"We're at the end of a vast, multi-faceted con of internet users"
March 11, 2024 11:52 AM   Subscribe

Are We Watching the Internet Die? (Edward Zitron's 'Where's Your Ed At' newsletter)
posted by box (67 comments total) 45 users marked this as a favorite
 
Terrifying that 2023 will become the last year where all these models can scrape things without hoovering up regurgitated AI stuff en mass. These AI models and future potential ones are basically all gonig to be forever tainted at their core of what shit was said most in 2023 and the years before, a dataset itself filled with countless malicious actors, astro-turfers, propagandists, trolls, hate groups and more. Think of the worst people you've seen writing comments online the past few years and know they are permanently enshrined inside these horrible programs.

They used to warn us about what we wrote online because it could come back to haunt you. This has never been more true with everything everyone has written online exist as barrow-wights and wretched phantoms inside dataset mulch. Things you were embarrassed by and grew from, learned to think, speak, or act better -- the LLM have no real sense of growing and learning from past mistakes, the sins of the past and present are the same to them when it comes to spewing output.
posted by GoblinHoney at 12:17 PM on March 11 [51 favorites]


Summary: We’re all going to be replaced by generative AI.

Even at smaller sites like Metafilter, humans.
posted by Tell Me No Lies at 12:18 PM on March 11 [2 favorites]


It makes it sound like the end result will be that everyone's entire internet experience will be mediated through AIs so that you don't really access the internet at all, just articles created by AI and AI-generated summaries of what is out there. That doesn't sound particularly enjoyable or healthy.
posted by any portmanteau in a storm at 12:19 PM on March 11 [22 favorites]


Like a plant turning to capture sunlight, parts of the internet have already twisted toward the satisfaction of algorithms.

what a great, chilling line.
posted by ZaphodB at 12:33 PM on March 11 [41 favorites]


Needs the "Betteridge's Law" tag.
posted by Greg_Ace at 12:33 PM on March 11 [7 favorites]


The Dark Forest of the Internet by Yancey Strickler, and a video by Kyle Hill based on it
posted by tigrrrlily at 12:34 PM on March 11 [6 favorites]


Their ideal situation isn't one where you visit distinct websites with content created by human beings, but a return to the dark ages of the internet where most traffic ran through a series of heavily-curated portals operated by a few select companies

As opposed to the dark age where most of what people see is from some kind of social media site? We're on stacked dark ages now.

The internet has always had a findability program. There's countless sites you would have enjoyed that died long before they met your eyes. Even Google, at its best, could only show you things you were immediately looking for. Algorithms promised to bring you into contact with things you never knew you wanted, but in fact, as Youtube's algorithm readily shows every day, can only bring you things related to what you were last watching, or sometimes, things related to what you last searched for on Google. It's terrible at finding new kinds of things you'd like.

The answer, I think, is link blogs. Like Metafilter at its best, where not everything might be interesting to you personally, but at least it's being brought to you by someone who cares about what they've found.
posted by JHarris at 12:36 PM on March 11 [75 favorites]


Well, time to fire up the terminal emulator and start telnetting into BBSs again.
posted by fimbulvetr at 12:39 PM on March 11 [21 favorites]


I noticed a couple of months ago that I was seeing listings on eBay that had ample seller text...but absolutely useless AI seller text full of business puffery. Rather than saying something like "these vintage boots were benchmade in England per their label; size UK 7.5 but fits narrow; leather still soft and pliable despite scuff on left heel" they all now say "stylish boots will make you fashionable in every season with their delightful leather and entrancing heels you can wear them all year round with happiness". It's always two to four sentences of sheer garbage; I'd rather get no information at all.
posted by Frowner at 12:40 PM on March 11 [37 favorites]


Perhaps one day we'll be scraping old vBulletin forums that haven't yet been exposed to AI, much like finding pre-nuclear age steel that is prized for it's lack of contamination.
posted by msbutah at 12:41 PM on March 11 [51 favorites]


In one or two generations, kids will apply the equivalent of 'cringe', 'ok zoomer', etc., to the olds and their obsession with the 'internet'.
And they will go outside and run and play and get high and fuck and all will be well with the world again.
Except for the climate and fascism and all that stuff.
posted by signal at 12:51 PM on March 11 [40 favorites]


Yes. And the climate and the fascists will do a great job of ensuring only certain kinds of kids who look a certain way and live in certain countries will be able or allowed to go outside to play.

