@bhawthorne you don't have an offline copy of Wikipedia?

@bhawthorne you might be interested in https://kiwix.org/en/ @kiwix

@bhawthorne so many recipes have 5/5 stars now too

@bhawthorne Out of interest, do the Google results contain as much AI noise as those on DuckDuckGo?

I can imagine that removing that noise from search results will become equally important for Google as their fight against SEO spam, and they have the resources to try it

@bhawthorne
No unit mismatches? Converting between Celsius, Fahrenheit, Réaumur, Kelvin, or Rankine may be confusing, even for U.S. folks. For that archive you mentioned, I recommend the Kiwix Offline Reader (via Fdroid, https://f-droid.org/packages/org.kiwix.kiwixmobile/)
Cc @dahie

@bhawthorne @Miniver in those situation I use kagi and filter out everything that has ad trackers. Even before llms, content sites with ads/affiliate links don’t have an incentive to provide quality information.

Ironically, tools like perplexity.ai also help a lot navigating through the nonsense.

@bhawthorne

#DataRot
More #Data is not more #information
#Entropy

@bhawthorne frustratingly, "AI", in this case LLMs, has been used to provide verbose webspam to maximise ad impressions. I know LLMs aren't popular in much of the Fedi, but ChatGPT gives a helpful and concise answer to the original question. There's mileage in LLMs to leapfrog all this web junk we have to wade through today.

@bhawthorne honestly cook books are becoming such a precious resource. Online recipe websites have become such a slurry of SEO spam that they border on the useless.

@bhawthorne That's the weird part. Those lengthy recipe sites were at least human written before by content farms because recipes themselves can't be copywritten but the long ad copy as they reminiscence about how it's a generational family recipe or whatever is usually what is copywritten. AI generated content has consistently shown itself to be really bad at copying the essentially needed parts

@bhawthorne I've started downloading the basic books I need, just in case.

This is a good start towards a decent library:

https://anarchosolarpunk.substack.com/p/offgridinternet

@bhawthorne As others have mentioned, do check out Kiwix - https://kiwix.org/en/ - You can grab an offline copy of Wikipedia and many other websites. (I’ve done so for when Internet access goes down… but you make a really good point to have one at a point in time before the LLMs started generating so much useless text.)

@bhawthorne Strangely enough I’m starting to think this is the best thing to happen to this world in a long time (not the way tech dudes believe, though).

@bhawthorne you don't think wikipedia policies will help it stay reliable in the future?

@bhawthorne - for anything not time sensitive, like recipes, I'm throwing "2013" into my searches.

@bhawthorne

It's worse than that. I remember trying to find out why you can't cook red kidney beans in a slow cooker and got recipes telling me time and temperature.

Those beans, and a few others, have a protein that you need to boil to denature, or they can be toxic. Yet Google was giving me an AI answer that could kill people.

@bhawthorne Our washer stopped working and the steps in the manual didn’t suffice. So I tried finding some DIY repair instructions. Kept the last one alive for over a decade after getting it second-hand. And it’s all just garbage. No instructions that could actually be followed, just crap that SOUNDS like instructions. Even YouTube is full of shitty generated text cards with vaguely related images. So I guess I will just throw away my five year old washer.

@bhawthorne Currently packing up my house for a refurb and I am so reluctant to part with *any* books right now. Remembering the time before the internet is a powerful incentive not to want to go back to that siloed life. Open access to knowledge is the ultimate democracy.

@bhawthorne
This was, in fact, the whole purpose behind "AI" in the first place. Can't have the plebs getting their grimy mitts on too much truth, ya know. Better to gum up the works with a bunch of auto-generated nonsense and then sell curated access to accurate information to the rich...imho.

@bhawthorne just ask your friendly #artificialintelligence for recipes and such. Works sorta okay.

@bhawthorne Don't tell Jason. He'll see it as an opportunity.

https://en.wikipedia.org/wiki/Mahalo.com

@bhawthorne I hate how right you are.

@bhawthorne
What was the correct answer?

@bhawthorne This is what SEO folks call the “truth range”

In order to rank high on SERPs, you want to be right in the middle of the range, *not* the most accurate 😭

@bhawthorne It's really too bad that search engines don't de-rank sites like those, but people click those links a lot. What a mess,

@bhawthorne what’s wrong with 2023 Wikipedia?

@bhawthorne honestly - society can just announce 2023 as the year search had been enshittified to oblivion. people are going to start going back to libraries to find shit - swear to god you nailed it. #ifmacaquescouldtalk

@bhawthorne IMO the most reliable source will still be Wikipedia. The reason is, that the content is generated by humans. If someone adds content without a reliable source reference, it’s very likely, that it will be removed within minutes by other editors.

@bhawthorne And people wonder why I still buy old reference books and cookbooks. (I guess some don’t wonder about it, because so many people here say they do the same!)

I am definitely seeing an uptick in LLM-generated advice that could harm people in my regular searches. I like the idea of storing an old copy of Wikipedia. It could be really useful.

I am actually on the lookout for an entire set of encyclopedias rn.

