I know this is an anecdote and very subjective but I've really never discovered something that I loved with a recommendation algorithm.
Whether it be videos, music, livestreams, books... Everything that I've considered a "10/10" has been recommended by a human or a non-personalized algorithm—such as "Most popular". Whether that's a direct recommendation by a friend, a comment on HackerNews, someone that I already follow that mentioned the thing in question.
My RSS client fetches my YouTube "subscriptions" and it's been years since I've been on the homepage.
I’ve had some great recommendations from YouTube. For example the video below, called “The Squirrel and the Peanut”, which by just adding some well chosen classical music to some found footage of a squirrel getting a peanut turns it into a moving story of overcoming your fears told in a mere 1:32.
I agree that the algorithm isn't always the best. Interestingly however I occasionally get "legendary pulls" (I think that's a gatcha term?) on youtube - a 1990s stop motion video of some legendary niche Japanese animator, or some other 8+ years old video in 240p that's absolutely amazing.
But "fortunately" for me, on my main account, my feed is 80%+ decent stuff - science channels, AOPA safety videos - stuff like that. But my other account has absolutely terrible suggestions on the algorithm, so it probably varies a ton on how much and what you watch.
> Everything that I've considered a "10/10" has been recommended by a human
Have you ever found a 10/10 on your own?
If so it's possible you were recommended that by an algorithm but you just didn't register it, because a human recommendation is more of a memorable event.
If I've ever did, it's after having been recommended something. For example if someone sends me a song and then I go to the "most popular" songs from that same artist. Sometimes I would stumble upon something I like even more than the recommendation.
For example if it's a genre that interests me I would search "[genre] mix" on YouTube. In which case it's usually mixed by a human.
Granted I have very specific tastes in music and 98% of the time I listen to things I've already listened to hundreds of times... I sometimes try the song radio but it's not to discover a "banger".
There are a lot of AI generated shorts around animals.
A common thing I see is a baby animal needing rescue by a human (which it does) and it comes back later on and rewards the human with a gift of some kind it thinks is valuable.
I watch a few podcasts as well and there are more that have their scripts generated and voiced by AI
Ohh the deep rabbit hole I went on this. There are so many variations of this like using trademark toys like Labubu, Spider-Man etc. I don’t suggest anyone looking for these unless they want their short feeds to be populated with AI nonsense.
I saw a video I wanted to share with someone, but it was part of a compilation. So you just search for it, right?
So I searched "cat lets brick fall onto mouse" and got... 100000 AI generated videos of cats with bricks? And cats with mice and cats being rescued by people (like you said). But not the video I was looking for.
We've totally passed the point where real information is impossible to find anymore. Video generation was really out of reach / delayed for a long time, and honestly all of those probably have a digital watermark in them that could be detected. YouTube could have prevented this if they'd have just been more proactive with detection and filtration. A simple "AI generated" and "not AI generated" filter would have prevented this.
> I saw a video I wanted to share with someone, but it was part of a compilation. So you just search for it, right?
There's something with these compilations. Almost as if deliberately AI slop is mixed in to numb the public to it, or for some AI startup to testdrive on an unaware public how good their stuff is.
Take compilations of lightning strikes for instance. There's always a couple that are just too spectacular or just unbelievably. Like a ball lightning going across the street.
It’s mostly content farms based out of Asia - Vietnam, India, Bangladesh etc pumping out these stuff to make a quick buck. It’s like 5 min crafts but now easier and faster due to AI and no personal overheads.
I see a lot of educational animal videos that copy the content of BBC Earth only to replace David Attenborough voice with AI, and it unreasonably irritates me.
That's an improvement as before people would abuse real animals to fake seemingly wholesome "before-after" videos by showing snippets in the wrong order.
Another common theme is to kill rare animals to stage cool photographs.
However, I don't assume AI slop will (fully) replace that cruelty, unfortunately. But maybe using AI slop will be easier than animal abuse/killing so that those business models run dry for the most part.
My feed is full of AI generated shorts summarizing books, animes, movies. The original piece name is never mentioned, and it tells the story in a very descriptive way, such as "The man was alone in the woods when...".
> A common thing I see is a baby animal needing rescue by a human (which it does) and it comes back later on and rewards the human with a gift of some kind it thinks is valuable.
Cats getting totally excited to see their owner, because they dearly missed them. Cats filmed in night cam dropping weird animals from the forest on sleeping boss. Olympic athletes, gorgeous, but not real. Countless disasters where people die, generated. Youtube shorts is a pile of steaming garbage. As long as it sells, your brain may rot.
Worst are imho on the regular long vids side, the geopolitical advisor deep fakes, giving background to the news. Some with well over a million followers. Many of those have the same "we are a fan of the real person" disclaimer, many have no disclaimer.
And no one in the comments, of which many look fake too, notices it is AI. That is the most scary part.
