“The filter relies on manually curated open-source blocklists, including the ‘nuclear’ list, provided by uBlockOrigin and uBlacklist Huge AI Blocklist,” DuckDuckGo said in a post on X.
I'm not sure the list has been provided by uBlock, looking at their repo[0] I understand that the list is available for uBlockOrigin and uBlacklist, but not made by them.
I recently opened Pinterest after a year or so, and did look for some nice architecture / houses, and all results had the tag 'AI enhanced'. No I don't want AI houses, I want to see real architecture! Even though the tags are there, you can't filter on 'No AI'. So I stopped using it again.
I never understood Pinterest and how it works, but I know that people who DO understand and use Pinterest are basically saying that it’s been completely ruined by AI images.
It was mostly ruined by ads/SEO spam many years before. AI is probably the final nail in the coffin for any of the users who appreciated its original appeal, which was basic image curation from mostly blogs.
I still have no desire to use it, but reading the complaints and explanations of AI wrecking Pinterest has finally given me some understanding of what people do there and why they bother.
Recipes are a sore spot for me. holy smokes is it a seo spam wasteland, I mean as far as our modern search engines go the whole web(statistically) is seo spam wasteland. but recipe sites are a specific particularly bad biome in that wasteland. they all follow the same infernal formula, about 5 paragraphs on how the recipe makes you feel, then the over specific formulation.
I know there are good recipe sites out there, you just will never find them, unless our search engine overlords slip and let one through.
The good web still exists, it is just that the search engines(our gateway to the web) are unable to find it under all the trash. My only solution so far has to been to start manually curating a list of good sites. A big part of the problem as I understand it is the search engines heavily prioritize new content over good content. To the point if I see a date in the current year next to the search result, I instantly reject it as seo garbage. If I were a better person I would sign up for Kagi, as I firmly believe a large part of the solution is to fix the business/customer dynamics, that is, you want to be the customer, not the product. But I am reluctant to sign up for another service so I am stuck shoveling slop to get through the web.
The use case of someone preferring AI generated material vs the real material seems only to strike the algorithm and retention time of users so they always feel a sense of having what they need - despite it not existing.
I miss when you could put `+foo -bar` into a search engine, ctrl-F on the top ten results, and reliably find `foo` on all ten of them and `bar` on none of them.
Nowadays, putting in `-bar` might increase the number of results that are directly about bar.
This bugs me no end. Image search is so challenging to use because the results are a vague interpretation of what you ask for, and don't try to be more descriptive because that's just more words to throw in the mix of results
Yesterday's discussion around Kagi[0] prompted me to poke through some settings there, as it had been a while. They've also added a "Exclude AI generated images" which they describe as "Best-effort removal of AI generated images from search results"
There's a group of people for whom AI images seem to just be a trigger, and they simply never want to see them. So having a filter to remove those images makes sense. I don't really have that issue with AI images, but I often do want to have some way of filtering photographs from illustrations—be they drawings by humans, or images generated by AI. And that could go either way: maybe I want to see what an actual baby peacock looks like in nature sometimes, and maybe I want to see a cute illustration of a baby peacock other times. It would be great to have a filter for 'real' vs 'artificial'. Perhaps ironically, using AI is the best way to make that distinction on my behalf.
If I’m looking for images because I’m curious what a thing really looks like, it’s counterproductive to see AI’s “artist’s conception” of it, which is often anywhere between not particularly accurate and a picture of something else entirely.
A risk is that it will give people a false sense of confidence that they are viewing real content. The only way out of this mess is cryptographic methodss (based on hardware in cameras) that can allow end-users to verify photos as real and then we assume every other photo may be AI.
Cryptographic components in cameras have roots of trust that can be compromised as well. Also, photos can be staged (many famous examples of this).
The only real solution is to build social infrastructure that helps people identify the trustworthiness of a source. Some efforts are being made in both the centralized and decentralized directions.
It turns out librarians and teachers teaching media literacy and critical thinking still has great value!
"Are the author's conclusions supported by the facts they present?" "Is their assertion internally consistent?" "What do you think the author is trying to make you feel and do by consuming this content?" "Is this a primary or secondary source?" "Can this be independently verified?" "Is the source credible?"
I really want to support alternate search engines. And removing/filtering AI generated imagery is really important for research and history.
