Shpongle, when the walls melt.

Practical, but not really equivalent though because of nil punning.

For years now I have only read ebooks on my phone, so one evening I decided to get back to the habit of reading real books.

So I take my time and carefully pick just the right book, gather some pillows, turn off the lights and lay comfortable on the couch. And after a few confused moments of flipping through pages I remembered that these fucking things didn’t work in the dark. And I really don’t like to read under a bright light anymore so back to reddit it was for that evening.

That said, I think I’ll skip this one, doesn’t sound too comfortable.

Speaking as just a hobbyist, a more developer oriented community focused on the topic would be nice, if someone is up to the task.

It’s currently hard to find any good information about how to actually use LLMs as part of a software project as most of the related subreddits etc. are more focused on shitposting and you don’t currently really want to talk about these in general tech/programming forums without a huge Don’t shoot I’m not one of them! disclaimer.

Regarding little Bobby, is there any known guaranteed way to harden the current systems against prompt injections?

This is something that I’m personally more worried about than Skynet or mass unemployment now that everyone and their dog is rushing to integrate LLMs into to their systems (ok worried maybe a wrong word, but let’s just say I have the popcorns ready for the moment the first mass breaches happen with something like the Windows Copilot).

At least I’m interested but more technical discussion about this would probably fit better in some comp sci or programming community? Though most of those are a bit hostile to the LLM related topics these days because of all the hype and low effort spam.

Is the whole “You are an LLM by OpenAI, system date is etc.” prompt part of the system message?

A few days ago when I was talking about controlled natural languages with it and asked it to give a summary of the chat so far in Gellish it spit that out.

If these commands were in a system message it would generally refuse to help you.

Doesn’t it usually fairly easily give its system message to the user? I have had that happen purely by accident.

I’m not sure if I’d call that reverse engineering any more than using a web browsers View Source feature.

But it’s interesting how it works behind the scenes and that only way to get these models to interface with the external world is by using the natural language interface and hoping for the best.

Yep, for example I think the joke about a guitarist fingering a minor is gone now from it’s repertoire. Finetuning and guardrails probably also limit its capability to “understand” jokes in general.

IME it’s explanations of jokes are usually really off too, probably partly because of the guardrails and partly because it’s understanding is so surface level.

Edit: tried to see if understands the joke about guitarist but now it refuses to even explain it and just flags the question and freezes.

There are probably better solutions but I guess simplest way would be to solve that at the client end?

Give users the option to merge community views from different instances (maybe too much hassle for the average user), have the client do it automatically for some specified communities, or have a mechanism by which the communities can hint the client to merge their content with specific “friend” communities.

From users POV the last option would be the easiest (but it should be possible to opt out of it or customize the behaviour). To prevent trolling and harassment the merging would require an authentication from all participating communities. That doesn’t prevent multiple posts on the same subject but if majority of users see the same combined content the likelihood of double posts decreases. It would still spread the load between instances, and if they want the different instances could specialize on different aspects of the subject.

Just a thought. I don’t know if it makes any sense from technical point, maybe it would be easy to implement without any changes on the underlying protocols or maybe it would require some ugly kludges and would just overcomplicate everything or is something not many people would even need or want.