Okay, after *years* of over-hype about AI, does the DeepSeek moment and the deflation of the AI bubble mean we can finally be normal about AI and treat it like any other tech, evaluating it on its pros and cons, and its actual utility? I wrote about how we might actually do that in our day jobs: https://www.fastly.com/blog/can-we-be-normal-about-ai-now-that-deepseek-happened
@johnzajac I'm not sure this is really true? Training can be relatively power-intensive, though this varies a lot. Usage is increasingly not that power-intensive, especially when efficient caching is in place, and overall the average consumer's impact is roughly in line with other activities that we don't generally critique, like video encoding or high-res video streaming. I agree with you about consent on content, but again... same issue with streaming content on (e.g.) YouTube.
I think absent unlimited clean energy, "AI" is probably always going to stink like the planet-killer it is.
I suppose it's possible some unforeseen utility is worth the energy expenditure, social and economic disruption, and widespread larceny needed to "train" it, but I'm highly skeptical.
But it seems clear that "being normal" about energy intensive technology with highly suss value during the accelerating climate catastrophe is to see it as evil AF.