But DDG scrapes some bottom feeder news publications. Newsweek and MSNBC wire reposts shouldn't be the top results. Ever. Those are ghost articles and link bait.
I don't know how they're going to keep this up long term but I appreciate / like the effort / feature. Even google is getting infected where I search for just an animal name and a bunch of the results are meme like AI images of something that looks "like" that animal.
Likewise, not sure how this will be sustainable. "Even google"... I'd instead say "Especially google". It's already a mess out there, and I hope Google and others take an active role in at least trying to make things usable.
I see a lot of Google search results thar are websites dedicated to sharing AI art. It's not hard to keep a list of those going, and that will presumably remain a bit portion of it.
Google has personalized search enabled by default for users, which makes it feel more "relevant", of course you get this by sacrifacing any means of privacy. For me, I don't mint unpersonalized results, however, my reason for stopping using ddg is a bit unorthodox one; the overall search performance (as in the time you need to wait until the results load) was significantly higher than Google's. I still believe that Kagi is a game changer.
> Google has personalized search enabled by default for users, which makes it feel more "relevant"
I am pretty convinced that one of the big reasons why Google search results have become borderline worthless to me is directly because it tries to "personalize" the results.
For me it has been the reverse. I tried DDG many times over the years, but results were always lackluster, especially for local Norwegian results.
But a few years ago Google became so bad, I changed my browser default to DDG, and frankly I can count on two hands the number of times I've needed to go to alternatives since. Especially Norwegian content has gotten good enough in almost all cases.
That said, it's not perfect, there's still stuff it struggles with.
I gave it a shot and toggling the button on definitely let some AI slip through. I assume they're doing their level best to heuristically determine whether something is AI, but if there aren't good pedigree indicators it's getting harder and harder to tell.
Helpful, but even more so would be a way to filter out AI content farms. It's getting to the point where I won't go to a site if I don't recognize its domain from the pre-GPT-3 era. Maybe domain age should be the filter. Not perfect, but no solution will be.
The amount of medical advice on those LLM content farms is very concerning, someone is going to get hurt eventually. I recently made a user script that tries to highlight LLM generated text and sometimes when I search things more than half the results are garbage and the entire articles (including the bios of the supposed professionals who write the articles!) get highlighted purple. One way to avoid them in a pinch when the content isn't time sensitive is to filter out anything from after November 1st 2021, but it's not ideal. I hope duck duck go finds a way to filter the AI pages out.
Jefit is my favorite app out of all I've tried. It's intuitive, let me create my CoolCicadaPPL workouts by day and automatically keeps track of my progress. When I'm working out after logging a set, it starts my rest timer, then loudly beeps when it's time to start the next set. The UI refresh was excellent.
More reasons to be thankful for uBlock!
“The filter relies on manually curated open-source blocklists, including the ‘nuclear’ list, provided by uBlockOrigin and uBlacklist Huge AI Blocklist,” DuckDuckGo said in a post on X.
I'm not sure the list has been provided by uBlock, looking at their repo[0] I understand that the list is available for uBlockOrigin and uBlacklist, but not made by them.
[0] https://github.com/laylavish/uBlockOrigin-HUGE-AI-Blocklist
I recently opened Pinterest after a year or so, and did look for some nice architecture / houses, and all results had the tag 'AI enhanced'. No I don't want AI houses, I want to see real architecture! Even though the tags are there, you can't filter on 'No AI'. So I stopped using it again.
I never understood Pinterest and how it works, but I know that people who DO understand and use Pinterest are basically saying that it’s been completely ruined by AI images.
It was mostly ruined by ads/SEO spam many years before. AI is probably the final nail in the coffin for any of the users who appreciated its original appeal, which was basic image curation from mostly blogs.
Wait, are you telling me that before LLMs, this was the good version of Pinterest?
This is the most common blocked site on ublacklist lists and kagi and probably it is also on many of the common dns blocklist.
It seems most sites that allow user generated content, and allow images/video, have been ruined by AI.
Before long we’ll need a second internet just for real people, as this internet will just be all bots and AI content. Dead internet indeed.
Or better yet, maybe I can get some real hobbies outside of tech.
Cara (cara.app) has images - art, photos, etc - that are not AI.
Incidentally, it is a site blessedly free of extra features that add nothing to the use case (sharing things you make).
I still have no desire to use it, but reading the complaints and explanations of AI wrecking Pinterest has finally given me some understanding of what people do there and why they bother.