> And no one in the comments, of which many look fake too, notices it is AI. That is the most scary part.
Many who notice won't bother commenting, because most who notice know how pointless that is (counterproductive in fact: a comment is an interaction, any interaction is a positive for the "content"). Those that do notice and comment are either drowned out by
• those too numbed on the brain to care, let alone notice, who lap it up, and praise it
• bots (either those being used to interact with the clip to drive it's interactions counters, or more general spam bots)
or if there is anyone/anybot monitoring the negative comments are removed.
It has been years now that I only care about my subscriptions. I also installed an extension to remove anything else (especially shorts!), and that works great for me.
The downside is perhaps that I rarely discover new content, but YT can't be trusted to give me that organically.
Every time I access YT without being logged to my account and this extension, I'm surprised by the amount of garbage that YT feeds me based on my IP and/or location they infer from it. I worry what effect that is having in the population that consume it without safeguarding.
Sure, there's always been garbage TV, but this is the next level, and on demand.
Thanks for the suggestion. I just bought the UnTrap creator’s bundle social media product and tested with YouTube, X, Facebook, and LinkedIn. This is a safari extension. My biggest surprise was how pleasant it was to spend t minutes in Facebook.
The end game is that computer generation beats human content for a subset of the population who becomes accessible to advertisers and propagandists through model alignment, and unresponsive to word-of-mouth.
The end game is eliminating the last remaining human element in the engagement optimization pipeline, so that the corporations can control 100% of it.
Platforms like YouTube and TikTok already have almost full control of how the majority of users spend their time on their platforms. They open the app, they immediately get a feed of content algorithmically selected to keep them on the app for as long as possible. They don't need to search, they don't need to think about what they want to watch, they just consume. Fully automated consumption with 0 human effort involved.
Well, almost. There's one last thing remaining: you still need humans to produce the content that you then put on people's feeds. Or rather, needed. Now that the actual production can also be automated, those platforms no longer need to put effort into finding existing human-created content that will keep people watching - they can just generate new, algorithmically perfect content. This is their endgame.
Other advantages to generating content: (1) fewer copyright issues. (2) No creators to pay, just GPU bills scaling with the use of the platform. (3) a much smaller critical mass.
Given these advantages I expect the current "social media" to be replaced with a new one, rather than them pivoting. The next big thing after tiktok might be something that only has generated content, where a last final bit of "social" is taken out of "social media".
Generating stuff is very cheap compared to building and training the model. When you have your model done you're incentivized to use it as much as possible. Maybe even considering the sunken costs.
Why can't it be the actual tiktok? By simply winning competition with humans i.e. whereas vast majority of humans see their pay go too low to bother to continue?
There is no “end game” it’s just hustlers out for a buck for themselves — YouTube, the slop makers, the ad companies, the bots scraping videos, all of em.
Same as it ever was. When this cash cow proves worthless or runs out it’ll be another thing.
Someone figured out how to make computers be able to create content that is costly to distinguish from human-made content. Someone else is using it to pump out AI slop in the hopes that it will make them a quick buck. The platform becomes unusable for anyone that values their own sanity. No "end game" to be found.
AI will be the worst thing that happened to society in a very long time.
I have noticed people on the subway watching them. I sneaked a peak on a few of their phones and it was legit AI slop with clear signs (for me at least) that it was AI generated. The end user (viewer?) seemed hooked but they are mostly shorts (10-20sec videos) and you can see their fingers swiping to the next one.
The other day my mother told me if I watched some random AI slop (Putin getting in a physical fight with Trump) and I asked her why she watches this stuff and her answer is that it comes up in her feed. She said it was funny.
I just blocked a fake Brian Cox one a couple of hours ago. There can be a certain time waste looking at the things before figuring they are rubbish.
Youtube should give you more options than just block or don't show me this. You should be able to click 'AI fake of real person' so they don't get inflicted on others unless they like that stuff.
What Youtube pushes vs what I watch are so completely different that I am shocked. I almost exclusively watch geology, space, tech and cooking and it pushes at me (quick check of the home page): disgusting cow medical procedures, random sexual videos, celebrity/influencer junk. Oh, wait, there is a geology thing....and it is a junk sci/scare video that has no actual geology related content. Yeah. Their algorithm is clearly working well. I would go to other platforms but there isn't one. I have started donating on patreon and watching there where practical (even though it is still yt serving it up) but that doesn't easily allow me to discover new things. The real challenge here is finding new things. When search is so completely broken, how do you find valid new things to watch?
Out of curiosity, do you have watch history enabled/disabled?
I found the feed with it enabled is much better than disabled and I have it finetuned to be more in line with the niches I care about.
I am also very proactive with marking channels or content I don't prefer with the "Not interested" or "Don't recommend channel" as well as going through and pruning any content I don't want from my watch history directly.