@bhawthorne

I've been very pleased with kagi.com search. Just did a test search for your chestnuts temperature and the top three sites appeared to be legit food resources. Kagi isn't free ($5/month) but sooooo totally worth it.

@bhawthorne is there a way to know if a site was not written by a human? Especially when searching for tech support, I see oddly phrased language that could just be a translation from another language

@bhawthorne I looked up how to calibrate a Frigidaire oven. A single AI generated website article had three different methods, all incorrect. My fave was the one that had me unplug the range, adjust something in the back, then check the temp again… without plugging it back in.

@bhawthorne download the 2022 Wikipedia, sure, but now a better time than ever to start editing and curating wikipedia!

The community of humans there are not perfect, but still puts an effective guard against generated crap vaguely similar to mastodon.

@bhawthorne
It's tempting to think that the whole LLM thing is an attempt to monetise the Wikipedia model.

@bhawthorne on the bright side, Ai might help libraries recover 😅

@bhawthorne

The Web is turning into the Library of Babel.

"The books contain every possible ordering of just 25 basic characters (22 letters, the period, the comma, and space). Though the vast majority of the books in this universe are pure gibberish, the library also must contain, somewhere, every coherent book ever written"

https://en.m.wikipedia.org/wiki/The_Library_of_Babel

@szescstopni

@bhawthorne in a growing digital world humanity seems to hold on to the things which are physical representations, to document that they was ever really here. where media can be edited so freely can you be sure of anything? humans cling on to the material possessions to validate themselves which seems futile in practice but in application they have to have faith in something or be lost & drown in the vast sea of oblivion

@bhawthorne And it will get worse and worse, as new generations LLMs get trained on online text produced by previous generations of LLMs, amplifying the errors and nonsense.

@bhawthorne I'm not sure this a question with a 'right' answer as (looking at the answers I see) you can use a higher temperature and cook them for less time.

So my top result (bbc.co.uk) says pre-heat to 200°C and roast for 5 minutes while most (presumably US) suggest 350°F (175-180°C) and about 10 minutes.
Or you can lower the temperature and cook for longer.

@bhawthorne Yes! The other day I ditched the idea of making cheese scones when the same kind of thing happened with me, also with DuckDuckGo, which has changed :(

@bhawthorne
Websites for different programming languages can now be found that are written by AI and posted for whatever reason. You have to be careful that you're getting an authentic site that will give an accurate answer to whatever programming problem you're having.

@bhawthorne In the meantime, we need to filter on date.

@bhawthorne
Remember the original Yahoo. It was actually an index not just a keyword search. And remember when there were print yellow pages directories of every website on the Internet. Keep your own index/bookmarks of trusted sources.

@bhawthorne

Are you using Google for your DuckDuckGo searching? If so, might as well use Google. I switched my DDG search engine to use DDG, not Google (or Bing, which was another option).

@bhawthorne If everybody used ad blockers then these sites wouldn't be rewarded for behaving this way.

@bhawthorne Quora (which I rarely use anyway) now pushed a ChatGPT response above any actual answers, and now actively hides answers from actual people so that's fun.

@bhawthorne

Just means we need to set up systems for automatic source tracing.

Article doesn't source its numbers? Browser extension should flag it.

@bhawthorne
And it's not like there is a song for hazelnuts. I mean a least we know chestnuts roast on an open fire. Easy peasy.

@bhawthorne Not only recipes, but workout information too. Most search results are garbage. I still trust exrx.net, it looks like a website from 15 years ago (just like it did 15 years ago) and doesn't peddle supplements.

@bhawthorne this seems to be the ideal use case for https://search.marginalia.nu

It rejects websites with too much advertising or JavaScript. Leaving just passion pages in its index.

Testing your case found a real person giving a real temperature and time for roasted hazelnut in the first result.

And it was readable too!

@bhawthorne For questions where the answer doesn't change over time, I would recommend limiting your search results to, say, before 2018.
Granted, this is merely a bandaid solution over an ongoing problem, but it does help, and at least Google and DuckDuckGo have that functionality.

@bhawthorne

I'm wondering, is this part of #enshittification too, @pluralistic ?

@bhawthorne seems like a block list of machine-generated domains would be a very useful thing to have to improve the speed and quality of online searches

@bhawthorne Welcome... to the INSHITIFICATIN!!! *poorly plays the Jurassic Park theme on recorder*

@bhawthorne You can grab a packaged version over at https://library.kiwix.org/#lang=eng&category=wikipedia

@bhawthorne

Gonna break out my Encyclopedia Britannica DVD 2005 edition.

@bhawthorne
Do mine in microwave 😉

@bhawthorne So I should hold on to my 1972 Joy Of Cooking ?

@bhawthorne @drwho i’ve been thinking about this too, getting the Wikipedia snapshot from 2022 seems right.

@bhawthorne Chatbot might deliver answers for a couple of month, and then then they will be useless too, since they drink from the well they poisoned.

@bhawthorne

Thanks to "AI" Bullshit Generating Plagiarism Engines, we now have the world's ignorance at our fingertips, masquerading as knowledge.

#AI

@bhawthorne At this point I might do that search, but also just look for websites I recognize, or simply go to them directly and search there. Filtering out these crappy websites must either be really difficult, or they don't care to bother.