But for some, “all” will be well with “the world” again.
posted by armoir from antproof case at 1:16 PM on March 11 [7 favorites]


I stopped shopping at Amazon a decade and a half ago because I like having bookstores near me, but the year before last I needed to buy something that was only available through them.

Out of curiosity I browsed around the site for a bit, or rather tried to browse, and found it remarkably difficult. Their whole thing used to be discoverability, and that seemed to have gone by the wayside. Whatever innovations they’d made I algorithmic sorting and searching didn’t make it easier to find the unexpectedly intriguing.

Bookstores are community hubs and important facets of local intellectual infrastructure, sure, but now they’re also really crucial in finding interesting things to read. Especially as there’s a real dearth of book reviews too, compared to the heyday of blogs, and before that newspapers and magazines.

I had realized a long time ago that all these cornerstones of culture were under threat by modern capitalism, but I assumed that big tech companies would get good at matching consumers to stuff they’d want to buy. But no, it’s entirely a volume business, it seems.

Bookstores and libraries are nodes in a network of exchanging ideas that is parallel to the internet, and is good at things the internet isn’t good at, and it’ll be increasingly important to keep those around as the internet drowns in text goo.
posted by Kattullus at 1:19 PM on March 11 [37 favorites]


.
posted by chronkite at 1:33 PM on March 11 [5 favorites]


Just yesterday I came across a question posted to stackexchange (not going to link because I don't want to single out that user) which was in the form of an AI prompt. It was obviously a prompt, because instead of asking "How do I...." or "is it okay to do this?" it started with "You have been asked to design a .... given the following details".

Even though the mods there had done the right thing and downvoted it, seeing a blatant prompt get posted made me a little angry. Was this someone who naively posted an AI prompt thinking that stack exchange was another LLM-based chatbot? Is this just the way people are going to communicate ideas on the internet from now on? In passive voiced descriptions of scenarios?
posted by RonButNotStupid at 1:40 PM on March 11 [13 favorites]


And they will go outside and run and play and get high and fuck and all will be well with the world again.

Do you WANT to be sent out to clean?
posted by The Bellman at 1:41 PM on March 11 [20 favorites]


I think "The Harvest" has been going on for a decade or more, and it has more to do with apps and walled gardens and anti-competitive behavior than AI. Maybe in 2010 the VCs were hot on API mashups and such, but that tune changed pretty quickly when Facebook started buying apps for $bns.
posted by credulous at 1:44 PM on March 11 [2 favorites]


> much like finding pre-nuclear age steel that is prized for it's lack of contamination.

Low-background memes.
posted by genpfault at 2:21 PM on March 11 [11 favorites]


I went down a little rabbithole trying to read this article because the links looked weird. I looked at the source code and the author is very deliberately creating an effect by using <u> tags to underline every link, with the <a> tag creating it's own underline, so you get a double underline effect on every link.

Which seems like a lot of work for an effect you can get using a CSS definition on your link tags. Especially so when the double effect is occasionally not lining up with the links due to mismatch in location of the <u> and <a> tags.

Then I thought about it some more and considered that the fact I know how to do this (look at source code and decipher what is or is not going on in the HTML and CSS) means that I remember the internet when it sucked but was wide open, while it was wonderful AND wide open, and now have watched it decline into suck and corporate lockdown.

Man I feel old.
posted by caution live frogs at 2:36 PM on March 11 [33 favorites]


From the article:
There are simply too many users, too many websites and too many content providers to manually organize and curate the contents of the internet, making algorithms necessary for platforms to provide a service.
This is an admission of defeat, and possibly premature. What's most wrong about today's internet, and the mindless deployment of sh!t-fed generative AI, is precisely the abdication of human responsibility and involvement in curation, in favour of algorithms. Best single example of this is the improvement in search results you get by including "reddit" in most Google searches; you're limiting results to human-generated and usually moderated content, which often finds more relevant results. (the reddit issues notwithstanding)

We could be ready for a new economic model. I would consider paying something like $10 a month for SEO- and sponsor-free search, and for well-managed and curated newsfeeds. I already support one newssite, and have tossed a few pesos to Metafilter.
posted by Artful Codger at 2:38 PM on March 11 [13 favorites]


