprinting towards discovering similar constraints for our information systems.

It won’t be exactly the same, and much like climate change I don’t think there will be a specific minute of a specific day where everything turns to shit.

But I think there are instructive similarities:

• The most harmful kinds of gen AI, as with fossil fuels, will have the highest ROI.
• There will be safe, responsible ways to use it, but it will be difficult to regulate from the top down and full of perverse incentives to cheat from the bottom up.
• It will probably continue to accelerate even as the problems become more noticeable and disruptive.
• It will be next to impossible to undo the damage at a significant scale.