The peeps focusing on finding out scheming prompted an llm to generate scheming. Yawn This is the only surprising if you don’t know that llms are fancy autocompletes.

Nah but it’s a discussion that spawned from my reply so I’m invested 😁

my alt? what alt?

The problem is that even if the sources are real, it’s no guarantee that the LLM even used them

I’m not an llm hater. I run one of the biggest Foss genai services. It’s because of that that I know their limitations.

You said that you’re not going to check every search result, which implied you’re not checking every time source either which will lead you to eventually believe some llm bullshit. And of you’re using an llm just to compile sources that you check with yourself it’s no difference than a search engine without llm

If you don’t check everyone source. It might be just bullshitting you. There’s people who followed your approach and got into hot shit with their bosses and judges

You don’t lose your account for approving a bot (well maybe if you approve dozens of them or something extraordinary malicious), you’re just not allowed to approve anymore.

You also don’t get dinged by having approved others who approved bots, unless that too becomes da trend.

Additionally, let’s assume I’m a really dumb bot creator, and I keep all of my bots in the same tree. I don’t bother to maintain a few legitimate accounts, and I don’t bother to have random users approve some of the bots. If my entire tree gets nuked, it’s still only a few weeks until I’m back at full force.

Even A few weeks is a big amount and there’s no guarantee it’s that little time.

If someone keeps approving accounts who end up getting caught generating spam trees, then that account might lose privileged to approve as well.

You got another one job at the the other starlink?

There’s a reason the system requires a certain percentage of unemployment to keep working. There’s also a reason there’s homeless people and children living in flood security.

The problem is that you need to check those sources today make sure it’s not just making up bullshit and at that point you didn’t gain anything from the genai

Problem is, you cannot trust it’s not hallucinating these stats

Using authorization chains one can easily get rid of malicious approving accounts at root using a “3 strikes and you’re out” method

For example, a bot on Twitter using an API call to GPT-4o ran out of funding and started posting their prompts and system information publicly.

While there’s obviously botspam out there, this post is clearly a fake as anyone with the programming experience will notice immediately. It’s just engagemeb bait

Corporate throws you on the street to starve.

Controlled by governments or controlled by corpos and the super rich? I say there’s hardly an improvement.

As the article says, yes.

I think having a way for something like liberapay to be more closely coupled with your fediverse account so people can easily see one can accept donations would be good, but adding sub-only posts is a step too far I think.

The free version is pretty braindead nowadays. Early on it was quite better.

Do not expect anything factual from llms. This is the wrong use case. You can role play with them if you guide them sufficiently and they can help with sone tasks like programming if you already know what you want but want to save time writing it, but anything factual is out of their scope.

What is this nonsense? This is a technical post explaining why it’s not encrypted.

Plus there’s plenty of other services like matrix which can do the same thing better without enriching a billionaire

I think there’s way too many salty nerds here that either didn’t see the video or didn’t even get the point Adam was making.