It's kind of like the tool manufacturers in the UK and US that used to make high quality planes and chisels but then got bought out by giant corporations that shuttered them, using their respected trademarks to brand poor imitations. The products were marketable because they were cheap and the shape of a tool, rather than having the utility of a tool. If these tools didn’t perform adequately, buyers would assume their skills were lacking and gradually people forgot how to even sharpen a tool, shifting to power tools having replaceable cutters, which changed the sort of work people did.…and that's why we're up to our armpits in river tables and end grain cutting boards.
posted by brachiopod at 2:57 PM on March 11 [18 favorites]


Meanwhile, this place is like the house in Up.
posted by pracowity at 3:13 PM on March 11 [24 favorites]


Why would you want to waste your time on unreliable google search when GPT4 will do it for you, synthesize the best articles into a cohesive answer, and provide citations to validate the work?

Search died last year, it just doesn't know it yet.
posted by Nanukthedog at 3:18 PM on March 11 [4 favorites]


Really though, what we're missing is a way to insert ad revenue into AI.
posted by Nanukthedog at 3:22 PM on March 11 [3 favorites]


These AI models and future potential ones are basically all gonig to be forever tainted at their core of what shit was said most in 2023 and the years before

I think it will be even worse; that by training AIs on AI output it will get more and more nonsensical, like a jpg that has been copied and re-saved and compressed over and over until the artifacts render the image into something else entirely. Any thin connection to reality and truth will eventually be lost.
posted by misskaz at 3:23 PM on March 11 [5 favorites]


Just yesterday I came across a question posted to stackexchange [...] which was in the form of an AI prompt. It was obviously a prompt, because instead of asking "How do I...." or "is it okay to do this?" it started with "You have been asked to design a .... given the following details".

Searching now, I do see questions like this on StackOverflow going back a few years, well before AI. It looks like they're often people wanting help with homework or job interview questions.
posted by smelendez at 3:47 PM on March 11 [1 favorite]


"I would consider paying something like $10 a month for SEO- and sponsor-free search, and for well-managed and curated newsfeeds."

Ah, but would you pay $20? How about $25? How about $15 for the Silver package, where we add just a little SEO in there, from our trusted sponsors, of course. Sorry, now that level costs $25, as well. It's only $30 for the Gold package though, although some say that our curated newsfeeds have gone downhill since we fired the curators right before that stock buyback. The Gold package is still pretty good though, I mean, when you compare it to the Silver package. That shit's unusable at this point. Gold package is only $40 now, if you buy it as part of your yearly contract bundle with Microsoft Office x Netflix Prime x RedditGPT.

You'll still get ads tho.
posted by Rudy_Wiser at 3:49 PM on March 11 [67 favorites]


Terrifying that 2023 will become the last year where all these models can scrape things without hoovering up regurgitated AI stuff en mass.

Whenever I see this line of thought it makes me stop and wonder: does the author understand they are accusing AI researchers of simultaneously having the intelligence to create insanely complex language emulation systems, and the lack of intelligence to not see this obvious and extremely well-publicized problem with several potential solutions? Yes, data after 2023 cannot be blanketly harvested as before; that’s great news! Earlier material should never have been blanketly harvested for a host of reasons - society’s racial bias, consent of the creators, etc. That it happened the way it did was because of Silicon Valley’s particularly toxic exercise of capitalism.

So other methods to obtain the necessary training data will be employed. Some of them will be more ethical (eg curated collections of public domain content akin to Project Gutenberg, filtered to produce responses on the progressive edge of the Overton window) and many of them will be even less (mass exploitation of impoverished people in developing economies as the “Human” in Learning from Human Feedback). Some will be entirely orthogonal to good/bad distinctions like continuous fine-tuning through environmental interaction. We can hope for and encourage the good (right now we’re kinda limited to less-bad, eg Mistral), and we can call out the most dedicated bad actors (OpenAI).

What we can’t do is pretend that the whole thing will fall apart because of model collapse - several teams both corporate and open source will attempt a multitude of solutions. The ones that succeed will continue and the ones that fail will not; that’s not even zero-sum capitalism just the reality of competing for limited attention.

