For someone out of the loop, what’s the deal with poa.st? I followed the link and it says its for shitposters which… doesn’t sound good… but could mean a few different things.

The validator would have access to real references it can use to ensure some form of correctness

That’s the crux of the problem, a LLM has no understanding of what it’s saying, it doesn’t know how to use references. All it knows is that in similar contexts this set of words tended to follow this other set of words. It doesn’t actually understand anything. It’s capable of producing output that looks correct to a casual glance but is often wildly wrong.

Just look at that legal filing that idiot lawyer used ChatGPT to generate. It produced fake references that were trivial for a real lawyer to spot because they used the wrong citation format for the district they were supposedly from. They looked like real citations because they were based on how real citations looked but it didn’t understand that citations have different styles depending on the court district and that the claimed district and citation style must match.

LLMs are very good at producing convincing sounding bullshit, particular for the uninformed.

I saw a post here the other day where someone was saying they thought LLMs were great for learning because beginners often don’t know where to start. There might be some merit to that if it’s used carefully, but by the same token that’s incredibly dangerous because it often takes very deep knowledge to see the various ways the LLMs output is wrong.

The problem with ALL the LLMs is that they don’t actually understand anything at all. They produce output that looks like other similar things but may or may not have any actual relationship to reality. So they’re incredibly advanced bullshit generators.

I would never trust a piece of code written by one of these things and you’d spend just as much time debugging what it wrote as it would have taken to write it in the first place.

For that matter you’d never be able to actually trust anything one told you about anything either as you never know if anything it has said is true or not so you literally need to research everything it tells you in order to know if it was true or not.

It could maybe work as a better interface to a search engine though. You ask it a question and it redirects you to what it thinks are the most relevant search results. E.G. “how do I do X” and it tells you “People who wanted to do X often needed to know about Y and Z, here are some of the top search results for that”, but you’d need to actually follow the links provided, not let it summarize them.

Smart devices aren’t so much the issue as internet connected ones. I use a number of ZWave smart devices, but I avoid buying any smart device that connects with wifi. You do need some kind of hub device ultimately but even in that case you can use your own hardware and software for that.

Yes there are several alternative voting systems that could serve as inspiration. The current upvote only system as done by beehaw is closest to Bloc Approval Voting (everyone picks as many options as they like, the ones with the highest votes win). A system like you describe would be more like Bloc Score Voting (everyone ranks each option on a scale, the ones with the highest scores win). Another interesting but probably confusing (and possibly not implementable in a federated system) option would be Cumulative Voting (each person gets a number of points to distribute as they see fit, the total number of points is equal to the total number of options).

90% of the time mobile via RIF, the remainder is desktop with old reddit + RES. Once the API apocalypse hits it will be 0% of the time mobile, and we’ll see if I even bother on the desktop after that.

Hmm… maybe there’s some interesting parallels to the problems we see with voting in elections. Specifically I’m referring to the problem where a first past the post voting system leads to an eventual devolution into an increasingly polarized two party system. There are various suggestions for how to address that problem like ranked choice voting. I’m not sure exactly how you’d translate something like that into a comment voting system, but maybe there’s some clues there? Maybe instead of a simple up/down and point tally you provide a series of reactions? “I agree”, “I disagree”, “This is factually wrong”, “This is spam”, “This is interesting”, etc. and allow people to pick multiple of them? Not sure how you’d translate that into a ranking algorithm but that also allows for some interesting choices like sort by most interesting, or sort by most agreed with.

While that is true and feeds into driving an echo chamber I’m not sure removing downvotes actually solves that problem. You’re still going to see posts that agree with the most popular opinion voted to the top while unpopular opinions will languish. All it really does is slow the speed of the feedback cycle, not truly stop it. The only way I see to actually do that would be to disallow voting entirely, but then you have a signal to noise problem. Strong moderation can help somewhat by removing the truly bad content (although then you have moderator bias leading to echo chambers), but you’re still drowning in a sea of comments of questionable quality.

There’s also a larger question of how federation and communities in general feed into the formation of echo chambers. There are definitely some lemmy servers I won’t be taking part in because I can already see I disagree with the vast majority of users on those instances. The entire thing also touches on the problem of tolerance for intolerance.

Honestly I would have joined beehaw except I disagree with the no downvote policy. It forces you into a terrible tradeoff where either you upvote literally everything that isn’t bad, but then have no way of actually indicating truly good content, or else only upvote the truly good content, but then have no way of indicating bad content. You could always block users that post bad content, but that’s a super heavy handed approach with no real nuance, and doesn’t help improve the community.