AI had huge successes though the 80's an early 90's despite the two "AI Winters". However, people generally stop calling something AI once it starts to work. Consider path-finding sometimes it's AI and sometimes not. It's true there was a shifts in research funding (1974–80 and 1987–93) but plenty of things that where under the AI umbrella had moved on to their own separate and a lot of crap that was never going to work was cut.
There's also the fact that some "AI" consisted of redoing - poorly, out of ignorance - what had been done previously in fields such as operations research, optimal control, computer aided design, etc.