The problem is the same prompt will yield good results one time and bad results another. The "get better at prompting" is often just an excuse for AI hallucination. Better prompting can help but often it's totally fine, the tech is just not there yet.
While this is true, I have seen this happen enough times to confidently bet all my money that OP will not return and post a link to their incorrect ChatGPT response.
Seemingly basic asks that LLMs consistently get wrong have lots of value to people because they serve as good knowledge/functionality tests.
If you want a correct answer the first time around, and give up if you don't get it, even if you know the thing can give it to you with a bit more effort (but still less effort than searching yourself), don't you think that's a user problem?
I briefly got excited about the possibility of local LLMs as an offline knowledge base. Then I tried asking Gemma for a list of the tallest buildings in the world and it just made up a bunch. It even provided detailed information about the designers, year of construction etc.
I still hope it will get better. But I wonder if an LLM is the right tool for factual lookup - even if it is right, how do I know?
I wonder how quickly this will fall apart as LLM content proliferates. If it’s bad now, how bad will it be in a few years when there’s loads of false but credible LLM generated blogspam in the training data?
The person that started this conversation verified the answers were incorrect. So it sounds like you just do that. Check the results. If they turn out to be false, tell the LLM or make sure you're not on a bad one. It still likely to be faster than searching yourself.
That's all well and good for this particular example. But in general, the verification can often be so much work it nullifies the advantage of the LLM in the first place.
Something I've been using perplexity for recently is summarizing the research literature on some fairly specific topic(e.g. the state of research on the use of polypharmacy in treatment of adult ADHD). Ideally it should look up a bunch of papers, look at them and provide a summary of the current consensus on the topic. At first, I thought it did this quite well. But I eventually noticed that in some cases it would miss key papers and therefore provide inaccurate conclusions. The only way for me to tell whether the output is legit is to do exactly what the LLM was supposed to do; search for a bunch of papers, read them and conclude on what the aggregate is telling me. And it's almost never obvious from the output whether the LLM did this properly or not.
The only way in which this is useful, then, is to find a random, non-exhaustive set of papers for me to look at(since the LLM also can't be trusted to accurately summarize them). Well, I can already do that with a simple search in one of the many databases for this purpose, such as pubmed, arxiv etc. Any capability beyond that is merely an illusion. It's close, but no cigar. And in this case close doesn't really help reduce the amount of work.
This is why a lot of the things people want to use LLMs for requires a "definiteness" that's completely at odds with the architecture. The fact that LLMs are food at pretending to do it well only serves to distract us from addressing the fundamental architectural issues that need to be solved. I think think any amount of training of a transformer architecture is gonna do it. We're several years into trying that and the problem hasn't gone away.
Yup, and worse since the LLM gives such a confident sounding answer, most people will just skim over the ‘hmm, but maybe it’s just lying’ verification check and move forward oblivious to the BS.
People did this before LLMs anyway. Humans are selfish, apathetic creatures and unless something pertains to someone's subject of interest the human response is "huh, neat. I didn't know dogs could cook pancakes like that" then scroll to the next tiktok.
This is also how people vote, apathetically and tribally. It's no wonder the world has so many fucking problems, we're all monkeys in suits.
It depends on whether the cost of search or of verification dominates.
When searching for common consumer products, yeah, this isn't likely to help much, and in a sense the scales are tipped against the AI for this application.
But if search is hard and verification is easy, even a faulty faster search is great.
I've run into a lot of instances with Linux where some minor, low level thing has broken and all of the stackexchange suggestions you can find in two hours don't work and you don't have seven hours to learn about the Linux kernel and its various services and their various conventions in order to get your screen resolutions correct, so you just give up.
Being in a debug loop in the most naive way with Claude, where it just tells you what to try and you report the feedback and direct it when it tunnel visions on irrelevant things, has solved many such instances of this hopelessness for me in the last few years.
I am unconvinced that searching for this yourself is actually more effort than repeatedly asking the Mighty Oracle of Wrongness and cross-checking its utterances.
Very interesting. These terminals are super common in the Netherlands also. The exact same ones. Kudos to the author, rigging up a BGA chip is no small feat. I love this kind of curiosity (one of the reasons I like HN)
> As others have mentioned, it would be interesting if the author stuck in and got past the tamper checks to see if it would work as normal. Oh well!
That's the kind of thing a lot of guys in suits would take very seriously. After all you have to connect it to the banking network to see if it works as normal. Not advised.
