We're currently at the part of the story where chess AI can beat amateurs and is rapidly improving. Kasparov's not worried, he's an expert. He's more creative and skilled than an AI.

Why would these machines stop improving at average this time?

I didn't say they weren't improving. My point is that if our justification is comparing the top 10% of performers to AI to justify human performance, that speaks poorly for the replacement of the bottom 10% of humans.