this post was submitted on 16 Mar 2025
64 points (100.0% liked)

TechTakes

1709 readers
73 users here now

Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.

This is not debate club. Unless it’s amusing debate.

For actually-good tech, you want our NotAwfulTech community

founded 2 years ago
MODERATORS
[email protected]
64
The broken search bar is symbiotic with the bullshitting chatbot (awful.systems)
submitted 3 days ago* (last edited 3 days ago) by [email protected] to c/[email protected]
19 comments fedilink hide all child comments
 

The other day I realised something cursed, and maybe it's obvious but if you didn't think of it either, I now have to further ruin the world for you too.

Do you know how Google took a nosedive some three-four years ago when managers decided that retention matters more for engagement than user success and, as this process continued, all the results are now so vague and corporatey as to make many searches downright unusable? The way that your keywords are now only vague suggestions at best?

And do you know how that downward spiral got even worse after "AI" took off, not only because the Internet is now drowning in signal-shaped noise, not only because of the "AI snippets" that I'm told USA folk are forced to see, but because tech companies have bought into their own scam and started to use "AI" technology internally, with the effect of an overnight qualitative downstep in accuracy, speed, and resource usage?

So. imagine what this all looks like for the people who have substituted the search bar by the "AI" chatbot.

You search something in Google, say, "arrow materials designs Amazonian peoples". You only get fluff articles, clickbait news, videogame wikis, and a ton of identical "AI" noise articles barely connected to the keywords. No depth no details no info. Very frustrating experience.

You ask ChatGPT or Google Gemini or Duck.AI, as if it was a person, as if it had any idea what it's saying: What were the arrows of Amazonian cultures made of? What type of designs did they use? Can you compare arrows from different peoples? How did they change over time, are today's arrows different?

The bot happily responds in a wise, knowledgeable tone, weaving fiction into fact and conjecture into truth. Where it doesn't know something it just makes up an answer-shaped string of words. If you use an academese tone it will respond in a convincing pastiche of a journal article, and even link to references, though if you read the references they don't say what they're claimed to say but who ever checks that? And if you speak like a question-and-answer section it will respond like a geography magazine, and if you ask in a casual tone it will chat like your old buddy; like a succubus it will adapt to what you need it to be, all the while draining all the fluids you need to live.

From your point of view you had a great experience. no irrelevant results, no intrusive suggestion boxes, no spam articles; just you and the wise oracle who answered exactly what you wanted. Sometimes the bot says it doesn't know the answer, but you just ask again with different words ("prompt engineering") and a full answer comes. You compare that experience to the broken search bar. "Wow this is so much better!"

And sure, sometimes you find out an answer was fake, but what did you expect, perfection? It's a new technology and already so impressive, soon¹ they will fix the hallucination problem. It's my own dang fault for being lazy and not double-checking, haha, I'll be more careful next time.²
(1: never.)
(2: never.)

Imagine growing up with this. You've never even seen search bars that work. From your point of view, "AI" is just superior. You see some cool youtuber you like make a 45min detailed analysis of why "AI" does not and cannot ever work, and you're confused: it's already useful for me, though?

Like saying Marconi the mafia don already helped with my shop, what do you mean extortion? Mr Marconi is already beneficial to me? Why he even protected me from those thugs...

Meanwhile, from the point of view of the souless ghouls at Google? Engagement was atrocious when we had search bars that worked. People click the top result and are off their merry way, already out of the site. The search bar that doesn't work is a great improvement, it makes them hang around and click many more things for several minutes, number go up, ad opportunities, great success. And Gemini? whoa. So much user engagement out of Gemini. And how will Ublock Origin ever manage to block Gemini ads when we start monetising it by subtly recommending this or that product seamlessly within the answer text...

top 19 comments
sorted by: hot top controversial new old
[–] [email protected] 12 points 3 days ago (1 children)

I think you’re absolutely correct, and this feels to me like the only reason why we’re seeing some of the bizarre shit we’ve been keeping an eye on:

  • several old forums, all of which are unique high-quality data sources, are being polluted by their own admins with backdated LLM-generated answers. this destroys that forum as a trustworthy data source and removes it as competition for the LLM that already scraped the forum — and, as a bonus, it also makes training a future LLM on that data source utterly impractical without risking model collapse.
  • Wikipedia refuses to compromise on quality in general, so it’s under increasing political pressure to change. the game here is to shut down or pollute the original data source by any means necessary, so that the only way to access that data becomes an LLM. the people behind the AI startups are experts at creating monopolies, and shutting down a world-class data source like Wikipedia or making it otherwise unusable would guarantee a monopoly position for them.
[–] [email protected] 5 points 2 days ago (1 children)

several old forums, [...] are being polluted by their own admins with backdated LLM-generated answers.

I've only heard about one specific physics forum. Are you telling me more than one person had this same idiotic idea?

[–] [email protected] 4 points 2 days ago

I vaguely remember that one of the articles talking about the physics forum mentioned it happening elsewhere, but I haven’t dug into it myself. it might just be one or two shitty admins doing this, but I suspect (without evidence, I just can’t think of another reason to do it) there’s some party offering a financial incentive for them to go back and fuck up their old forums

[–] [email protected] 26 points 3 days ago (1 children)

Imagine growing up with this.

I know this isn't your central point, but this is the scariest part to me. They're never going to convince people like us that this is a superior way to communicate with tech because we know better. Google, Microsoft, Meta, and Amazon all know this. They don't have to convince us. They just have to make it the new norm.

