Perhaps ironically, there are some aspects of this project (parallelism and AI) that are now massively important. It's just that they were ahead of their time on those aspects, and Prolog is now largely a historical bywater.
[1]: https://en.wikipedia.org/wiki/Fifth_Generation_Computer_Syst...
That research in the US was defunded in a hurry after they realized they were training foreigners how do to this stuff
https://sgp.fas.org/crs/nuke/RL30143.pdf
If you look at Hadoop and other developments in commercial parallel computing I'd say many of them (all the ones that don't involve some special interconnect between the processors) are more closely related to what the 5th gen project had converged on at the end... At that point they had given up on Prolog (as we know it, they did settle on a subset called KL0), Dataflow processors and other things and settled for plain ordinary message passing and "shared nothing".
"VRML has never seen much serious widespread use.[18] One reason for this may have been the lack of available bandwidth.[19] At the time of VRML's popularity, a majority of users, both business and personal, were using slow dial-up Internet access."
https://www.goodreads.com/book/show/22617.Seduction https://en.wikipedia.org/wiki/Simulacra_and_Simulation
and then how there was a craze for it in the early 1990s (like this failed TV show)
https://en.wikipedia.org/wiki/VR.5
then there was The Matrix and by the time Sword Art Online: Ready Player One came out people were taking it for granted.
From a hardware perspective modern VR headsets are "good enough", the success or failure on VR depends on (1) software that is compelling and (2) will VR use ever "move the needle" compared to reading, tweeting, watching TV, conventional video games, ...
The web3 is a generation too early.
https://news.ycombinator.com/item?id=18531
Segway, Google+, Google Glass, Theranos