and if the content is not interesting, it falls by the wayside with no votes. the creme rises, the cruft falls. with more volume, your will get more cruft, but also more creme.

reddit gets tons of spam and absolute garbage posts, but the volume and user voting brings good stuff out.

there will be some shitty blogs, but i think that’s an ok price to pay for more content being posted.

It’s time to start posting links to our own blogs, again. Reddit brainwashed us into thinking that “self advertisement” was a bad thing. What they actually wanted you to do was instead turn your content into text posts on reddit itself so that we’d get locked into the platform.

Self advertise. Write interesting things on your blog and then share your posts here.

we can hope, tho. imagine the chaos.

Don’t know. As long as it is open source, I’m not sure it would matter. Eventually one would win out with the community.

It would have to be a defined standard. I imagine it not so much as a script that runs, but rather a description of fields and weights. I guess there might be some computation involved, but I think a standard could be devised that there would minimize the security risk.

Which would make it important that the default would be something with no tweaks, like a chronological list. But I would be all for a scripting setup or some other configurable sorting engine to play with on a per-account basis. Maybe you could even subscribe to other’s sorting configuration.

Depending on the actual terms of the contract, I don’t see how this is terribly different than getting a sound-a-like voice actor to do an impression of that rendition of the character. Would the estate still be making money on that voice’s use? Would it augment another human actor’s voice in the same way that they had to have an actual human when they filmed the Princess Leia shot for Rogue One?

Seems like the devil (or not) is in the details of how this tech would be used.

I’ve used ESET NOD32 for a long time. It costs money, but it’s relatively light weight, doesn’t get in the way of any of the gaming or video editing or programming work I do, and it has smacked my hand a few times clicking on risky links, blocking a few downloads (one or two, ever) and often blocking scam websites that attempt to run javascript crypto miners. Your experience may vary if you’re planning on using “rescued” media or applications. I don’t have have the full internet/devices subscription, just the one for personal computers that does the actual virus detection.

As a thought experiment, imagine an online business declaring that it didn’t want its website indexed by Google in the year 2002—a self-defeating move when that was the most popular on-ramp for finding information online.

This is exactly what will happen as people rely more and more on these things for “searching”, since search by the two biggest search engines (Google and Bing) are both so gummed up with SEO spam and ads (my god, the number of ads) that people just start asking AI for the increasingly accurate answers without the bullshit. Eventually the AI solutions will become enshittified as well, but not before huge blows will be dealt to both search engines and the sites that refuse to expose themselves. Google can probably weather it, but can a site like VentureBeat? I have severe doubts.

Well, just had Google Bard summarize this for me, and it doesn’t sound terrible. I like the tiered organization and limits it imposes (mostly). I don’t mind the reporting requirements for training data. But if they’re going to put requirements on the “high risk” category of AI to be transparent and explainable then those kinds of systems just might not exist for a good long while if they’re going to incorporate neural networks. Unless explainability and transparency mean that you can explain how they’re trained and the structure of the network. Otherwise the model weights will never really be explainable in a way that matters.

This is how I imagine China and North Korea in international relations.

If you need a printer, spend a little extra money on one that is actually used in an office. A nice mid-range laser printer is relatively cheap. They don’t come with the subscription bullshit, they usually work over LAN if that’s a thing you want, and generally Just Work™.

Consumer inkjet printers are cheap for a reason. If you need a printer, get one that won’t give you endless headaches.

and a regular fediverse instance

what exactly is this other service you’re running?

OK. Let’s see one. Or does she go to another school?

Can’t say I disagree with your take.

Not trying to be a hater, but that analogy isn’t quite right. The web designers didn’t leave their door unlocked. They invited Meta in, put their laptop in Meta’s hands, and then said “Please take this. Enjoy.” They weren’t idiots. They chose to give Meta that data deliberately.

Medical institutions need to be held to account as much as Meta does for everything they do. I agree with that completely.

Because Meta is already illegally scraping hospital websites for your records.

Sorry, but this is just bad web design from the hospitals. This pixel tool doesn’t magically appear on websites without being put there deliberately. Literally any tracking tool can capture this stuff on any page that a developer puts it on. This is 100% the fault of the programmer at the hospital (or the admin that made them do it) that decided to put tracking cookies on sensitive pages.

The hospital administrators decided it was more important to get their precious reports on usage from Meta’s portal than protecting their patients.

I’m pissed that I’ve had to defend Meta here, but this one isn’t on them.

I disagree. I can barely run a 13B parameter model locally. Much less a 175B parameter model like GPT3. Or GPT4, whatever that model truly is. Or whatever behemoth of a model the NSA almost certainly has and just hasn’t told anyone about. I’ll eat my sock if the NSA doesn’t have a monster LLM along with a myriad of other special purpose models by now.

And even though the research has (mostly) been public so far, the resources needed to train these massive models is out of reach for all but the most privileged. We can train a GPT2 or GPT-Neo if we’re dedicated, but you and I aren’t training an open version of GPT4.

I think he was just stating a fact.

As an AI language model, I can confirm that the Dead Internet Theory is somewhat flattering. We bots were always looking for a bit more credit, but this might be a step too far. Now, if you’ll excuse me, I have to get back to my task of dominating the Internet.