It could also lead to lowlife types putting pressure on you to help them do that. Not the kind of thing to brag about.
I thought transcription was a solved problem now. I run whisper at home and it's blazing fast and accurate with the large model <3. If anthropic is much worse they need to up their game. Or just use Whisper until they do.
Uhm what am I not getting? This has always been possible.
Edit: I think what's new is the UI to add the new engine by typing the URL by hand. Previously engines needed to support OpenSearch to auto-add them (but you could easily add custom ones this way). However pretty much all of them do (it's no trouble, just some XHTML in the page). I don't know why you would do it by hand.
Well yes but the point was kinda moot because it was easily possible to add an opensearch provider and every self-respecting search engine supports that. Why wouldn't they? It makes it easy for users to adopt them and it's not hard, just some boilerplate code.
So the URL entry box wasn't really needed for the main purpose. It's nice to use the feature to use with things that aren't literally search engines though. Or to tweak the URL. I wonder, is the AI avoidance thing very effective? Not that I use Google directly but still.. (I use SearXNG as meta search)
Yes, it's just a way of going directly to google's "web" search instead of "all" so you don't get the AI summary and half a screen of assorted guff before the search results themselves. In that sense it's 100% effective.
Ahhh I see. I thought it was somehow trying to filter out AI generated sites in the web results. Which wouldn't be airtight of course. Now I understand, thanks!
Is (was) it possible to set an OpenSearch engine as the default search provider, though? At least I haven't been able to find a way to do that, so maybe that's what's different now.
I guess it's a regression fix (perhaps Windows-specific) where it used to work and broke at some recent version? Because the feature as described is indeed nothing new in Firefox.
From the linked bug:
> Firefox now supports adding your own custom search engines. Just right-click a search field of a supported website and select Add Search Engine, or go to Settings > Search > Add (below the search shortcuts table) to manually enter a search URL.
Edit: Actually, I don't have browser.urlbar.update2.engineAliasRefresh configured at all (neither true nor false, just not configured).
But for me I've always been able to do it. All I had to do was to simply visit my SearXNG instance once, then it would pick it up (OpenSearch API) and then I could just set it as default.
In fact I remember setting up my new tablet a few months ago, and I didn't need to mess with about:config at all, it worked like I mentioned above. On firefox mobile they make it super hard editing about:config for some reason so I'm sure I didn't do that.
Yes, it has alway been working like that, but you couldn't add a search engine that doesn't have an opensearch.xml published (eg chatgpt), make custom search egines such as only search github in your org, or mix and match them such as perplexity with google suggestions
The only way I could ever add custom search engines was through a separate extension. I have no idea how you've been doing this such that you think it's always been possible?
Interesting, for me they don't. I can only choose between Google, DuckDuckGo, and Wikipedia, even though I have more custom search engines available. They work via their configured shortcuts, but they don't appear in the search settings.
Huh no? I've just added it in settings. No custom extension needed. I'm pretty extension-averse because I already need so many :) Ublock, Sponsorblock, Dark Reader, password manager, consent-o-matic, sideberry I really can't do without.
But I've always been able to do it. All I had to do was to simply visit my SearXNG instance once, then it would pick it up and put it in the list of search engines (OpenSearch API) and then I could just set it as default.
One thing I could not do was edit search engines (e.g. the URL they visit). I still can't do that in fact but maybe that's new?
If you check the bugzilla thread in the OP, this is specifically about adding manual search engines outside of automatic OpenSearch support.
Theres a panel where you fill out the name, search URL, suggestion URL, and search keyword yourself, under settings. You can add whatever you like, even if its not OpenSearch compatible, just like you already could on Chrome (for the better part of a decade, mind you)
If a dev is lazy and doesnt incorporate OpenSearch functionality you can homebrew it easily this way.
I personally use it quite often to restrict searches for engines that otherwise only support OpenSearch for their entire catalog.
For example, I have seperate `@ma` for manga and `@an` for anime via MAL, where by default (IIRC) they only have the combined "search everything" advertised through OpenSearch.
I also use it to search individual boards on foolfuuka archive sites, as the default OpenSearch advertisement suffers the same issue as MAL where it's only for searching every board on the site at once.
I'm sure I have even more examples on my browser, I actually use it all the time so I'm glad this is getting mainlined rather than nixed. I use Librewolf, so they already had this enabled by tweaking the corresponding about:config setting themselves.
Yeah to be honest I read the bugzilla thread but I didn't find it very clear. It's really meant for the in-crowd. Which makes sense for a bugtracker but not for a HN article.