Recognize these reactions for what they are: defense mechanisms borne of fear prompted by a feeling of losing control. And respond in the only healthy and sane way: accept that you never actually had control, and that all things are impermanent. And it’s never all bad because that includes stuff like the divine right of kings.
posted by Ryvar at 3:54 PM on March 11 [9 favorites]


Huh. I thought the problem was going to be a paperclip maximizer. I guess it's a content optimizer.
posted by ryoshu at 4:08 PM on March 11 [6 favorites]


We've already made paperclip maximizers. They run on wetware and are called "for profit corporations."
posted by Zalzidrax at 4:13 PM on March 11 [18 favorites]


It's good to know that the invisible hand of the market will save us.
For a while there I thought we were in trouble!
posted by thatwhichfalls at 4:16 PM on March 11 [8 favorites]


What we can’t do is pretend that the whole thing will fall apart because of model collapse

I think what we can do is recognize that model collapse will immiserate us even further once we surrender to these model-makers, and conduct ourselves accordingly. If AI developers were actually interested in solving problems well as opposed to profitably (that is, in the cheapest, shittiest form that they can coerce society into accepting), we wouldn't even be here in the current state of play in 2024.
posted by praemunire at 4:17 PM on March 11 [12 favorites]


The unfolding saga of Reddit's IPO highlights a dilemma in the cyber age: the exploitation of community-driven content for corporate profit. This shift towards commodification raises significant ethical concerns, particularly as generative AI threatens to further dilute the authenticity of online spaces.

It's a critical moment to reassess the balance between profit and the preservation of the internet as a platform for actually human expression and collaboration.
posted by chaz at 4:19 PM on March 11 [7 favorites]


Really though, what we're missing is a way to insert ad revenue into AI.

what if we are the ad and revenue.
posted by clavdivs at 4:30 PM on March 11 [7 favorites]


Why would you want to waste your time on unreliable google search when GPT4 will do it for you, synthesize the best articles into a cohesive answer, and provide citations to validate the work?

I don't know if you've ever tried to check a citation from ChatGPT, but much like the text itself it's plausible-sounding crap. Invented titles attached to a leading author in the field, using a completely unrelated DOI. Fake legal case citations.

The ability to provide actual citations for information requires technological developments in the opposite direction of everything these models are. You can't unsupervised learn your way to that kind of information hierarchy.
posted by irrediated at 4:54 PM on March 11 [10 favorites]


The unfolding saga of Reddit's IPO highlights a dilemma in the cyber age: the exploitation of community-driven content for corporate profit.

Why is this a problem again? People get a thriving community but if someone makes a buck off it the well is poisoned and everyone goes home?

Reddit was not shy about declaring it was a company and that it intended to make a profit one day. So they provided free services for 20 years. One day a (frankly, very modest) bill shows up for the users and they seem to be deeply surprised and offended.
posted by Tell Me No Lies at 5:08 PM on March 11 [2 favorites]


We've already made paperclip maximizers. They run on wetware and are called "for profit corporations."

This is not an uncommon suggestion, to the point where an AI safety researcher has made a video addressing it.
posted by tigrrrlily at 5:10 PM on March 11


I swear this latest AI / LLM hype cycle is the 21st century version of Chinese peasants throwing all their iron possessions into improvised backyard blast furnaces, with VCs leading the way in place of Mao Zedong. All of these layoffs to replace creatives with AI, and eventually we’ll realize the output is nothing but useless pig iron.
posted by qxntpqbbbqxl at 5:25 PM on March 11 [12 favorites]


Whenever I see this line of thought it makes me stop and wonder: does the author understand they are accusing AI researchers of simultaneously having the intelligence to create insanely complex language emulation systems, and the lack of intelligence to not see this obvious and extremely well-publicized problem with several potential solutions?

