How many real world production systems have benefited from formal methods, significantly more than if the same resources used to do so had been applied in other ways?
Let’s assume you can name a few, how general in the class of systems this can apply to? Or this just an obscure corner of computer science that applies to NASA and various low level system components?
I don’t count out formal methods, but can you really say none of this reminds you of the AI winter?
Is there evidence that it’s more reliable compared to equally complex systems? Or that it’s reliability would have been impractical or more expensive using traditional methods?
I didn’t hold out much hope of a surgical response (because it’s not the most common thing being done in projects) but that looks like a great link exactly on point.
It may sound cynical but as I mentioned anecdotally is does really feel like the AI winter, where simultaneously people felt there was incredible promise while hearing a steady stream of discouraging progress reports.
I look forward to reading their paper, maybe there are a lot more bright spots than can be easily seen from the outside.
I think that formal verification is going to be size limited. seL4 is a microkernel which means small, about 9000 LOC. I don't think xen at 150,000 is within the abilities of verification. So it's promising for small stuff.
How many real world production systems have benefited from formal methods, significantly more than if the same resources used to do so had been applied in other ways?
Let’s assume you can name a few, how general in the class of systems this can apply to? Or this just an obscure corner of computer science that applies to NASA and various low level system components?
I don’t count out formal methods, but can you really say none of this reminds you of the AI winter?