@bhawthorne
I quit using DuckDuckGo because of this. Every time I tried to look something up it was at least two full pages of garbage sites which were often inaccurate or conflicting at best. Incredibly disappointing.

As for recipes, there is this site. Not sure what else* you’re looking for or if they’ll have it. https://nobullshitrecipes.com/

Recipes without the tragic/nostalgic backstory behind every one.

@bhawthorne When I moved, I donated a lot of books to the library. That Rodale book was one of them. Foolish me.

@bhawthorne sounds like #analog becoming the new gold standard. 🤦‍♂️😔

@bhawthorne i find the same even with peoples blogs sometimes. books will likely be all the rage again

@bhawthorne Can I ask what search terms you used? I tried "hazelnut roasting temperature" just now and the top results include some stuff that could well be AI generated but also actual websites like BBC food and wikiHow. And basically all of them say to preheat the oven to either 175 or 180 °C.

@bhawthorne
The renaissance of that ancient database technology — books. 📚 🙂

@bhawthorne I wonder if my copy of Encarta still works...

@bhawthorne

This could also be an opportunity to better know who has good info. I’ll probably be leaning harder on those blogs and sites that are obviously real people that you know or can grow to trust through interaction.

@bhawthorne I used to think, I know how to word a websearch in a way that will show me the exact thing I am looking for. Lately Bing started throwing random porn at me for absolutely innocent searches and on "moderate". It's not even just websites, the whole net is going to shits. Everything is fucked up, search engines don't do shit anymore. I don't know how they broke it so badly.

@bhawthorne Yep. I download the offline version of Wikipedia every couple years and load it onto my ancient e-reader. (First-gen “Kindle Keyboard” whose cellular access stopped working years ago.)

@bhawthorne we reached the local maximum two quarters ago. It's downhill from here

@bhawthorne 😿

@bhawthorne Auto-generated listicles has been a growing problem for some time. With AI it is metastasizing rapidly.

Are we going to have to documents out with a date stamp to avoid the creation of fake facts?

@bhawthorne recipes, datasheets, lyrics and various "how to" are worst offenders...

@bhawthorne Googles knowledge panes were supposed to be based on knowledge graphs, so some standards of 'truth',...if DDG is just crawling a giant unregulated informational flea market containing a lot of machine generated content this is not surprising. Google shouldn't have a monopoly on knowledge graphs but KGs ARE necessary to validate shitty AI data.

@bhawthorne sturgeon's Law states that 90% of everything is crap. he underestimated the modern web.

@bhawthorne this sounds crazy but I’m starting to dabble with Kagi as a search engine. The free tier has about convinced me that it’s worth it.

@bhawthorne Isn’t this sort of a moderation/curation issue? The machine-generated websites are spam. They need to be identified, reported and blocked. We’ve all gotten used to farming out this task of blocking spammy websites to Google (or DDG or Bing); maybe we need a Wikipedia-like collective approach to it now.

@bhawthorne interestingly i found relavant links immediately with kagi.com, it also gave me dates when the article was made, some of which are pre 2023 :>

@bhawthorne

IT related searches are nearly as bad. Recipe sites have been a shit show for some years and the same is being seen elsewhere. ChatGPT etc (generative AI) are now being used as tools to "enhance" these dubious offerings. Those random musings are known as "hallucinations", which is a side effect of how a word guesser works.

I've almost gone back to using the web like its 1999: bookmarks and trusted sites. I still search a lot but much less, these days.

@bhawthorne I've been thinking we need an open source search engine or maybe a directory (which seems easier to achieve). We're going to have to track for ourselves which websites we find authoritative rather than those with the best SEO.

@bhawthorne

So true; the first few pages of links on any general info search are clearly useless unreliable bot-generated dreck.

For those of us who are not tech savvy,* is there an easy way to download an archive of the 2022 wikipedia?

*is dystechtic a word yet?

@bhawthorne you don't trust nutroastingtemps dot com slash 12-new-hazelnut-temperatures-to-try-in-2023?

@bhawthorne

I now have a print copy of the 2023 World Book encyclopedia.

There are so many threats to the knowledge landscape it isn't funny.

@bhawthorne It's so bad that I don't results for search terms where I *know* there is a website with that info.
For recipes especially the first hits I get are not all AI generated but they're all from big, general sites where anyone can submit recipes, so the quality is hit questionable. Because of this, I started seeking out cooking blogs and Youtube accounts I know are run by actual humans.

@bhawthorne I've given away many books. But I still have The Joy of Cooking. Guess I'm gonna be needing that after all.

@bhawthorne This is why I have 4000 books.

@bhawthorne

The companies that create the tools to generate these kind of sites will have tools to search for higher quality and accurate content.. For a premium.

@bhawthorne Bugger I was thinking of making DDG my search engine but if it's pulling up the same shit as Google, what's the point.

@bhawthorne *gasp* You can't possibly mean people are going to have to resort to analog books?!

@bhawthorne @grrrr_shark Kagi shows some variation in temps, but none of the hits look generated, and several (including the one they use to extract an answer) is from 2021.