Basically every art related social media hit that state unfortunately.
Aside from making search useless, it just demotivates everyone. No reason to use a social media site that appears full of bots.
Searching for recipes is definitely an exercise in ai suck.
Recipes are a sore spot for me. holy smokes is it a seo spam wasteland, I mean as far as our modern search engines go the whole web(statistically) is seo spam wasteland. but recipe sites are a specific particularly bad biome in that wasteland. they all follow the same infernal formula, about 5 paragraphs on how the recipe makes you feel, then the over specific formulation.
I know there are good recipe sites out there, you just will never find them, unless our search engine overlords slip and let one through.
The good web still exists, it is just that the search engines(our gateway to the web) are unable to find it under all the trash. My only solution so far has to been to start manually curating a list of good sites. A big part of the problem as I understand it is the search engines heavily prioritize new content over good content. To the point if I see a date in the current year next to the search result, I instantly reject it as seo garbage. If I were a better person I would sign up for Kagi, as I firmly believe a large part of the solution is to fix the business/customer dynamics, that is, you want to be the customer, not the product. But I am reluctant to sign up for another service so I am stuck shoveling slop to get through the web.
I’ve noticed Etsy has a lot of AI art showing up (prints, mugs, etc).
I think even in this case, I might want to see CAD renderings, which I can't help but think might get swept closer to AI generated than "rendered".
> No I don't want AI houses
Currently looking for an actual house on real estate aggregators, and feeling the same way
I’m annoyed for you.
The use case of someone preferring AI generated material vs the real material seems only to strike the algorithm and retention time of users so they always feel a sense of having what they need - despite it not existing.
Dear DDG, I genuinely, truly appreciate an AI filter on images.
But for the love of Bob, would you please enable absolute operands ("" or +) in image searching? It's been on your todo list for over a decade.
https://duckduckgo.com/duckduckgo-help-pages/results/syntax
https://duckduckgo.com/duckduckgo-help-pages/company/contact...
I miss when you could put `+foo -bar` into a search engine, ctrl-F on the top ten results, and reliably find `foo` on all ten of them and `bar` on none of them.
Nowadays, putting in `-bar` might increase the number of results that are directly about bar.
This bugs me no end. Image search is so challenging to use because the results are a vague interpretation of what you ask for, and don't try to be more descriptive because that's just more words to throw in the mix of results
Yesterday's discussion around Kagi[0] prompted me to poke through some settings there, as it had been a while. They've also added a "Exclude AI generated images" which they describe as "Best-effort removal of AI generated images from search results"
[0] https://news.ycombinator.com/item?id=44594475
There's a group of people for whom AI images seem to just be a trigger, and they simply never want to see them. So having a filter to remove those images makes sense. I don't really have that issue with AI images, but I often do want to have some way of filtering photographs from illustrations—be they drawings by humans, or images generated by AI. And that could go either way: maybe I want to see what an actual baby peacock looks like in nature sometimes, and maybe I want to see a cute illustration of a baby peacock other times. It would be great to have a filter for 'real' vs 'artificial'. Perhaps ironically, using AI is the best way to make that distinction on my behalf.
If I’m looking for images because I’m curious what a thing really looks like, it’s counterproductive to see AI’s “artist’s conception” of it, which is often anywhere between not particularly accurate and a picture of something else entirely.
A risk is that it will give people a false sense of confidence that they are viewing real content. The only way out of this mess is cryptographic methodss (based on hardware in cameras) that can allow end-users to verify photos as real and then we assume every other photo may be AI.
Cryptographic components in cameras have roots of trust that can be compromised as well. Also, photos can be staged (many famous examples of this).
The only real solution is to build social infrastructure that helps people identify the trustworthiness of a source. Some efforts are being made in both the centralized and decentralized directions.
It turns out librarians and teachers teaching media literacy and critical thinking still has great value!
"Are the author's conclusions supported by the facts they present?" "Is their assertion internally consistent?" "What do you think the author is trying to make you feel and do by consuming this content?" "Is this a primary or secondary source?" "Can this be independently verified?" "Is the source credible?"
If only it were that easy. The camera manufacturer keys would leak. Key revocation would be too difficult. PKI is hard.
Take a photo of an ai image, profit! Would the signature even survive the editing process of any photographer?
It would be an endless source of memes to try and make the most obviously fake image have the "verified real" badge.