It is not a question of their intelligence. It's a question of whether they are motivated to care.
posted by srboisvert at 5:48 PM on March 11 [8 favorites]


The most precious thing on the internet, for me, are the contents of the old PhpBB forums for cars and motorbikes, before about 2008 or so, full of the advice and knowledge of the weird old guys who'd tell you, for example, exactly how to replace the steering stem bearings on a Yamaha SR400, and why a tapered bearing kit is superior to the OEM. That was when all the old guys got onto facebook, moved the advice-asking there. Anything before 2008 is precious.
posted by Fiasco da Gama at 5:51 PM on March 11 [22 favorites]


caution live frogs: i suspect the problem is that Zitron is writing in Word and then pasting into the post interface without cleanup (or with insufficient cleanup); it's well known for leaving artifacts like extraneous tags.
posted by adrienneleigh at 6:05 PM on March 11 [3 favorites]


Reddit is "one of the internet's largest corpuses..."
Who are we to argue with Mr. Huffman, I ask you?
posted by Western Infidels at 6:11 PM on March 11


I work at a venerable, smallish (20 million dollar/year) social services nonprofit. The place has great programming for the community. But the executive end, marketing (me) etc is so backwards and behind the times that it's damn near hilarious. One of the reasons I took the job aside from needing money was I thought I could make a genuine difference.

Anyway, my boss, not the brightest bulb, has been talking about AI and how we have to get into and start using AI. I wish I could link the website to show you what an utter backwards joke it is. There's thousands of these manager types who are just convinced that AI is something that we all need to get into and start using, stat!

Even our logo is from 1968 and looks like it was designed by a bright high school student with Letraset rub-on type and Rapidograph pens. We have no brand. We have no style guide, or budget for anything. We take photos with our iPhones and use those in our annual report!

Yet somehow this guy is convinced that we need to start "using AI."
posted by SoberHighland at 7:18 PM on March 11 [19 favorites]


SoberHighland, although I am fortunate to run for communications for a much more advanced (in the marketing realm) organization, I have had exactly this conversation with board members, some of whom are university faculty. It is baffling. We have a team of skilled writers and an extensive image library. What would using AI do for us?

At the same time, we do use “AI” extensively to do first passes on transcripts, upscale video and images that our partners send us from what I assume are vintage 2005 NOKIA phones, etc.
posted by Just the one swan, actually at 7:34 PM on March 11 [3 favorites]


Whenever I see this line of thought it makes me stop and wonder: does the author understand they are accusing AI researchers of simultaneously having the intelligence to create insanely complex language emulation systems, and the lack of intelligence to not see this obvious and extremely well-publicized problem with several potential solutions?

I mean, not the author, but: I have no problem believing some people are good at a specific thing and bad at examining their pre-existing biases because they assume their skill at that thing protects them from bias. We're talking about folks in an industry that valorizes breaking things because you didn't take the time to think things through before starting your plan.

Motivated thinking is a hell of a drug.
posted by Gygesringtone at 7:56 PM on March 11 [8 favorites]


...all things are impermanent. And it’s never all bad because that includes stuff like the divine right of kings.
posted by Ryvar


The ubiquitous techno-surveillance Panopticon the kings and wannabe kings now have at their disposal kinda complicates that.

It is a change from pre-internet era – the entire previous history of the human race – whose full implications and power has yet to sink into the minds of both the kings and the peasants.

But the kings are grokking onto it faster than the peasants at the moment, and I don't give much for our chances of reversing it if the kings win this round. Certainly not without serious violence.

If we cannot come up with a reliable robust means to distinguish between human and machine generated content, we are going to be royally fucked. It was bad enough during the era when we could be certain that the bad faith actors online were human.

I am not comforted by claims that those who grew up with the internet are going to be better equipped to handle all this. I fear the opposite. Once those who grew up without it die off we will have no living corporate memory of what it was like in the Before Times.

Few things in this world scare me more than that.
posted by Pouteria at 8:33 PM on March 11 [19 favorites]


> If we cannot come up with a reliable robust means to distinguish between human and machine generated content

the most reliable method is to just ask the person you got the link from. or if you found the link on a web page instead of by word of mouth, ask the person who runs that page.

obviously this is not possible if you found the content via a search engine, but search engines are a terrible idea and should only be used as an absolute last resort.

i’ve got this theory that the old Internet still exists, that it is in fact bigger than it was in the 1990s, but that you can’t find it on search engines or social media, only by word of mouth.
posted by bombastic lowercase pronouncements at 10:07 PM on March 11 [10 favorites]


The counter argument with regard to Reddit at least is that Reddit’s revenues went from about $60 million in 2017 to over $800 million last year. That kind of growth will justify their share price in the short term. They lost $90 million last year on $810 million in revenue but forecast over $900 million this year. If they’ve frozen their expenses at last 2023 levels they will be able to show a profit pretty quickly.

