> Figuring out what pages are good is no longer a feasible task for software. Of course it is. It shouldn't be at all difficult to train a large natural langua

> Figuring out what pages are good is no longer a feasible task for software. Of course it is. It shouldn't be at all difficult to train a large natural language AI model on SEO spam pages. But nobody will do it unless they can figure out a business model that makes it pay. Or a non-profit organization is formed to do search spidering for the benefit of the whole world, much like Wikipedia or PBS.