It's not perfect but it's orders of magnitude better than my logged out or watch history disabled account (though not sure if they have since updated it to not show anything at all)
It’s sad that we replaced manual curation with recommendation algorithms because they were doing the hard job of surfacing relevant content for us and the end result is that you’re spending more effort and intellectual energy into steering the algorithms in the right direction than we ever did before
Manual curation would be waste of time. Why should I spend hours looking through slop when I can get algorithm do it reasonably well.
Youtube actually works reasonably well, as long as you remove from history content you disliked, use do not recommend channel and possibly not interested option. It might not be highest of quality, but at least content stream then is passable.
I follow N YouTubers who produce genuinely useful and insightful videos on the state of the market and personal finance etc, but the algorithm just recommends people with stock tips and scaremongering junk
Probably there not enough content. Just the same repackaged in human or machine produced slop.
People are addicted to YouTube but I think the key to the healthy watching habits would be restricting the screen time.
Someone knowing about things he's interested in has few problems separating the new and informative content and if he had, say, two hours a week for watching he'd probably enjoy what he sees.
Two hours is just an estimate I came up with, it can be an two hours a month or hour a day. The important thing is that YT just doesn't create enough real new information and after that it is just slop and brain rot, regardless of your habits and filters.
Not only that but also the same short videos keep repeating. I have tons of great suggestions from Youtube on the main feed from creators I like, yet the Shorts feed is almost 90% garbage and AI slop.
It's deliberate, I'm sure. People say they want the vegetables, but then go on to watch hours of fast food / Shorts. Clearly the algorithm knows.
do you have an account about which "you care somehow", like banning/disabling specific channels explicitly?
Ive found out: If I tell youtube what I want by sometimes thumb-up/down and explicitly clicking on "no more vids from this channel", then the results a) much better and most of the trash is removed.
Though, new AI-slop channels popup some days - esp. new/fresh channels, but these you can ban anyway since most new channels are AI slop. or "XY ... official" or "XY .. best of" or "XY .. clips" or "XY... fails"
I honestly haven't tried putting effort into the platform for a while now. I have very targeted things I watch (links to the channels) and go direct instead of navigating the homepage at all. I think my upper confidence bound on the value of the YT homepage and improving it is pretty low so it will be a while until I give it a try again.
If you somehow curate your YT account, the shit wont show up anymore (only new shit not banned already) and you will see one thing: A curated youtube feed is actually rather empty :-D
On some days, I really have to actually look precisely if there is something new at all: Im a subscriber of 30-35 technical/business channels - its very rare that i could "spend a whole day watching YT", because there is just not enough new stuff.
Before curating my account, I also had the impression that "its endless and not any sort of structure" - but it has.
That ain't exactly a study as people think of, something scientist do and that get published in a journal, peer reviewed, approved, etc.
It's just kapwing employee checking YouTube channel views and reporting it, same for the feed, so I can't say that the 21-33% number can be trusted.
Now the fact that YouTube has ai slop isn't new, but my bet is that many of these 'subscriber' are in fact bots used to inflate their numbers.
And on a more personal touch, I suggest you install YouTube unhook extension to your computer AND the ones of your relatives. The one less tech savvy that are the more susceptible to fall for this. I surprised my father once watching these kind of crap, he couldn't understand it wasnt even human made. Now I know he's safe from at least that.
Breaking this down by country must be the least interesting analysis possible. I'm much more interested in a breakdown by video genre and the type of person watching.
Youtube is horrible, specially the force-fed "shorts". Its 95% AI slop with the same generic graphics and voice-over. From the actual videos most are also AI generated, making me quit the video as soon as i see it.
My shorts are mostly of the creators I follow or adjecent creators. I do get the occasional AI slop, but it looks more like the algorithm is testing someone new, to learn what they are, than the algorithm feeding me slop.
Now Facebook/insta shorts, they are somehow just trash. But maybe that is because I don’t follow any creators on those platforms.
At a societal level [if that's what we should be worried about], it's not what percentage of results are slop. It's what percentage of people believe the slop. Arguably, that seems to be decreasing proportionally. What I mean is: Slop is on a parabolic, hockey stick upward, and trust is on a logarithmic decline. So, good? - let it blow itself out? Bad information is worse than no information, and evolutionary pressures have many ways of proving that.
Skimmed through the article, some interesting numbers but not a single statistic is per capita (or per million, whatever). How do I understand the scale of the phenomenon without the per capita figures? Sorry but seems a bit useless.
AI is OK. I used to ignore AI stuffs on YouTube but in recent months the channel “Napoleons Hill Notes” helped me tremendously. It is an AI voice reading channel.
That's fair. I don't even know whether they are authentic NP notes (never read his books), but the channel is the only thing that helped me to be very close (a few days/a couple of weeks away) to complete a non-trivial learning project out of school -- first time ever in my life, how weird that sounds. I guess as long as it works it's fine.