Reddit seems to have survived the community drama at the mod strike last year. People mostly stayed and it seems like traffic is the same.
posted by interogative mood at 12:11 AM on March 12 [1 favorite]


So other methods to obtain the necessary training data will be employed.

Buying the right to train on a backend archive of all of Tumblr from 2013-2023, before generative AI was part of the equation.

Oh wait..
posted by subdee at 4:58 AM on March 12


The internet has always had quality issues. It's easy for anyone so inclined to have their say and to get it out there. What's sad is that they have nothing much to say but that's not going to stop them. They insist on sharing. (Sort of like '.' here)

This explains a lot of the noise and turbulence on the internet. Instant reactions when something happens. There's no longer time to reflect on the event. Either be first, be contrarian or add spin to provoke outrage. Great. Everyone loves righteous indignation and outrage.

We're now moved beyond the enshittification of popular platforms, to the taint the internet itself. Almost. The problem we face is still the same - curation. Curation with a very necessary triple dose of scepticism.

How it goes now? I don't think anyone can say at this point. I guess that there will still be pockets of taste dependent quality hiding in plain sight while the gross internet is just going to be gross, like it is now. These pockets are going to be dynamic, with rapid changes and short life cycles because any glimmer of mass interest and success is going to attract the sharks and predators. Maybe something like lobsters but even more locked down.
posted by w.fugawe at 5:15 AM on March 12 [6 favorites]


Why would you want to waste your time on unreliable google search when GPT4 will do it for you, synthesize the best articles into a cohesive answer, and provide citations to validate the work?

New York lawyers sanctioned for using fake ChatGPT cases in legal brief

ChatGPT is making up fake Guardian articles. Here’s how we’re responding

Michael Cohen says he unwittingly sent AI-generated fake legal cases to his attorney

OpenAI Hit With First Defamation Suit Over ChatGPT Hallucination

If you had a research assistant who was great at writing summaries but often just made up citations to support their work, why would you waste time reading what they produce and double-checking it against its sources to make sure it's a valid synthesis of the original material?
posted by RonButNotStupid at 5:39 AM on March 12 [6 favorites]


I get that some people never want to do the heavy lifting of actual research and just want the answers handed to them on a silver technology-enabled platter, but understanding how to take information from multiple sources, evaluate those sources, and synthesize understanding is pretty important.

So much of the mess we're in re: Trump was caused by people trusting the algorithm to give them the information they need to know. Encouraging people to further abdicate responsibility and just run with the answers from ChatGPT is dangerous.
posted by RonButNotStupid at 5:45 AM on March 12 [5 favorites]


I was waiting for you to show up in this thread, bombastic, because it perfectly aligns with your unwavering thesis re: search engines. And I was not disappointed!

But re: the article, yeah, we're at the end of large scale social media. Is that "the internet?" I guess to lots of people. But there is absolutely nothing stopping anyone from putting up their own, non-algorithmic website, or using places like Metafilter. And stuff will be harder to find using Google et al. But the great thing about the capital-I Internet is that nobody really owns it, and anything is still possible.
posted by grumpybear69 at 5:45 AM on March 12 [4 favorites]


"Habsburg AI" is a great phrase.
posted by doctornemo at 6:36 AM on March 12 [6 favorites]


Meanwhile, I'll keep writing my blog. And making my videos.
posted by doctornemo at 6:36 AM on March 12 [5 favorites]


I want to know who's rinsing Discourse for being a closed walled-garden with inadequate search and link-archiving tools.

The 'Dark Forest' is mentioned up top, is Discourse going to monetize the chatter and histories on its platform? Or is it a model of communities with moderation that's not going to be swallowed by the circus of stochastic parrots?
posted by k3ninho at 7:53 AM on March 12


> I was waiting for you to show up in this thread, bombastic, because it perfectly aligns with your unwavering thesis re: search engines. And I was not disappointed!

generally what i do with this sort of thing is i keep doing the thing until i see other people doing the thing and then i step back to see if enough people are doing the thing that new people start doing the thing without me doing anything

i'm just having a hard time doing that "step back" part with this thing because fuck search engines
posted by bombastic lowercase pronouncements at 8:00 AM on March 12 [6 favorites]


irrediated: The ability to provide actual citations for information requires technological developments in the opposite direction of everything these models are. You can't unsupervised learn your way to that kind of information hierarchy.

