frezik

Just to address this from a high level, I see this as typical of Nvidia and AMD approaches. Nvidia makes something that's engineered to perfection, but adds a bunch of requirements on it that make it expensive and supports vendor lock-in. Even if you're willing to put with that to have The Best, you might hesitate when finding out what assholes Nvidia are about everything.

AMD then makes something 95% as good, and it's cheap and you can work with them without yelling.

See also: FSR vs DLSS.

I went through an exercise with a few other developers to see if we could use it for transferring sensitive information. I was using Windows w/WSL2 at the time (now I'm full Linux for my work machine), and I believe the other two were on Macs.

Our conclusions were that while it might be useful alongside other ways, it was too clunky to use in general. One of the more useful things we could do is have developers sign git commits.

The email plugins for various clients make it easy to mistakenly think you're sending an encrypted email. When even technical people are making this mistake, then it's a big issue for widespread adoption. The plugins also don't always send it in a format that works for every client out there. We found the most consistent way was to encrypt the message in a file and attach it to the email.

The plugins don't work with modern webmail, anyway.

Public key servers are unreliable. They're largely maintained by volunteers, so this is understandable, but we couldn't recommend that the company use them. If we wanted reliability, we'd need to run our own internal keysever.

Then there's the key signing meetings we'd need to have. Even technical people find these a bother. These are, unfortunately, inherent to the web of trust model.

I really wanted to make it work. The decentralized nature of the web of trust--as opposed to the hierarchical model of TLS--is appealing to me personally. But this shit hasn't gotten better in 20 years, and at least some of it is unfixable.

There are plenty of dual-use technologies. That is, one's that have both a private sector and military application. The big secret agencies rarely keep these things to themselves. The economic advantages of QC are too great to just sit on.

QC is likely to remain the domain of liquid nitrogen-cooled machines for a long time to come, possibly forever. I can run a basic LLM on a Raspberry Pi--and I have--but it's highly unlikely QC will ever be that easy.

As for my opinion, comparing QC to early silicon computers is very misleading, because early computers improved by becoming way smaller. QC is far closer to the minimum possible size already, so there won’t be a comparable

Thanks for saying this. I see a lot of people who assume all technology always gets better all the time. Truth is, things do have limits, and sometimes things hit a dead end and never get better than they are. Those things tend to get stuck in the lab and you never hear about them.

Oh, for fucks sake.

Eh? I tried it about a year ago, and I found all the same clunky problems that were there 20 years ago.

Inflated Expectations. Most people who are aware of them will still talk about how they're going to destroy crypto. We are very, very far off from the size of QC that could possibly do that. It may not even be feasible to do the quantum juggling act necessary to handle that many qbits. It primarily effects public key crypto, with relatively minor effects on block ciphers and hashes. Plus, we already have post-quantum crypto making its way into TLS and other cryptographic suites.

And don't get me started on the morons who think the NSA already has some super secret breakthrough QC that can already break all crypto. Often from the same sorts of people who (correctly) throw Russell's Teapot at creationists.

Meanwhile, there are far more interesting possibilities that don't need so many qbits. Things like improving logistics or molecular simulation.

Sometimes, PTSD almost diagnosis itself.

In the US, the specifics depend on the state. It's often not "primary enforcement", which means you can't be pulled over specifically for that, but it can be added on when pulled over for something else. In some states, yes, the driver can be held responsible for passengers not having seatbelts on. It may also matter if the passenger is a minor.

Primary enforcement of seatbelt laws tends to make it more likely for black people to be pulled over. It's amazing how many good ideas get ruined by racism once you dig into the details.

(almost) free from dis- and mis-informations (because, at the time of BBS and Gopher, the Internet was still being born

Yeah, no. There was tons of bullshit. I ran across a post back then saying you could get psychic powers by eating the Americium sensor in your fire detector.

Goalposts keep moving, but perhaps not in a useful way. A 10,000 Wh/kg battery would be amazing, but EVs will get along fine with 160 Wh/kg. Especially if they're cheap and made of abundant materials.