AI 公司都是寄生虫

Merriam-Webster:

An organism living in, on, or with another organism in order to obtain nutrients, grow, or multiply often in a state that directly or indirectly harms the host.

That's it, right? The whole thing, their entire business model.

1. Scrape the whole of the internet, destroy books, siphon everything they can from whoever they can without compensation, and call it fair use.
2. Train models on said stolen data.
3. Sell metered access to the models.
4. Repeat until you've killed the host.
5. Hope you can head off the effects of AI inbreeding.

The old bargain of traffic for indexing is gone. Scrapers overload infrastructure. Clickthroughs collapse. Communities collapse.¹

Yes, licensing deals exist. Reddit is selling access to its data, but that data is human-generated and depends on a flagging community the company seems, at best, indifferent to.

They'll even show up in your community and set up a data center that nobody (well, except for the people profiting from it) wants. Maybe they'll spin up gas generators and send pollution your way. Or they'll drive up your electricity and water rates. But what about jobs? Most of those only exist during construction and, who knows, they might just hire folks from outside your area for that.

They never should've attached themselves to society, and it may be too late to dislodge them. At least we get stilted prose, images that look like Pixar rehashes, and code nobody understands.

1. How's Stack Overflow doing lately? ↩︎