On the bright side, once AI actually CAN answer things like this we’d have “Artificial Ethics”. A way to judge questions objectively or at least without emotional tempering for each specific case. This could solve some fundamental problems like “who watches the watchers” - we simply build the perfect watcher.

Of course we’re far away from that yet. And then we’d just ignore our perfect watcher anyway 🤣

The US helped Saudi Arabia bomb the shit out of Yemen for a long time (search). I’m not properly educated about this myself but there are war crimes, starvation, a lot of shit. I mean you can’t keep up. But to them, enforcing sanctions against Israel creates local support, otherwise they are a pretty bad Islamist regime. The Houthis are also not proxys of Iran. So US bombing them probably helps them too internally.

But there is so much shit going on that it’s hard to keep up.

Using IP laws to legislate this could also lead to disastrous consequences, like the monopolization of effective AI. If only those with billions in capital can make use of these tools, while free or open source models become illegal to distribute, it could mean a permanent power grab. If the capitalists end up controlling the “means of generation” and we the common folk can’t use it.

But aren’t these key pairs generated per session and/or per contact? So once you switch to a more secure / auditable client this only matters when communicating with people on whatsapp. But they presumably have a backdoor in their app for the NSA anyway.

Even if you can just add to it, you could have some sort of journalism file system to replace or delete previous files in newer records.

I mean “federate” with whatsapp. Apparently there is a bridge https://github.com/tulir/whatsmeow

Yeah this is very stupid. But I never liked Signal anyway.

Is there a matrix protocol based app that is planning to “federate”?

Ghostery works too for me

Not sure what you mean, of course WhatsApp can disable it’s own encryption. That would be an argument for open source third party apps and interoperability.

Just compare the civilian deaths: UN Human Rights Monitoring Mission in Ukraine has verified 30,457 civilian casualties since 24 February 2022, including 10,582 killed and 19,875 injured.

In order to justify the Genocide you need to completely dehumanize Palestinians. That is the purpose of this propaganda. The same propaganda tactic has been used before.

Hmm. It would definitely had helped if you could reply with emoticons like “lol” to classify jokes, not just with thumbs up.

Advances in AI could then also tweak the content sorting so that people are always kept in the optimal engagement mood. I mean they try to do that now.

Not sure what you’re trying to say either, but fascist speech using lies is fascist recruitment. That is why autonomous anti-fascism is right to disrupt fascist recruitment events in universities. Because the state or moderates care more about maintaining order. So you have to disrupt the recruiting by any means.

So if your argument is that “sunlight is the best disinfectant” then no, it definitely isn’t. There is historical evidence.

Ideally the AI can actually learn to differentiate unhinged vs reasonable posts. To learn if a post is progressive, libertarian or fascist. This could be used for evil of course, but it could also help stem the tide of bots or fascists brigading or Russia’s or China’s troll farms or all the special interests trying to promote their shit. Instead of tracing IPs you could have the AI actually learn how to identify networks of shitposters.

Obviously this could also be used to suppress legitimate dissenters. But the potential to use this for good on e.g. lemmy to add tags to posts and downrate them could be amazing.

There is nothing to keep you from using factors of 1024 (except he slightly ludicrous prefix “kibi” and “mebi”), but other than low level stuff like disc sectors or bios where you might want to use bit logic instead of division it’s rather rare. I too started in the time when division op was more costly than bit level logic.

I’d argue that any user facing applications are better off with base 1000, except by convention. Like a majority of users don’t know or care or need to care what bits or bytes do. It’s programmers that like the beauty of the bit logic, not users. @mb_@lemm.ee

That would be amazing! You could store the entire 450TB of ebooks in annas-archive on 4 of those disks!!!

Then comes the discussion if “1KB” is 1024 bytes or if 1000 bytes is a kilobyte.

It’s the metric system and it’s standard now. 1 kilobyte is 1000 bytes, just like 1 kilometer is 1000 meters. It is much easier to convert 20.415.823 bytes into megabytes - 23.4 MB.

Only windows insists on mislabeling the base 1024 kibibyte as kilobyte. The metric unit is much easier to use.

Sorry, I’m just curious in general how fast these local LLMs are. Maybe someone else can give some rough info.

How fast are they with a good GPU?

Huh? Doesn’t everyone assume he has a whole PR team that probably runs hundreds of sock puppet accounts? Like human bots? I mean isn’t a PR team basically a given for a billionaire or any special interest group?