I don't think that validation is actually that hard a problem, especially when specialized data are available. It doesn't seem so difficult to direct an AI to, for example, validate its own results re case law by then searching a known-good legal database to look up every case it cites. If nothing else, it would catch the cites to nonexistent cases. And it could present its results, together with links and a short precis of the cases it looked up for verification, for human review.

Much of the problem we currently have with AI deployment is still the age-old GARBAGE IN -> GARBAGE OUT issue. You can't use an excavator to dig up an entire vinyard - vines, stakes, soil etc - macerate it all, and expect fine wine at the spigot.
posted by Artful Codger at 10:10 AM on March 12 [1 favorite]


w.fugawe: The internet has always had quality issues. It's easy for anyone so inclined to have their say and to get it out there. What's sad is that they have nothing much to say but that's not going to stop them. They insist on sharing. (Sort of like '.' here)

Unless you’re talking about something different, I’ve always thought the obit post period was a really respectful way to mark that a MeFite is indicating that a person’s death has meaning for them, without taking up space for people who have deeper connections with the the deceased. It’s almost the opposite, in that it means ‘I have nothing to say, and I will stop.’
posted by Kattullus at 10:18 AM on March 12 [11 favorites]


generally what i do with this sort of thing is i keep doing the thing until i see other people doing the thing and then i step back to see if enough people are doing the thing that new people start doing the thing without me doing anything


You just described my approach with cryptocurrency bullshit! Was a time where I felt like the lone nutcase on a soapbox, railing against crypto crap. Not sure if I had anything to do with the overall attitude change, or if it just all became more obvious over time.
posted by notoriety public at 10:23 AM on March 12


interrogative mood: Reddit seems to have survived the community drama at the mod strike last year. People mostly stayed and it seems like traffic is the same.

After never visiting there, I joined Reddit in 2022 just to follow a few technical subjects i was interested in... and when all the sh!t went down last year, I quit it. (and gained a few hours a week of my life back)

I still pop by those few subreddits occasionally, but just to read. There's still the same quantity (or maybe a bit more) of absolute n00bs asking the same damn questions every week: ("What do I buy to get into [thing]?" or "I want to do [thing] at a professional level. How can I get there? My budget is $150."). I when I was a member, I did my share of trying to help even the n00bs, but it grinds you down. Now, THERE'S an application for AI: parse new questions, and if they match one of the already-answered n00b questions, point them to the FAQ or send them a canned or generated answer, and don't post the question. (or enable users to filter out seeing those n00b questions) This would cut the noise way down.
posted by Artful Codger at 11:32 AM on March 12


If we're at the end of the 'social' Internet, I for one am not displeased.
posted by signal at 12:13 PM on March 12


@artful codger “I would consider paying something like $10 a month for SEO- and sponsor-free search, and for well-managed and curated newsfeeds.“

Have you tried kagi.com?
posted by underwater at 6:07 PM on March 12


... kagi looks interesting. Any MeFites using it? If yes, what's your experience with it?
posted by Artful Codger at 7:56 PM on March 12


Not great. I wanted a search engine, I just got another "here's what we think you really meant to search for" engine. Plus lower coverage, so my experience ended up being inferior to using Google with an ad blocker.

But it could be fine for people without specialized search needs, I think.
posted by Not A Thing at 5:48 AM on March 13 [2 favorites]


I signed up for Kagi a month or so ago and use it pretty much exclusively and it's been good so far. I'm not a power searcher, though, so ymmv. They have a free tier you can use to try it out, which is what I did. Liked it enough to sign up for the unlimited plan, which maybe I won't need but I don't mind kicking some extra cash their way for a good project.
posted by snwod at 6:44 AM on March 13 [1 favorite]


w.fugawe: "the taint the internet itself"

Hello new sockpuppet name! Oh, wait, duh!

Metafilter: the taint the internet itself
posted by team lowkey at 12:16 PM on March 13 [2 favorites]


« Older Political demands at the level of biology itself   |   Tintreach (AC/DC in Gaeilge) Newer »


This thread has been archived and is closed to new comments