AIs benefit from negative examples too. As long as the inputs are accurately labelled then it enhances what it knows.

For example, if I go to ChatGPT I can ask it to produce a wild parody of a far-right rant about something and it'll be able to do so. I could feed it an existing far-right rant and ask it to analyze what it's actually talking about and it'll have a better ability to figure out what's going on if it knows that style from its training. I could tell it "write an article about this subject that specifically avoids the tropes and pitfalls of far-right media."

We've seen frequently the pitfalls of trying to censor a model's training data to exclude something that's part of the common world around it, the model either "fills in the gaps" itself or just ends up insane. When the nude human form was excluded from training data for early image AIs it wound up making weird Cronenbergs because it had no idea what was really going on underneath the clothing, for example.

Security robots should be able to fill most of those roles, I would imagine.

Trump doesn't joke. He doesn't really comprehend the concept. What he does is say something outrageous and then if there's pushback he'll claim he was joking so it doesn't look like he's backing down.

Honestly, the "distractions" are at this point so far far worse than the Epstein files themselves are that I do kind of want to focus on them.

Last I heard the judge hadn't decided yet whether the current dismissal of the case would be "with prejudice", which would prevent reopening it. Presumably his lawyers will be bringing statements like this to the judge's attention to further drive home the argument that this is deliberate persecution.

Alas, sometimes smart people do really dumb things now and then.

It really baffles me sometimes how clever people can make such basic mistakes when doing crimes.

If you're planning to do something illegal don't literally document your crime. Or if you do need to do some written-down planning ahead of time, a major part of that plan should include how to destroy that documentation at the earliest possible moment.

It's a real thing.

Oh man, I used to look forward to deleting lots of stuff so I could watch this run.

Choice is a good thing. Small communites are good. Decentralized is good because then big bullies can't come in and take your community away. Small size and fragmentation are features, not bugs.

Would be nice if the Fediverse had more of that.

There are three or four different "Technology" communities. But there might as well just be one, it's the same bubble in each of them.

2-10 million per launch, but each launch will carry ~60 satellites. So that's as little as $30,000 per satellite.

Starship is going to fundamentally change what the launch industry is. It's easy to overlook.

The only people inconvenienced by "poisoning" your old comments are actual humans who come across old threads through Google or whatever in search of helpful information.

When Reddit sells data to AI trainers or whatever, they sell their internal database. It's got the deleted content in it still. So do the public archives that third parties have collected and made available as torrents.