I saw one Warren Buffett video where the voice was Buffet's and he said "in this video I will show you...". Most videos featuring Buffet's voice are snippets from actual interviews etc, but this one threw me off. I stopped it immediately. Sigh.
It has really become horrible in the last some days, probably weeks
now. It is not just fake-videos AI generated, without Google having
any decency to mark it, as it wastes my time and the time of others -
but there are now also videos where some videos are real (I know
because some of these videos came from years ago), mixed in with
AI fake crap. Now, I am able to spot many AI videos, but I bet many
other people simply don't have the knowledge. That is also a generational
problem, where people have a harder and harder time to separate real
from fiction. But I have no desire to waste my time with fake, so
Google now has started to kill youtube. I still have use cases of
youtube without this problem, e. g. good music (here, whether it is
AI generated or not, makes no real difference IF the music is good;
but most AI music is crap anyway but I can not listen to it and only
focus on good music), but this is getting more and more of a dead end
here.
Google already killed its search engine and other things. It is continuing
on its path to now kill Youtube. And, mind you - Youtube already had
problems before AI. Many content creators felt violated and abused by
Google. I really think we should end Google as a company - it is not doing
the world any good now. It changed completely; the old Google is permanently
gone. Nobody needs the AI slop infected money-milking-via-ads machine.
Also, Google further ruined its already by-now-total-crap search engine, with crap videos nobody really cares about in 99% of the cases. Or the "others searched for xyz" - what the heck do I care what others did? If I want to find something, I don't want google to distract with excuses. Google abuses people here. It is an EVIL company now. These are not "accidents" - this is deliberately aimed at wasting people's time. I want compensation money for Google wasting my time here. This has been different in the past, so it is 100% Google's fault. No more excuses here.
There are already a lot of impersonating AI-Slop videos appearing, not just faking Yanis Varoufakis, but also many other political commentators. It's hard to find the real videos by now.
I tried to flag these videos, but the process of doing so is so cumbersome that I finally abstained from it. AI-Slop is slowly destroying everything: books, youtube, education, in the end everything that is data driven...
Where it could be useful, e.g. high quality video translation, it fails utterly.
I noticed that too and it’s kinda scary. Soon we will have the opposite of canceling, where the target will be deepfaked to say everything and its opposite to nullify their signal to noise ratio.
Yeah, I hate that also. I just subscribed to Yanis Varoufakis‘s official feed and installed a Safari extension to block everything on YT not in my subscriptions. I think this will work out: use the YouTube app occasionally to discover new material to subscribe to, but usually access YT on Safari using this extensikn.
I'd be more concerned about just slop, whether AI or human created. And the fact that Youtube content is overwhelmingly slop - regardless of creator type - is not news at all.
Reminder that you don’t have to play by YouTube rules.
Unsubscribe from everything and disable video history, that also stops the slop recommendation engine.
YouTube search still works, you’ll still come across interesting YouTube links and if there are channels you want to subscribe to there is a way to get them in your RSS reader.
You can also get the feed to be full-length videos only, ignoring all the noise coming from shorts.
Does it? Last time I tried YouTube search it was useless. It showed two or three relevant results, then a bunch of shorts, then a long list of totally unrelated "you might also like" results.
Turn off watch history and there’s no feed whatsoever, they started to refuse to show anything on the home page with watch history off since a couple years ago. Yes that includes even Shorts, you’re not allowed to doomscroll without watch history.
> Recommendations are off
> Your watch history is off, and we rely on watch history to tailor your feed. You can change your setting at any time, or try searching for videos instead. Learn more.
> they started to refuse to show anything on the home page with watch history off since a couple years ago.
I remember when that happened. It was hilarious to me because the way they phrased it felt like they think of it as a punishment or incentive, but not having anything on the homepage is incredibly refreshing, exactly what I wanted. YouTube’s homepage (at least in my country) is just garbage videos no one should be watching.
E.g.: Just for entertainment I recently searched for information on a new LLM model. The result was filled with AI garbage. I stumbled upon videos with AI generated content, presented by an AI generated voice and an AI generated human. And if it is not something like that, YouTube at least lies to me about the language of the video and tries to make me listen to an AI voice which is equally vile.
There are times where I constantly have to close videos after a couple of seconds. It is unbearable. At least they should mark these videos clearly. At best they should allow me to filter them out but of course they won't do that or else they couldn't make the experience on their platform even worse than it already is, what seems to be their real goal.
That's a skill issue on your part. Do not use YouTube to search for information. If you need to find information on a new LLM model, look for official documentation from the model's maker. This applies to everything.
No, it's not. I search for videos and i get AI garbage as a result. It shouldn't matter what i search for or if you think that it is dumb to search for it. It shouldn't return garbage in the first place unless I explicitly ask for it.
> Do not use YouTube to search for information. If you need to find information on a new LLM model, look for official documentation from the model's maker.