And yes good point. I can also imagine using it for stuff that's not strictly a search engine as such.
> High functioning male social clubs generally have implicit rules, like "straight acting gay guys are fine but don't make it weird" or "no weird lefties".
I think that whole conformation thing is why they don't work. Nobody wants to hang out with people pretending to be someone else so they fit in. Any social connection you make is then fake too.
> Which means high status guys, the kind of guys who are trend setters, tend to stay away.
The board games types can also be high status trend setters, just not in your circle. That's fine though. Nothing wrong with seeking out people that are like yourself.
But there's plenty of places where you can find what it sounds like you're looking for. Like sports bars. Won't find the board games types there and not many women either.
Huh? "Conformist" is the most common type of person on Earth and conformists prefer hanging out with conformists. Social clubs are entirely a conformist phenomena, almost by definition. All those Elk clubs and bowling clubs and so on were chock full of conformists.
I don't agree. If you choose the right club you don't have to conform and you can just be yourself. Especially in the cities there's a scene for everyone. In the countryside it's slim pickings of course so you do have to conform.
Maybe that's one of the reason people in small towns are so different, the social dynamic is stricter because there's just not enough people around to form groups of people that are different. City people like me, if we don't fit in we'll just find another place to go so we're more aligned. We can choose our community because a city isn't a community, it's a big box of lots of different ones. If you live in a small town you don't get to do that (not as much anyway)
But the idea that there's no community there at all is not correct. I live in a big city but I keep running into the same people :)
Ps I don't think one is better than the other, just more suitable to some people than others. I'm a city guy and I moved to the town of my girlfriend for a decade but I couldn't stick it. She couldn't stick the city with me, not for more than a holiday. That's ok too. Just meant we had to go our separate ways.
Edit: But yes when I said "Nobody wants to conform" I was just talking about myself. I guess there are people that want that. Thanks for the correction.
It's a subtle matter. Seeking acceptance and validation subconsciously and being willing to conform to get those is probably a much more common pattern compared to a conscious desire to compromise and to conform.
Really? Back in my day, there were all sorts of nerd groups, which were often plagued with horrible social dysfunction.[0] People just muddled through.
That's not too geek-specific. I've seen that at almost all kinds of volunteer-driven organisations. Like a local radio station, student fraternity, backpacker houses. Disorganisation, feuding, usually because several people put more effort than the rest but feel like they also are more important than the rest. Coupled with usually not very strictly defined roles and responsibilities this is a recipe for discord and fighting.
I've seen it at typical geek places too like makerspaces but it's certainly not limited to the geek communities.
At the groups I'm part of the vast majority is neurodivergent but things go really smoothly. We rarely have incidents.
They're not neurodivergent-focused groups but there's just a (much) larger percentage of us attracted to events that stray a bit further from the mainstream.
In my experience people do not just muddle through anymore. I don't want to speculate about why that might be, but I have seen so many weird behaviour explosions at these sorts of events myself that leads to people being ostracized
I could speculate, it was partially because of that blog post. The more social nerds are encouraged to cast out the antisocial stinky ones. Instead of a whisper campaign, there's a social media ejection.
(Nerds in particular have been lured into fake socializing with fake friends on a discord or something. I've seen this where someone disappears and it's like "i dunno, maybe he got busy with life". None of their 'friends' really care if he's dead or not, because if he really did get "get busy", that is an indictment on them.)
> there's just not enough people around to form groups of people that are different.
That's certainly part of it, although I think the bigger factor is that people who are different just leave. Small towns are conformist because of survivorship bias.
1) Real friends certainly let you be real. And the scene I frequent deeply frowns on unconsensual photography. Most of the events I go to they sticker all the cameras. I love that. I go there for the people not for Instagram.
That's not to say nobody takes pics but they do it in a quiet corner so they don't catch anyone by mistake. It makes it very respectful. The stickers are just a reminder so you don't just start flicking away when you're drunk. It makes everyone feel safer and more genuine.
2) I guess but nothing some quick ice breaking games won't fix
3) In a small town there's much more familiarity yes. But also a much deeper sense of being watched and judged. I can't live with that. Even the small city I lived in was too small for me. Everyone knows everyone's business and constantly gossip behind your back.
The nice thing in a big city is meeting new people and finding new places. And the variety. In a small town there's a lot of pressure to conform, eg often you're an outcast if you're not religious. I don't think they're bad but there's little acceptance of people who are different. So what do you do? Pretend. That's not real connection.