DDG reminds me of what Google used to be. Hopefully they stay that way.
While I like the effort, I can imagine the amount of false-positive images that this will flag, but still, better than nothing.
As the technology nears perfection, whether something is "AI" will become a test like whether something is porn: "I know it when I see it."
I really want to support alternate search engines. And removing/filtering AI generated imagery is really important for research and history.
But DDG scrapes some bottom feeder news publications. Newsweek and MSNBC wire reposts shouldn't be the top results. Ever. Those are ghost articles and link bait.
Does Kagi skip these kinds of results?
I don't know how they're going to keep this up long term but I appreciate / like the effort / feature. Even google is getting infected where I search for just an animal name and a bunch of the results are meme like AI images of something that looks "like" that animal.
Likewise, not sure how this will be sustainable. "Even google"... I'd instead say "Especially google". It's already a mess out there, and I hope Google and others take an active role in at least trying to make things usable.
I agree, although I don't know if Google cares. They're happy to provide tools to generate this stuff too ...
I see a lot of Google search results thar are websites dedicated to sharing AI art. It's not hard to keep a list of those going, and that will presumably remain a bit portion of it.
Do they still hide tankman images in search results too?
https://news.ycombinator.com/item?id=27394925
This is easy to check, and no they do not. I can see many photos on Bing of tank man.
I have used ddg for many years. But I think the search results were too irrelevant to me most of the time now I have moved on
Google has personalized search enabled by default for users, which makes it feel more "relevant", of course you get this by sacrifacing any means of privacy. For me, I don't mint unpersonalized results, however, my reason for stopping using ddg is a bit unorthodox one; the overall search performance (as in the time you need to wait until the results load) was significantly higher than Google's. I still believe that Kagi is a game changer.
> Google has personalized search enabled by default for users, which makes it feel more "relevant"
I am pretty convinced that one of the big reasons why Google search results have become borderline worthless to me is directly because it tries to "personalize" the results.
For me it has been the reverse. I tried DDG many times over the years, but results were always lackluster, especially for local Norwegian results.
But a few years ago Google became so bad, I changed my browser default to DDG, and frankly I can count on two hands the number of times I've needed to go to alternatives since. Especially Norwegian content has gotten good enough in almost all cases.
That said, it's not perfect, there's still stuff it struggles with.
That was because Bing wasn't as good as Google, but now they're kinda even... because Google became much worse.
ddg pretty consistently fails to return expected a results, compared to google.
It’s a shame, I don’t want to stay with google, but it still does best fit my needs for search.
Ex: tumblr.com/ddgvsggl
Well, kind of. Credit for them trying.
I gave it a shot and toggling the button on definitely let some AI slip through. I assume they're doing their level best to heuristically determine whether something is AI, but if there aren't good pedigree indicators it's getting harder and harder to tell.
Helpful, but even more so would be a way to filter out AI content farms. It's getting to the point where I won't go to a site if I don't recognize its domain from the pre-GPT-3 era. Maybe domain age should be the filter. Not perfect, but no solution will be.
The amount of medical advice on those LLM content farms is very concerning, someone is going to get hurt eventually. I recently made a user script that tries to highlight LLM generated text and sometimes when I search things more than half the results are garbage and the entire articles (including the bios of the supposed professionals who write the articles!) get highlighted purple. One way to avoid them in a pinch when the content isn't time sensitive is to filter out anything from after November 1st 2021, but it's not ideal. I hope duck duck go finds a way to filter the AI pages out.
> I recently made a user script that tries to highlight LLM generated text
How does that work?
It uses one of Mozilla's models: https://huggingface.co/fakespot-ai/roberta-base-ai-text-dete...
It has a pretty high false positive rate though, but it reliably highlights AI generated spam websites and saves me from having to read them.
Random number generator and vibes probably.
Not just that, the text results rewrite results to be AI slop: https://news.ycombinator.com/item?id=39808531
Excellent.
Jefit is my favorite app out of all I've tried. It's intuitive, let me create my CoolCicadaPPL workouts by day and automatically keeps track of my progress. When I'm working out after logging a set, it starts my rest timer, then loudly beeps when it's time to start the next set. The UI refresh was excellent.
Wrong thread?
Nice! Although the filter is unlikely to be perfect, it will be great to have another way to filter AI slop.
Now if only HN had a filter for negative articles, especially about AI.
Finally! Now let's do the rest of the internet.