What I got from his interviews is not that he disliked the GoL, it's just he disliked the GoL overshadowing everything else he did (basically, becoming the GoL guy). He personally didn't see much more interesting mathematics that could be done after answering basic questions like universality (although it's likely he wasn't aware of everything the community was up to). Also, it's clear he seemed to come to terms with it in his final interviews (including the second one you linked) :)

I've played around with several CAs and Conway's rules stands out to me as one of the most interesting still, for many reasons (like simplicity, interesting patterns, long lived structures).

Reminds me of Steve Paxton, an amazing dancer who passed away recently. He led a project called “Contact Improvisations”, which became a movement form called Contact Improvisation. He taught some classes and many others contributed. 50 years later, it’s still going strong. But, he didn’t embrace this role of “Contact Improv guy” that was really available to him. He just kept doing other stuff, even as this community exploded.

I think that’s partly the nature of pure researchers. They usually have something more interesting to them than what they got famous for, and they probably don’t want to lead an organization. This is different from BDFLs like Guido van Rossum and Rich Hickey. Neither type is good or bad, and I appreciate them all.

It's almost by definition that if you get popular for something, it is not the thing you're most interested in or best at - because you're an expert in the craft and for something to be popular it has to be at least somewhat approachable by non-experts.