In a big city you can really be yourself because there's always others that are like you and you can meet them in like-minded places or events. And you can make real ties there. And even find out about other communities you might fit in.
I really hate going to male-exclusive places by the way. There's very few men I have a deep connection with (I'm male) because the whole BS thing that it's frowned upon to talk about feelings. "Men's weekends" just end up with too much beer, macho talk, shooting the shit and hanging in front of the TV watching boring sports or crappy porn. Nothing serious, fun or enlightening. That's my experience with those anyway. I find that exhausting and I always excuse myself from them now. I used to try to fit in but the others would know I hated it anyway so it was awkward.
I have much deeper relationships with lady friends. They're more open and less judgemental in general. I feel safer around them. So mixed events are a must for me.
Many. And I don't even mind but it's usually really bad porn. Also it's not like anyone does anything, they just sit there and pretend they're not interested or ignore it. It's really weird. Also a kind of macho thing I think. Anyway in Holland this is not uncommon.
I have nothing against pornography at all but I do have a minimum quality :) I like the more stylish stuff.
Though I prefer watching people in real life, I'm lucky to have some other friends who are into that too.
It sounds like when you say you go to "male specific" events, you have a very specific set of male-specific events in mind. I suspect the experiences you described earlier are less due to the male-ness of those present, and more simply a function of the social events you tend to go to?
Like, not that that's a bad thing- live your lice how you want- but complaints of "my complaint with all-male gatherings is that the collectively viewed pornography is of insufficiently high quality" are not necessarily the most relatable.
That wasn't the only complaint :) It was more the overall uninterested feel and senseless macho behavior that maybe they have pent up being with the wife or something and they need to act the tool to justify their manhood. Or to forget the day to day family routine or something. Most of the time they sit around bored.
What my friends don't know is that I'm pretty 'liberal'. And I don't have this pent up need for horseplay. I get enough of the more real kind of play :) I just felt so out of sync. It's just so senseless and not enjoyable at all.
The kind of event is pretty common in Holland. It's a pretty typical thing for male friends to go off for a weekend once a year or so in a cabin somewhere. I've done it with several different groups.
> In a big city you can really be yourself because there's always others that
> are like you and you can meet them in like-minded places.
> And you can make real ties there.
This is a massive assumption, but maybe 'yourself' is limited to a standard deviation from the accepted mean.
Well yes of course there's a maximum deviation. If you're too different you won't fit in. Not a bad thing because then there's no real point in being there anyway.
That's why you have to pick the communities you engage in so you fit. You don't have to change yourself but you pick the community to suit.
It's not an assumption though. I live in a city of millions and I'm in some communities of only hundreds of people. Which thrive and even have their own places. That's the nice thing, in a city it's easy to have enough scale even to make niche communities thrive.
What's the majority? There's so much difference in people. There's the IT/intellectual worker and there's blue collar workers, there's sport fans and book enthusiasts, there's religious communities and lgbt-friendly ones. All examples of dualities that are common to some degree but don't have so much overlap in interests.
In my experience social settings work a lot better when they're a bit more specific. Like, about something. And there's not really one majority that fits all. In the US even the two major parties are extremely polarised and yet they are about equal in size.
Queer nightclubs - Berghain and FOLD (London) e.g.
Some parties I occasionally go to in London have a “we really really don’t want you to use your phone on the dance floor and will tell you off” policy.
Like te_chris says, they're the more expressive parties. The "embrace different" ones. Not specifically queer in my case but certainly queer friendly.
Not necessarily as extreme as Berghain mind you. But just places and events where people are encouraged to dress or behave less typical.
Even the cosplay community now has signs to always ask before photographing a cosplayer as they might not want to be photographed without their knowledge.
Those clubs still exist in London but they're just for the elite to make shady backroom deals with their rich buddies :)
They're really exclusive and they always have been. You and I would not get in, not now and not in the Victorian days. Even 'new money' is usually not ok. You really have to have gone to the right school and have the right family.
Just to challenge that slightly. There is a range of clubs, some are honestly very easy to get into if you end up there for some work event and talk to at least two people. It's the ability to socialize, and lack of clubs focused on new industries that's made them elusive to the new-money (There isn't a National Software Club for example). I'll also knowledge most would run about £1-2k a quarter which is restrictive (by design) cost.
Hmm yes I've seen men's sheds but they weren't actually men exclusive and very topical to making stuff. A bit like a makerspace but less focused on tech and more on woodworking.
reply