I used to mod a good-sized gaming discord server, and I saw so many young people say things like "Well, what do you expect? To watch someone's YouTube content for free?" They don't remember when YouTube was just a bunch of us randos posting fun stuff, and the most popular channels were shot in 360p on someone's flip phone in a basement, and the site itself had some banner ads at most. Even the very first video shot by Jawed as an example practically fits this description. Now it's become normal for it to be a monetized career, professionally shot and produced to be pushed to the top of the algorithm. I hate it, but all I have to do is take one look at YouTube's home page to determine such is the case.

Search engines used to be things we managed by entering specific commands and conventions long forgotten, but if I try to explain to someone the benefits of such a system, they'd think I'm just some nostalgic boomer.

Forums used to be anonymous, and many still are, but not the most popular ones. It's become normal to give at a bare minimum your personal location and phone number.

And it will unfortunately become normal for "AI" to hallucinate bullshit at us and accept the answer as "good enough" or whatever. Maybe it already has.

[–] [email protected] 16 points 3 days ago

It is admittedly only tangential here, but it recently occurred to me that at school, there are usually no demerit points for wrong answers. You can therefore - to some extent - “game” the system by doing as much guesswork as possible. However, my work is related to law and accounting, where wrong answers - of course - can have disastrous consequences. That's why I'm always alarmed when young coworkers confidently use chatbots whenever they are unable to answer a question by themselves. I guess in such moments, they are just treating their job like a school assignment. I can well imagine that this will only get worse in the future, for the reasons described here.

[–] [email protected] 16 points 3 days ago (1 children)

drowning in signal-shaped noise

Ooh, I love that phrasing, wonderful :D

But yeah, it's an interesting point... It's weird to think that "good search" may just be permanently gone. Somehow I thought that it would come back eventually... but maybe it won't? Wouldn't be the first time a good thing just disappears from the internet...

[–] [email protected] 10 points 3 days ago (2 children)

If good search does come back, it'll likely require heavy human curation to keep LLM noise as low as humanly possible. Automated methods can be easily SEO'd to death, but human curation's gonna be rather tough to game.

[–] [email protected] 9 points 3 days ago (1 children)

I hate programming but if I wanted to waste any time programming stuff my idea would be something akin to Yahoo! Directory from before Google, or del.icio.us from the 2000s, but distributed, and tied to a PGP-like web of trust system.

You search for a topic, you get links saved with that tag by people you personally validated and trust first, and then by people they trust, and people you don't know but added as probably fine, and so on. Dunno how doable it would be to do something like this.

[–] [email protected] 3 points 2 days ago

Great idea, but i can see a lot of bias ripping through that

[–] [email protected] 8 points 3 days ago (2 children)

I believe I read somewhere that Wikimedia was some time ago (a decade ago? who knows and no point in trying to search for the article) exploring the idea of a human curated search engine. Perhaps an idea who's time has come.

[–] [email protected] 7 points 3 days ago (1 children)

A human-curated search engine would likely be easy to sell as well - the obvious approach to marketing it would be to bring attention to the human-curation involved, and claim no algorithms are involved in determining search results. This is arguably bullshit - you'll need an algorithm to sort the search results at minimum - but it'd evoke the idea that the engine is giving customers what they want, and not what someone else wants.

Additionally, you can pull out the somewhat old standby of claiming the search engine to be AI-free - with LLMs and slop generators defining how the public views AI, presenting yourself as a bulwark against the slop-nami will be an easy marketing win.

(Sidenote: Between the ever-growing backlash against AI, boiling resentment against Silicon Valley, and the fact I found this an easy sell, I suspect this idea's time has indeed come.)

[–] [email protected] 4 points 3 days ago

I keep stopping myself from doing this exact project, with the fediverse as the curation source, several times. I’ve talked about this before, but interestingly Postgres’ full-text search is effectively the complete core of a search engine, minus what you’d need for crawling and ranking (which is where curation and a bit of scripting would come in)

other than resources and time, one big open question is how to do this kind of thing as a positive part of the fediverse — to not make the same mistake that a bunch of techbros already have and index the fediverse without consent. how does one make the curation process simultaneously consensual and also automated enough that it can be reasonably ruggedized against abuse?

[–] [email protected] 4 points 3 days ago

It was a Jimmy Wales personal project, not an official WIkimedia project, and no, it didn't go anywhere

[–] [email protected] 22 points 3 days ago

One day I'll tell my grandkids that the internet used to look more like a library--mostly good information with an easy way to find what you're looking for--instead of the grocery store magazine rack it is today.

I doubt they'll believe me.

[–] [email protected] 6 points 3 days ago (1 children)

The way that your keywords are now only vague suggestions at best?

Duckduckgo does that for me, ignores a word even if there are only two. I then use the !g bang for Google, they still work like they always did here (Swiss).

[–] [email protected] 5 points 2 days ago (1 children)

It is extremely annoying the frequency with which i have to use !g to find the specific thing i'm thinking about. I'll still use ddg because the bangs are great, but as a search engine it might actually be worse than fucking google.

[–] [email protected] 2 points 2 days ago (2 children)

Btw, why do search engines never expose a API, in which you can filter and sort how you're happy?

[–] [email protected] 4 points 1 day ago

because they sell that! e.g. Kagi resells Google's actual-search-results API

[–] [email protected] 5 points 2 days ago

@MonkderVierte The answer is obvious—if they did that, they couldn't force-feed you advertisements or bamboozle you with promoted (paid for) content (hint: filtering/sorting is exactly what they DON'T want their users to do).