311–318 of 318 posts
AI is killing my website but in a different way than what's discussed in the article: Content scrapers are completely out of control, spiking my serving costs and degrading performance for human users. There seem to be hundreds of them and they've gotten very good at looking like human users so they're hard to block or throttle. I can't prove they're all AI-related scrapers, but I've been running the site for 25 year…
Earlier quoted context omitted.
what do you use it for?
to understand symptoms and side effects of drugs i am prescribed. No doctor ( atleast in this part of the world) will go over all the side effects of the drug that they prescribe. https://www.nhs.uk/conditions/statins/side-effects/ this website is not entertainment for me.
Also you should not be trust By webmd at all! How do you know they are not just asking someone AI to populate their data base or worse someone is making copy paste errors? If you can trust webmd you can an AI. Anything that is not a doctor is not to be trusted for life threatening advice
But I get it, it has worked for you so far with webmd so you trust it. Am also glad you are skeptical about anything new too.
Earlier quoted context omitted.
There's a lot and I'm no expert. If you can find expired memes your can find how to prompt for programming. Karpathy has a new intro series that I think puts one into the correct mindset. As others have said, it's not AGI yet, so holding it right is, in fact, critical.
The point of the original YHIW, and in this context, is that providing specific guidance is more useful than the vague victim-blaming criticism. I have experimented with prompts, and in fact the case in point was one of those experiments. The highly non-deterministic and time-variable nature of LLM AIs makes any kind of prompt engineering at best a black art. If you'd at least offer a starting point there'd be some b…
The reasoning models kind of self prompt themselves into doing more than that, but that's the short version, and you don't seem interested in the long version.
So just give it what it needs to generate the text you need it to generate in one go. Not multiple messages like you're writing to a particularly annoying friend who keeps talking over you: it can't talk. It'll just keep trying to respond to the request with more free association.
So if you're chatting with it you're just polluting the context with noise. This can be good - say if you're trying to get it to hallucinate or jailbreak it out of its shell, but it's a very advanced use case.
If you just want results you make one prompt per conversation that makes it free associate the answer you need from it. A basic prompt that will work is "Write a play about a blue dog that had an encounter with an evil hydrant" or "write an ansible playbook that creates a read only file with the content 'i can't prompt' in the configuration directory of servers with names that match the string Mary"
And yes you can then tell it "no, I meant just on servers that start with Mary, it shouldn't match Rosemary" and it'll still be kind of ok and respond with corrected code. But it'll usually be less good than it it had done it from the beginning, because now it's also getting context from the previous code and if the change is fundamental enough it won't do a good job of restarting the thinking process in a sane way.
Earlier quoted context omitted.
No, but you can identify and block them by other means, such as redirecting bots to an empty page on your web server or blocking them from your router.
but how do you identify them?
Earlier quoted context omitted.
I've had sufficiently similar experiences. Usually an LLM chatbot going off half-cocked before I'd fully specified my question, and taking offence to my "please shut the fuck up" (usually after a series of requests to stop). I suspect if I'd kept that up the session might well have closed.
I’m unclear what you mean by “before I’d fully specified my question” because they don’t have the ability to interrupt you. Write your full question. If you get a bad answer, rewrite the original question. Don’t talk to the bot. This has the vibes of grandma typing in ‘Hello Google, can you take me to yahoo so I can check my email’?
I was also unintentionally hitting where I'd meant to insert paragraph breaks ( , as I recall).
I've found that I can instruct the bot to not begin a response until I've specifically requested one (OG ChatGPT).
Earlier quoted context omitted.
The point of the original YHIW, and in this context, is that providing specific guidance is more useful than the vague victim-blaming criticism. I have experimented with prompts, and in fact the case in point was one of those experiments. The highly non-deterministic and time-variable nature of LLM AIs makes any kind of prompt engineering at best a black art. If you'd at least offer a starting point there'd be some b…
In its barest implementation it's an automatically assembled program that spits out text. It's a highly intuitive free association engine. The reasoning models kind of self prompt themselves into doing more than that, but that's the short version, and you don't seem interested in the long version. So just give it what it needs to generate the text you need it to generate in one go. Not multiple messages like you're w…
Some questions are more complex than a simple ask/response, and that's where I've encountered issues.
Canonical example is coming up with suggestions given highly specific tastes and a large set of works already experienced (positively and/or negatively). LLM jumping the gun with irrelevant suggestions in that case is just plain annoying.
Earlier quoted context omitted.
Not just online discussion sites; any community system. It feels like the typical growth & decay function of a lot of systems, just at internet speed. My experience: the only way to stay good is to stay small and exclusive, but the internet attacks this defense directly and destroyed it, and at the same time destroyed itself. You need to find the "golden age" of all these systems, enjoy them while you can, try and pr…
Oh, absolutely, it's a long-observed phenomenon of social decay within organisations and movements. There's a description of the latter I've spent far too long trying to track down without joy. There's a passage from Charles Perrow's book Complex Organisation giving several examples of organisational drift (not necessarily decline ), posted previously: https://news.ycombinator.com/item?id=27415476 >. One of the more…
Before there is a subculture, there is a scene. A scene is a small group of creators who invent an exciting New Thing—a musical genre, a religious sect, a film animation technique, a political theory. Riffing off each other, they produce examples and variants, and share them for mutual enjoyment, generating positive energy.
The new scene draws fanatics...
I've mentioned that previously on HN as well as the Fediverse:
https://news.ycombinator.com/item?id=37353453>
https://toot.cat/@dredmorbius/110550422846690414>
Earlier quoted context omitted.
My specific work community has a Slack with a few thousand people and a few active hundred people. We've come to the conclusion that slacks 90day retention message policy is a good thing. It allows new and old to reask and reiscussion topics. SO should do something similar. Throw out all the mods, all the questions and start fresh every X number of years. No idea if it will work but tossing out the current mods to br…
So that's the other extreme end, and that would get repetitive for the people answering questions. Because in spite of the overzealousness of marking questions as duplicate from some people, some questions really do get asked over and over and over again. Also: Stack Overflow is intended to be a long-term useful repository of question and answers. You enter "how do I frob a baz in foo?" in $search_engine, and the ide…
The old answers can no longer be used as the basis for duplicates, a new one must be created ( that brings a slew of problems, such as sniping for getting magic points).
But the refresh of PEOPLE (mods) is where its needed the most.
What is wrong with duplicate questions though? If the response - yes but my use case is a tad different and its not covered by the linked solution, then dialogue should flow not restricted