That is what I do but I like the entertainment factor of some YouTubers. I know that I do not get the best information from them but sometimes I'm entertained. That's why I'm using YouTube after all. Entertainment (and I'm not entertained by AI).
> This applies to everything.
So you are saying: In this glorious new AI world I need a search diploma just to evade the endless AI slop YouTube is filled with nowadays. And than there are topics that are taboo to search for at all. From what angle do I have to look at this to discover the good part?
I avoid watching any AI content, and I still periodically have music channels where every video has view counts in the low hundreds get recommended, with a description implying it's AI.
My wife and I pay for YouTube and the prevalence of AI Slop definitely reduces the value we get for our money.
YouTube has done a fairly good job over the years of providing useful recommendations but lately I always immediately ‘back out’ of any AI generated videos or materials with robot-voiced narrations. I keep hoping that the algorithms quickly learn to not show me this stuff, but in the meantime I have had to change my viewing habits to following specific channels - and this reduces ‘discoverability.’
YouTube promotes slop and "brainrot" irrespective of AI use, this through the client being tuned for clickbait title presentation, forcing back shorts after every relaunch and promoting clickbait select single frame thumbnails.
AI slop is a significant improvement over countless videos in the past few years that were little more than slide decks with bad TTS. The current equivalents are better written, better spoken, better visuals, etc.
I thought it was simply the way all youtube videos are now made :) my feed has little but those, i need to search for something specific to see anything else.
And ~90% of google searches now gives you 10 videos as answers to any query rather than just the web pages that have been turned into videos because youtube makes them more money than regular ads. Assholes.
HN is the ONLY place I've seen Kagi recommended as useful. So much so I bought it for a year and had to move back to google. I was missing so many search results.
Maybe I didn't configure Kagi enough? But I don't want to configure a search engine.
This is interesting to me. I've done basically no configuring (except frequent use of a "site:reddit.com" lens) and it's been great. Now when I see Google or Bing results it's jarring how bad they seem.
For search I'm increasing just using Claude + web search tool, it's better than Google has been in years. I have my own local ollama-powered RAG setup as well which isn't as good but may serve alright if/when Anthropic starts with ad insertions.
I find doing more focused searches with bang patterns work well for me, I.E searching wikipedia, manpages, mdn etc directly. Sometimes coupled with LLM to more quickly understand the syntax because docs tend to be sparse with examples.
I still miss something for quickly finding random facts like "how long to boil an egg" but for programming the above works quite well.
DuckDuckGo has also been my goto for years, but it is also getting swamped with SEO-rigged spam sites.
2 days ago I was making multiple searches looking for the websites of specific hotels and spa facilities, and none of them showed up in the first 4 pages of results, even when searching by exact name.
Out of desperation I switched back to Google, and, surprisingly, it was willing to give me what I was looking for on the first page. (But not as first result, of course..)
> DuckDuckGo has also been my goto for years, but it is also getting swamped with SEO-rigged spam sites.
True, but DDG seems, at the very least, no more afflicted than other engines. If I'm supposed to pay for a search engine, I expect something better than DDG.
Kagi systematically doesn’t return any good answer whatsoever. It’s such an awful quality that I can’t imagine anyone seriously promoting it, therefore I bet these comments about Kagi are advertising posted by bots. It’s not possible otherwise.
I find it to be more reliably useful than the other non-Google alternatives I’ve tried. I find its “PDFs Only” filter to be awfully handy too. Proudly not a bot, but I freely admit that, because I pay for it and our incentives seem aligned, I’m biased toward giving it the benefit of the doubt.
Always open to new horizons, though—are there non-Google search products that you find to consistently work better?
I share the same sentiment. I know many people praise Kagi, and I respect the effort behind it too. I tried it for three months and realized it was not for me. Google works just fine for my needs.
Haha… no bot here. Been using Kagi for years now. Not sure what you‘re searching for. My own tests, admittedly early, found no instances, where Google gave better results.
> And ~90% of google searches now gives you 10 videos as answers to any query rather than just the web pages that have been turned into videos because youtube makes them more money than regular ads. Assholes.
Which also reinforces already declining literacy rates.
The most insidious aspect is that tech CEOs and influencers keep parroting how "AI" will bring global prosperity and empower the masses. When it's clear to anyone with an ounce of reason that the exact opposite is happening. The amount of hypocrisy and manipulation is sickening.
I feel like the search result of YouTube videos on Google search is much worse than the result on YouTube itself. It's strange because Google develop both Google and YouTube search. It's like reverse Reddit, where the website's search is so unusable you have to search on Google like "xxx reddit".
9 crappy videos before they figure maybe they should link to the documentation after all.
To give them some credit, at least that search did not also come with a crappy AI summary that is broken, not applicable and wastes valuable space and bandwidth.
edit: corrected count, I missed one. And a very prominent 'see all' link before the actual one.
I searched for that exact phrase out of curiosity because I haven’t used Google in years… and it’s even worse than you say.
