In this, he has been proven wrong, as nearly all machines of industry are slowly becoming more and more automated, using hardware and software specialized to that particular machine.
As far as I can tell, AI is turning out like Philosophy. Back in the day, all science was philosophy. As a particular field of philosophy became advanced enough that we can actually do things with it, we broke it out into a science, like, say, physics.
The same is happening with AI. It used to be that natural language and image recognition were AI. Now that we actually have that technology working to the point where it can do useful things, it's no longer considered AI.
Both philosophy and AI, I think, are catchall buckets for "We want to study these things, but we don't understand these things well enough to properly define them."
I mean, it's a useful catchall, especially at first, but I think that breaking down a poorly defined field into many smaller but specifically defined fields is a natural part of scientific progress.
My guess is that we will reach the capability to build something that looks like asimov-style robots, but by that point we won't see any of the technologies involved as "AI"
Yes, we know that he was wildly wrong about what would be easy and hard about robots and computers. It is bizarre to go back and read his stories in which robots are walking and talking, but it takes a massive robot brain installed in a permanent emplacement to get little more "raw" computational power than a 1970s mainframe. Unless you travel back in time and reveal the secrets of robotic brains to those in the 1950s, it's a done deal.
However one note is that Asimov's stories have one common theme that could not be known until now. Hardware is what drives everything. Today we know hardware is an enabler, software is everything.