Because, as the article goes on to say, the problem isn't that genAI repeats copyrighted material. Rather, its automation and massive scale upends the labor incentives that copyright is intended to provide, in a way that a human synthesizing knowledge from multiple sources can't.
In other words, it doesn't really matter whether ChatGPT or Bing Chat or DALL-E are tweaked so that they no longer violate the letter of copyright law, if they deprive -- at scale -- all their data sources of the views, recognition, and revenue that incentivize them to go gather and produce this valuable data in the first place. Why bother researching and writing an article if you're never going to get any pageviews for it, because Google places your main point atop the search results page?
I find the inverse question much more interesting. The whole point of the copyright act was to advance the "useful arts and sciences", by giving them a temporary monopoly right. Now the question is what's the point of copyright if we have a machine that can generate infinite art at near 0 cost. Copyright isn't there to feed artists for the sake of it.
Sadly there are some powerful players with a vested interest in making sure copyright doesn't "regress". Will be interesting to see how it all develops, but one can dream
In other words, it doesn't really matter whether ChatGPT or Bing Chat or DALL-E are tweaked so that they no longer violate the letter of copyright law, if they deprive -- at scale -- all their data sources of the views, recognition, and revenue that incentivize them to go gather and produce this valuable data in the first place. Why bother researching and writing an article if you're never going to get any pageviews for it, because Google places your main point atop the search results page?