I got an AI summary that takes up half the screen, which doesn’t even give the right answer, then 5 YouTube videos with thumbnails and extra crud (most not even related to the question but just mention “ArduPilot” somewhere on the title), then half a page of “Other people searched for..”
Then about 3 “screens” down the page I get the ArduPilot homepage and then a bunch of embedded Reddit/Facebook discussions about ArduPilot in general, none about setting up GPS.
It seems to differ extremely by account or region?
I have to say AI summaries work pretty well for me on Google and my overall subjective sense is the results have been getting better the past 1-2 years.
Here's what I get for your exact search and it looks pretty reasonable?
I knew you'd come back with a comment like this one. You never argue in good faith. No, I wasn't talking about AI generated videos because then I would have written a different comment. Bye now.
I sometimes wonder how much other browsers, countries, or languages play into the search results for a particular user.
Every time I read someone complaining about Google search results, I try to use another search engine for awhile, just to see if I'm missing something. Every time I do this, the alternatives are awful and I shake my head and go back to Google search.
> AI summaries are actually rather good, in the general case.
The google ones for me seem to be talking about something unrelated to what I'm looking for a significant percent of the time because it interprets it as a much more popular concept that looks like my query if you squint very hard. Other times it pulls up seo-spam tier answers that are plainly wrong.
I hate to say it, but Zuckerberg is right about where online content is going. The future is just an endless feed of personalized short form AI brain candy.
This is definitely true for my mom's feed because she watches so many random pet videos.
Mine is like ~1%. I'm actually surprised about how good quality the feed algo has been, and for especially showing small creators and older videos too.
Each person has their own filter-bubble, I guess.
Search engines are definitely worse. DDG shows so much AI shit websites now for specific search queries. Google is almost as bad. I'm having to double quote almost everything because the quality and fuzziness of matching is much worse now. I miss the OR operator and groups parens that were possible way back.
The current utilization of generative video is almost universally horrible, which this article does suggest, so I'm not too surprised there are players trying to differentiate themselves. Slop for thee, not for me.
How can I block AI generated videos (especially AI generated scripts) in my browser? Youtube has to give the viewer an option or the website is going suffer greatly from AI slop...
Why is youtube in general making it so hard to block content or channels? and now they made it harder to clean up the user's own viewing history, it just doesn't make sense...
Do you believe we can automatically detect AI generated videos, without having too many false positives and without spending huge amounts of resources playing the cat and mouse game?
It's not possible and it's not economically viable. Best we could do is some sort of signature to prove that the source of a video is trusted by the proof issuer.
I know this is an anecdote and very subjective but I've really never discovered something that I loved with a recommendation algorithm.
Whether it be videos, music, livestreams, books... Everything that I've considered a "10/10" has been recommended by a human or a non-personalized algorithm—such as "Most popular". Whether that's a direct recommendation by a friend, a comment on HackerNews, someone that I already follow that mentioned the thing in question.
My RSS client fetches my YouTube "subscriptions" and it's been years since I've been on the homepage.
Surely I'm not alone but it surely feels like it.
I’ve had some great recommendations from YouTube. For example the video below, called “The Squirrel and the Peanut”, which by just adding some well chosen classical music to some found footage of a squirrel getting a peanut turns it into a moving story of overcoming your fears told in a mere 1:32.
Watch with the sound turned up.
https://www.youtube.com/watch?v=LQ1ZYGHmtN8
I agree that the algorithm isn't always the best. Interestingly however I occasionally get "legendary pulls" (I think that's a gatcha term?) on youtube - a 1990s stop motion video of some legendary niche Japanese animator, or some other 8+ years old video in 240p that's absolutely amazing.
But "fortunately" for me, on my main account, my feed is 80%+ decent stuff - science channels, AOPA safety videos - stuff like that. But my other account has absolutely terrible suggestions on the algorithm, so it probably varies a ton on how much and what you watch.
Akshually I disagree, I discovered the song "Plastic Love" from a YouTube recommendation. Then I dived into city pop as a result.
But taking into account all the thousands of not-so-good recommendations I guess you are right
> Everything that I've considered a "10/10" has been recommended by a human
Have you ever found a 10/10 on your own?
If so it's possible you were recommended that by an algorithm but you just didn't register it, because a human recommendation is more of a memorable event.
> Have you ever found a 10/10 on your own?
If I've ever did, it's after having been recommended something. For example if someone sends me a song and then I go to the "most popular" songs from that same artist. Sometimes I would stumble upon something I like even more than the recommendation.
For example if it's a genre that interests me I would search "[genre] mix" on YouTube. In which case it's usually mixed by a human.
Most of the great music I've discovered is thanks to Spotify's "Play song radio" feature.
Granted I have very specific tastes in music and 98% of the time I listen to things I've already listened to hundreds of times... I sometimes try the song radio but it's not to discover a "banger".
