To be clear, Neural network systems will not become sentient super intelligences by themselves - another system, or multiple other systems, likely would need to be working in tandem. It's also entirely likely that Neural Network systems lead to a dead end and it's actually some other machine learning technology that would actually cause an intelligence explosion.
It's inarguable, though, imo, that no one should be taken seriously who says LLMs or other Neural Network technologies would improve the efficiency or reliability of companies or government systems.