AI designed circuits are now the next thing
>>108500068Very cool, twitterman. Now post the transfer function.
>>108500068I have no idea about those things but the first two look clean and compact while the third one screams just fuck my shit up.
>>108500088Practical use seems to be that they can make tech that is harder to reverse engineer
>>108500068No need, there are SAT solvers for (locally) minimizing circuit complexity. However, in reality, the complexity class when designing a... say circuit or an assembly line or a logistics hub or a chemical plant or... is much more troublesome (Don't forget that concurrent temporal planning is EXPSPACE hard https://users.aalto.fi/~rintanj1/papers/Rintanen07icaps.pdf). In conjunction you also have to deal with packing problems.
>>108500068https://archive.is/AeM5XMeh, that's not anything new. If anything it will make new hardware harder to debug when some defect is found
>>108500068looks like typical ai hallucinations
Ah, so the goal is to make an entirely new "Black Box Architecture"
>>108500068Any test data?
>>108500111>>108500089Problem is: at some point it has to interact with sane architecture in order to work effectively. They already tried the black box design with chip blobs that are impossible to read and it doesn't work there either
>>108500068>nowif by "AI" you mean standard ML and algos yes has been for a while its nothing new.
>>108500088First two are trivial shit designed by an EE undergrad. 3rd is closer to the actual mess that is RF antenna design.
>>108500213ever wonder if there could be a reason they're not designed this way in small, compact circuits?