There are a lot of AI generated shorts around animals.
A common thing I see is a baby animal needing rescue by a human (which it does) and it comes back later on and rewards the human with a gift of some kind it thinks is valuable.
I watch a few podcasts as well and there are more that have their scripts generated and voiced by AI
Ohh the deep rabbit hole I went on this. There are so many variations of this like using trademark toys like Labubu, Spider-Man etc. I don’t suggest anyone looking for these unless they want their short feeds to be populated with AI nonsense.
It's awful :(
I saw a video I wanted to share with someone, but it was part of a compilation. So you just search for it, right?
So I searched "cat lets brick fall onto mouse" and got... 100000 AI generated videos of cats with bricks? And cats with mice and cats being rescued by people (like you said). But not the video I was looking for.
We've totally passed the point where real information is impossible to find anymore. Video generation was really out of reach / delayed for a long time, and honestly all of those probably have a digital watermark in them that could be detected. YouTube could have prevented this if they'd have just been more proactive with detection and filtration. A simple "AI generated" and "not AI generated" filter would have prevented this.
> I saw a video I wanted to share with someone, but it was part of a compilation. So you just search for it, right?
There's something with these compilations. Almost as if deliberately AI slop is mixed in to numb the public to it, or for some AI startup to testdrive on an unaware public how good their stuff is.
Take compilations of lightning strikes for instance. There's always a couple that are just too spectacular or just unbelievably. Like a ball lightning going across the street.
It’s mostly content farms based out of Asia - Vietnam, India, Bangladesh etc pumping out these stuff to make a quick buck. It’s like 5 min crafts but now easier and faster due to AI and no personal overheads.
Detecting AI is and will always be reactive.
It will always be subject to the delay in detecting the bypasses of the latest AI techniques.
I see a lot of educational animal videos that copy the content of BBC Earth only to replace David Attenborough voice with AI, and it unreasonably irritates me.
That's an improvement as before people would abuse real animals to fake seemingly wholesome "before-after" videos by showing snippets in the wrong order.
Source?
https://www.nationalgeographic.com/animals/article/how-fake-...
Another common theme is to kill rare animals to stage cool photographs. However, I don't assume AI slop will (fully) replace that cruelty, unfortunately. But maybe using AI slop will be easier than animal abuse/killing so that those business models run dry for the most part.
My feed is full of AI generated shorts summarizing books, animes, movies. The original piece name is never mentioned, and it tells the story in a very descriptive way, such as "The man was alone in the woods when...".
I think you need to start clicking "do not recommend similar" or whatever this option is called.
> A common thing I see is a baby animal needing rescue by a human (which it does) and it comes back later on and rewards the human with a gift of some kind it thinks is valuable.
I've posted an article on this matter: https://news.ycombinator.com/item?id=46121555
Cats getting totally excited to see their owner, because they dearly missed them. Cats filmed in night cam dropping weird animals from the forest on sleeping boss. Olympic athletes, gorgeous, but not real. Countless disasters where people die, generated. Youtube shorts is a pile of steaming garbage. As long as it sells, your brain may rot.
Worst are imho on the regular long vids side, the geopolitical advisor deep fakes, giving background to the news. Some with well over a million followers. Many of those have the same "we are a fan of the real person" disclaimer, many have no disclaimer.
And no one in the comments, of which many look fake too, notices it is AI. That is the most scary part.
> And no one in the comments, of which many look fake too, notices it is AI. That is the most scary part.
Many who notice won't bother commenting, because most who notice know how pointless that is (counterproductive in fact: a comment is an interaction, any interaction is a positive for the "content"). Those that do notice and comment are either drowned out by
• those too numbed on the brain to care, let alone notice, who lap it up, and praise it
• bots (either those being used to interact with the clip to drive it's interactions counters, or more general spam bots)
or if there is anyone/anybot monitoring the negative comments are removed.
It has been years now that I only care about my subscriptions. I also installed an extension to remove anything else (especially shorts!), and that works great for me.
The downside is perhaps that I rarely discover new content, but YT can't be trusted to give me that organically.
Every time I access YT without being logged to my account and this extension, I'm surprised by the amount of garbage that YT feeds me based on my IP and/or location they infer from it. I worry what effect that is having in the population that consume it without safeguarding.
Sure, there's always been garbage TV, but this is the next level, and on demand.
I did the same. Unhooked for desktop and UnTrap for iOS. No suggestions, no shorts, and no comments. Just the videos from creators I subscribed to.
Thanks for the suggestion. I just bought the UnTrap creator’s bundle social media product and tested with YouTube, X, Facebook, and LinkedIn. This is a safari extension. My biggest surprise was how pleasant it was to spend t minutes in Facebook.
What's the end game?
AI sloop ads for dating apps full of ai chat bots , YouTube watched by AI bots.
