Problem is that the LLM operators don't care if you're rude. They copy and paste your response to the LLM and probably don't read it themselves. If they do read it they don't suffer any ego hit from it because if there was any error it was the LLM's not theirs and their LLM is busy telling them how brilliant and unparalleled they are and how wrong the haters are in any case.
CURL is free to try it, but I'm doubtful being rude will meaningfully improve things. I'm confident it won't improve the ratio of good to bad reports because non-chatbot powered submitters are sensitive to rudeness or even the threat of potential rudeness, and so this approach could easily reduce total volume some but mostly in the reduction of good submissions.
When people don't provide a citation online when discussing some specific instance like this-- which could be provided with a couple clicks and would radically improve their argument a reasonable assumption is that the citation would undermine their argument.
The original model works, the new model significantly fails. LLMs have taken many cases that were on the border over the line into failure, by changing the resource management tradeoffs. (Both by giving valuable contributors a cheap way to get 'extra eyes' on their own terms, and by empowering a new generation of trisectors and trolls to flood out even the most efficient public submission pipelines).
To review documents received from a hostile and dishonest actor in litigation I used disposable VMs in qubes on a computer with a one way (in only) network connection[1], while running the tools (e.g. evince) in valgrind and with another terminal watching attempted network traffic (an approach that did detect attempted network callbacks from some documents but I don't think any were PDFs).
This would have been useful-- but I think I would have layered it on top of other isolation.
([1] constructed from a media converter pair, a fiber splitter to bring the link up on the tx side, and some off the shelf software for multicast file distribution).
True, but I've been consistently amazed by the improvements of the software in such a short amount of time (even running on the same hardware). I only have 8GB of VRAM, but it's amazing what I can do with those 8GB compared to a year or two ago.
Why is polymarket worse than cell phone games? At least polymarket essentially self identifies as gambling and isn't specifically marketed to children.
Cell phone games are knowingly losing money for fun. Polymarket is a sign of a failing, rapidly deregulatory economy.
And that's before we get into discussing the social damage to country that already sees more school shootings than weeks in a year (actually, 4x more), with rising political and civil tensions including assassinated politicians, adding potential "lose your house" to random events. As if it'll help calm things down and let us all keep a level head.
Or the implications of news companies reporting on these odds as if they reflect actual statistical likelihood, and how that gives the ultra wealthy yet another lever to control the view of reality the common people have.
I've been following a conman fantisist for a number of years and of late he's gone full LLM powered and has been churning out graduate degrees from respectable sounding places. Years ago he merely claimed to have varrious degrees, but now with the help of chatgpt he's just pumping them out.
While I'm sure a few places care many very clearly don't.
The advice here is now increasingly out of date in the era of LLMs. Trisectors are more numerous and voluminous than ever, while being less obvious and more effective at wasting your time, and approaching a far greater variety of subjects, and any kind of response to one has a much greater risk of outright aggressive and even threatening responses as sycophantic AIs escalate their users otherwise benign hyperfocus into outright paranoid delusions.
If what you say is true, shouldn’t that make the understanding of trisectors more urgent and important? If anything, this is more relevant than before.
I'm not sure. I think LLMs makes different people into trisectors and they trisect in different ways-- so not just much more numerous but also somewhat different in character.
CURL is free to try it, but I'm doubtful being rude will meaningfully improve things. I'm confident it won't improve the ratio of good to bad reports because non-chatbot powered submitters are sensitive to rudeness or even the threat of potential rudeness, and so this approach could easily reduce total volume some but mostly in the reduction of good submissions.
reply