I worked on formal proof in 2013 and 2014 and I don’t think they are important. There is value that mathematicians trust Lean’s proofs and trust that Mathlib has correct names to certain definitions, but there are many formal proof systems that could substitute.
That trust is valuable, because top mathematicians say that if there is a Lean proof, there is a proof. A few mathematicians reluctantly trusted HOL Light a decade ago, when the Kepler Conjecture was proven using it. That trust in Lean means that mathematicians are facing the truth now, rather than denying it.
But, from a technology perspective, Mizar and other systems existed in the 1970s. This accomplishment is all by AI.