On 12/29/21, k <gmkarl@gmail.com> wrote:
And without crowdfunding and starting up #OpenFabs , #OpenHW , #OpenAudit which is the better way forward, and which you can now easily startup via the global crypto crowd (over 2.5T mktcap, go spend 0.5%, ending secret closed HW better than hodling), you're just not ever going to know what's inside.
Y'know, I haven't personally tried it, but there's a pretty good chance that ending most closed hardware and software is just as far away
0.5% is enough to get open chips rolling out the door to funders and millions of paying customers. Ending closed could happen, look at all the open on (centralised) github that didn't exist at all 10+ years ago. But what must happen is putting down the $$$ to get the first one going, to put open chips out into the free market, that is the seed.
as training transformer models to automatically reverse them. The biggest impetuses to that are collecting example data and setting up the training environment. The best example data might be from chip design simulators if such exist, as then you could autogenerate variations and also automatically change points at which voltages are probed.
So go crowdfund and spend another trivial 0.01% to do that too. However fuzzing and exhaustively probing is hard, not least because today's instruction and data bus and transistor spaces require 2^n ops, reaching atoms in the universe levels. And all the effort is good for only whatever closed chips you test all of which get obsoleted as new secret ones drop every year. Decades and decades of CVE's against closed GovCorp shitHW. Whack-a-mole and embarrassment are not solving the problem. #OpenFabs , #OpenHW , #OpenAudit , #FormalVerification Those must all be crowdfunded started up and tried.