It’s not going to be become a major problem. We have radiation hardened computing hardware, and ways to deal with single event effects, we’ve in fact got a lot of practice doing these things, because guess what: Satellites also need working computing hardware, and they’re exposed to orders of magnitude more radiation than aircraft.
Manufacturers will just have to start taking it into consideration more in the future, and ensure that the flight computers have redundant ECC memory.
And eventually “we” might come to the thought that for many things analog computing is enough. Symbolic calculation, cryptography and such, of course, need digital. But when we are talking about airplanes and satellites, perhaps not.
One thing I somewhat like about the general idea of all those LLMs is that in theory they are closer to something that can work on non-deterministic technology.
I wonder if some sort of FPGA but for analog circuits is possible. To have the advantages of re-configuration that programmable things have, but also advantages of continuous signals.
It happened once, to one aircraft, and it’s solvable with a software update.
You’re more likely to be struck by lightning the next time you leave your house than to run into this problem on a flight, and that was before the software update.
Not even once according to the article. They don’t actually know what happened on that flight, but their simulations can’t test test for cosmic radiation and didn’t reveal any other errors, so they presume it must be the cause. Then made up a story about that being a day of heavy day of solar activity, which the article refutes.
It’s not going to be become a major problem. We have radiation hardened computing hardware, and ways to deal with single event effects, we’ve in fact got a lot of practice doing these things, because guess what: Satellites also need working computing hardware, and they’re exposed to orders of magnitude more radiation than aircraft.
Manufacturers will just have to start taking it into consideration more in the future, and ensure that the flight computers have redundant ECC memory.
And eventually “we” might come to the thought that for many things analog computing is enough. Symbolic calculation, cryptography and such, of course, need digital. But when we are talking about airplanes and satellites, perhaps not.
One thing I somewhat like about the general idea of all those LLMs is that in theory they are closer to something that can work on non-deterministic technology.
I wonder if some sort of FPGA but for analog circuits is possible. To have the advantages of re-configuration that programmable things have, but also advantages of continuous signals.
Not to mention that cosmic ray bit flips are extremely rare. A sys admin might encounter one or two during their entire career, if any.
Yeah, but the problem is, the airplanes don’t. The company didn’t think it would be a problem and now it’s a fucking problem.
It happened once, to one aircraft, and it’s solvable with a software update.
You’re more likely to be struck by lightning the next time you leave your house than to run into this problem on a flight, and that was before the software update.
Not even once according to the article. They don’t actually know what happened on that flight, but their simulations can’t test test for cosmic radiation and didn’t reveal any other errors, so they presume it must be the cause. Then made up a story about that being a day of heavy day of solar activity, which the article refutes.