I'm not so convinced. A lot of people have been noting the rapid development of ML systems in the past few years and projecting continued exponential improvements based on previous growth rates, but unbounded exponential improvement doesn't happen. This is an S curve and I think we're already well into the diminishing returns part of the curve. I think future growth is going to require increasingly impractical amounts of hardware for ever smaller levels of improvement.

People saying that this version is flawed but still amazing, so the next version is going to be perfect and mind blowing are going to be disappointed I think. The next version will be slightly better but still flawed. The version after that will be a touch better but annoyingly still not quite there. Constantly teasing you that full success is just around the corner while never quite getting there.