I don’t really know how I feel about this. On one hand the algorithm is presumably spitting out something unique based on other work rather than regurgitating other people’s work. On the other hand, they are making use of a huge body of work to create that new unique work. Is that acceptable? I don’t know.

The other side of this is, can you really copyright code that has been produced by an ai? If something has been created by a mechanism, my very limited input from a human, can you really call that a creative work? In the monkey photo case, it was determined that the photograph that was taken by the monkey could not be copyrighted by the photographer because the photographer did not take the photo. If you have a mechanical monkey spitting out code for you, can you copyright the equivalent of a mechanical monkey pressing a button?

I don’t use desktop on most of the Linux machines I remote into, but if I did I’d use tightvnc-server.

Usually included in the distro, and the app is super straightforward on every platform. I was doing it 20 years ago so it’s easy enough a dumb kid from the sticks could figure it out.

Letting regulators nose under the tent is bad. It might feel good to gotcha Twitter and Facebook, but they’re always coming for us next. :(

It only happens every time. But for some reason people keep on trusting megacorps to keep them secure.

When you are letting big companies that have huge targets on their back have your most critical data, it’s just a matter of time. I prefer self-hosted solutions, particularly ones you can hide from the rest of the world in addition to being secure on their own…

I think there are some real dangers of having non-humans involved with court proceedings.

First there’s the obvious slippery slope of first your lawyer is an AI, then the prosecutor is an AI, then the judge is an AI and suddenly we’re living entirely off the dictates of an AI system arguing with itself.

Second, there’s the fact that no AI is a human. This might not seem important, but there’s a lot of truth that a human can perceive that an AI can’t. The law isn’t computer code, it’s extremely squishy and that fact is important to it being just but it’s also important because you can’t just enter text into a prompt and expect to get the results out of the system you want. There’s a big difference between the same question asked by a judge who appears to be convinced by your argument and a judge who appears to be skeptical of your argument.

You might make an argument that it’s just traffic violations, but there’s a slippery slope there as well. First it’s traffic violations, eventually you might have poor people making use of the AI for serious crimes because through degrees you go “oh, it’s just a traffic violation, oh it’s just a low level possession charge, oh it’s just for crimes with a guilty plea anyway, oh it’s just a tort claim, oh it’s just a real estate case…”

Another thing is as AI expands, suddenly you get a potential risk with hackers. If you have a really important court case, it might be justifiable to pay someone to break into the AI and sabotage it so you win the case.

The argument is pretty weak. I suppose if grammarly is also a cybersecurity threat…

I’m hoping 2023 is the year tech giants continue to be humbled. Getting treated like prophets because they lucked into being part of a giant asset bubble isn’t good for anyone.

There’s definitely things that need to stay human. What a nightmare if customer service lines are replaced with voice recognition and speech synthesis and a perfect corpospeak ai wasting your time, so you hire your own perfect corpospeak ai to talk to theirs… What a dystopian nightmare.

That being said, I could see it being really useful if guided properly by a human being for certain things. For communication, if you can have the person expressing the message and the ai expanding upon the ideas and the people choosing whether the output is acceptable or not and either changing it or changing the message it’s trying to express you could have people communicating better because the ai can patiently explain basic assumptions over and over again while a human may get frustrated and neglect to do so.

Sort of looks like reverse net neutrality.

Imagine that!

There could be a bunch of bottlenecks. Disk I/O to memory, memory I/O to the CPU, and the number of threads the program uses.

It appears that there’s a -threads option in ffmpeg you could set to 8.

One great tool for finding out if you’re being stalled by disk IO as well as other things is glances under Linux. If it’s an Io wait issue, you might be able to improve things by closing processes that are eating up memory, or by tuning swappiness or other memory parameters.

Extremely interesting work, it probably isn’t easy getting so much stuff together to pull data out of so many different types of software.

For obvious reasons, I wish more places would migrate their communities to something that supports activitypub. FediBB looks pretty interesting for people who are used to phpBB.

Hopefully it ends up supporting porting to platforms that support ActivityPub in the future!

After my experiences self-hosting, I think every house should have a low electricity server people do most of their stuff from. Not having to ask permission to use your own services is a liberating feeling.

I live somewhere that solar isn’t an option for about 9 months a year, but it would be awfully cool running my hardware using directly collected renewable energy for those 3 months a year!

True, repurposing low energy systems like those is a clear win.

A lot of people buy Raspberry Pis for projects that just need a cheap computer. To me that’s a shame, because there’s a lot of cheap fanless computers out there that are as good or better for the purpose of being a computer.

Really interesting video… but I ain’t eatin’ that.