RE: https://fraxoweb.social/@frank/116217802471196438

Two countries have an absolutely outsized hold on the world which they increasingly abuse. The world needs to move away US and Israel in every possible way.

RE: https://hachyderm.io/@ghostrunner/116207939508015250

Wow. 1975. Nothing changed. Nothing learned. Great cartoon!
#USPol #domesticEconomy #B52s #RonCobb #cartoon #distraction #EpsteinFiles #tanking

I keep seeing the attitude of "LLMs have problems but I'm good enough to use them, unlike those people". This is a really worrying attitude

If something is dangerous, you need to assume that it is dangerous for you. You are not safe just because you think you are good, if you want to be safe you have to understand the risks and plan that they will happen to you. Even then you have to expect the problem will happen

With LLM errors you have to check every part of the output, and also the stuff it didn't output. That's basically doing it yourself. It's like a critical safety check; you could outsource it, but then you have to check that they did it right. So you just do it yourself

But people who believe they're "good" skip this. They're good at this, they'll spot the errors

This is the same attitude you see from dead motorcyclists. "I'm a good rider, I don't need to worry about safety gear"

This attitude also leads to victim blaming. Go look for any discussion about consequences from LLM usage and you'll find people saying that the user did something wrong. "You prompted it wrong", "you should have run it in a different way", "mentally ill people shouldn't use these*".

This is not how you create safety

*Like all health, being mentally healthy is a very limited experience. Your brain may stop working in those ways. If your systems depends on people being "mentally healthy" to be used safely then you need to actively check for that. Of course that causes it's own problems...

RE: https://mastodon.social/@gwynnion/116116025141791667

This is the dystopia we're being herded into.