Hell, if we're talking about the next emergent layer of software, autonomous code / artificial intelligence, I'd say LISP has a better shot than most languages. Performance-critical code can still be refactored into C but LISP has much more built-in capability for advanced genetic coding and reflection.
If someone picks up Bel and is changed by it, then something has been changed.
YOUR problem is that you have an all-or-nothing perspective that change only matters if it is massive, global, change-at-scale.
Change matters in the small. If building Bel changed nobody except PG for the better, then it is still change and it still matters. Just not to you.
As a bonus—and this is a happy accident, not core to my point—sometimes change in the small unexpectedly leads to change in the large. See microcomputers, web browsers, and so many other things scoffed at as things that simply could never displace the incumbent technologies.
Lisp is fine for what it is, one extreme of the spectrum and a child of its time; but far from the final answer to anything.
We would be better off triangulating new ideas than polishing our crufty icons. Lisp was a giant leap, hopefully not the last.
[0] https://github.com/codr7/g-fu