Dearest cryptography experts, I have a theoretical question
How feasible would it be to have an AI generating cryptographic algorithms, such that new algorithms may be generated faster than ASICs can be developed? And for it to continue creating new algorithms (like one every 6 months, or whatever the minimum ASIC development time might be), for decades?
I can intuit that clearly, this is not currently possible, but just wondering if there might already be theoretical basis out there to suggest its existence in the future.
We have a bigger issue if a true AI (rather than just a well coded program) is actually created. For it to actively look for new implementations it would need to be able to replicate thought patterns (in the same way humans think at least) which presents the possibility that it outstrips any conventional barriers or failsafes we code.
In reality, I'm not sure if we'll actually ever have AI, but if we do there'll be much more use in other fields rather than new crypto implementations. Much more likely humans come up with new solutions. As for as ASICs go I honestly don't really see the problem, what are you trying to do by making ASICs obsolete (make everyone use a CPU?).