Totally could be a defect in the lamination we don’t find out for years yet

Having had drives which sat for many years and spun right back up without corruption makes me think 1% is too generous maybe 0.05% per year at most

The claim you're responding to is that hard drives lose "magnetic charge" at a rate of 1% per year, not that bits get corrupted at a rate of 1% per year. The error correction in hard drives is far simpler and weaker than what's used in SSDs, but it does exist. So we should expect that there's a significant margin for data degradation before any observable data corruption begins. (This is true for SSDs, too; the first symptom of data degradation is reduced read performance as slower, more complex error correction methods kick in, then much later the host starts to actually get read errors or bad data.)

The magnetic strength of particles on the disk can decay at 1% per year, but the drive won't have issues reading them until they fall below a threshold where they can no longer be read. It could take decades.

Magnetic HDDs also tend to have inbuilt SMART features to monitor disk performance and health, so they can inform when they are beginning to give problems. So the advanced user can recover the data before the HDD fully fails.

In my experience, flash drives tend to get problems suddenly leading to data corruption, and it may not be immediately apparent to the user till it is too late. I haven't such problems in recent years though, so maybe flash drives have become smarter too.

Not just China but almost all of the world. We have parity in most countries right now. The whole point of the WTO was to do technology transfer and allow the US to double down on high margin, finance parts of business since the manufacturing game had the low fruit all picked and wasn’t valuable property anymore. The exception for the west would be to specialize on optics and silicon, two places China is still far behind.

“Gee I wonder what reputational harm could come to me for spamming the world with slop, let’s find out… for science!”

Doesn’t that make it worse? Lmao

Several of the candidate variable objects are characterized in the results section of the paper. The model is also tested for effectiveness against synthetic data. It appears to be a useful method and the paper describes a plausible path for it to aid future discovery.

But it can’t because the error rate is still too high even for the most trivial examples

This processor is state-of-the-art for silicon quantum computing. It's where modalities like superconducting were 15 years ago, and superconducting does not create noise these days https://www.nature.com/articles/s41586-024-08449-y

Gate fidelity significantly less than 100 is always noisy, regardless of the qubit itself

Sure, I'm not disagreeing that this processor is noisy, just providing enough context to say that it's fine. Historically, these devices improve enough to be under threshold at which point it doesn't matter that they are noisy cause error correction protocols can be run on top of them.

Very cool! Thanks for sharing. I would be making stuff like this if I didn't work for a legal team with a consumer electronics company attached to it.

Haha, I also have a background in law (in addition to the programming) and I've found that a stark context switch from e.g. reviewing contracts to writing code keeps the brain quite young! Or then it just deteriorates the brain faster and I'm already starting to rot... I guess the future will tell