Is There a Scalability Between a G I and a G I?

In his last interview, Marvin minski said he didn't know a single person almost who worked on a g i. And to be honest, i was kind of flabber gasted but when he said that,  he was referring to the seventies and eighties. So i couldn't get around to the thought that back then there were more people working in a g i than narrow a a i compared to to day. Are we putting more effort now into general ai, or is it still all narrow ai? Because one of the criticisms i heard from some computer scientists is that there is this presumption that narrow a i would lead us towards a general a i.