I was a bit surprised Spain has the most subscribers to ai sloop. Kinda weird considering the population size compared to the US
The end game is that computer generation beats human content for a subset of the population who becomes accessible to advertisers and propagandists through model alignment, and unresponsive to word-of-mouth.
The end game is eliminating the last remaining human element in the engagement optimization pipeline, so that the corporations can control 100% of it.
Platforms like YouTube and TikTok already have almost full control of how the majority of users spend their time on their platforms. They open the app, they immediately get a feed of content algorithmically selected to keep them on the app for as long as possible. They don't need to search, they don't need to think about what they want to watch, they just consume. Fully automated consumption with 0 human effort involved.
Well, almost. There's one last thing remaining: you still need humans to produce the content that you then put on people's feeds. Or rather, needed. Now that the actual production can also be automated, those platforms no longer need to put effort into finding existing human-created content that will keep people watching - they can just generate new, algorithmically perfect content. This is their endgame.
Other advantages to generating content: (1) fewer copyright issues. (2) No creators to pay, just GPU bills scaling with the use of the platform. (3) a much smaller critical mass.
Given these advantages I expect the current "social media" to be replaced with a new one, rather than them pivoting. The next big thing after tiktok might be something that only has generated content, where a last final bit of "social" is taken out of "social media".
Generating stuff is very cheap compared to building and training the model. When you have your model done you're incentivized to use it as much as possible. Maybe even considering the sunken costs.
Why can't it be the actual tiktok? By simply winning competition with humans i.e. whereas vast majority of humans see their pay go too low to bother to continue?
There is no “end game” it’s just hustlers out for a buck for themselves — YouTube, the slop makers, the ad companies, the bots scraping videos, all of em.
Same as it ever was. When this cash cow proves worthless or runs out it’ll be another thing.
> What's the end game?
What makes you think that there is an "end game"?
Someone figured out how to make computers be able to create content that is costly to distinguish from human-made content. Someone else is using it to pump out AI slop in the hopes that it will make them a quick buck. The platform becomes unusable for anyone that values their own sanity. No "end game" to be found.
AI will be the worst thing that happened to society in a very long time.
Its channels from Spain, so presumably appeals to Spanish speaking countries.
the 'slop' is generally at either end of the extremes of video length, either shorts or multiple hour videos.
shorts get paid by the view, ppl put on long videos to fall a sleep to and youtube premium does a rev share based on watchtime of the premium user.
this is why you have like 10 hour playlists and white noise videos.
I have noticed people on the subway watching them. I sneaked a peak on a few of their phones and it was legit AI slop with clear signs (for me at least) that it was AI generated. The end user (viewer?) seemed hooked but they are mostly shorts (10-20sec videos) and you can see their fingers swiping to the next one.
The other day my mother told me if I watched some random AI slop (Putin getting in a physical fight with Trump) and I asked her why she watches this stuff and her answer is that it comes up in her feed. She said it was funny.
I don't know what to make of any of this.
People do not want reality; they want to comfort.
Same for podcasts (and other types of online contents) -
Here's a dataset of 26,000+ ai-generated "podcasts"
https://www.kaggle.com/datasets/listennotes/ai-generated-fak...
Every time I watch a video it’s like I’m running a Turing test.
Mispronunciations could be a giveaway, but then some people may have naïve pronounciations.
So many videos about nerd-sniping niche subjects.
As though we need to have a new regimen of thought discipline since so much could so easily be list.
Don't care. I use my own domain index, which also contains youtube channels, which I use through RSS.
https://github.com/rumca-js/Internet-Places-Database
Most important links are available through a simple search
https://rumca-js.github.io/search
I have also RSS search
https://rumca-js.github.io/feeds
Fake Brian Cox and Richard Fynemen are in abundance.
Imagine 50 years down the road impossible to tell which things Richard Feynman really said in his lectures and which are all made up.
I just blocked a fake Brian Cox one a couple of hours ago. There can be a certain time waste looking at the things before figuring they are rubbish.
Youtube should give you more options than just block or don't show me this. You should be able to click 'AI fake of real person' so they don't get inflicted on others unless they like that stuff.
Unfortunately it's not possible to block a channel on YouTube, or I'd have blocked thousands. All you can do is tell it not to recommend a channel.
Google’s two moats, YouTube and Search, both succumbing to the same infestation to destroy its market grip I see.
Ah, that's what the TPUs are for.
What Youtube pushes vs what I watch are so completely different that I am shocked. I almost exclusively watch geology, space, tech and cooking and it pushes at me (quick check of the home page): disgusting cow medical procedures, random sexual videos, celebrity/influencer junk. Oh, wait, there is a geology thing....and it is a junk sci/scare video that has no actual geology related content. Yeah. Their algorithm is clearly working well. I would go to other platforms but there isn't one. I have started donating on patreon and watching there where practical (even though it is still yt serving it up) but that doesn't easily allow me to discover new things. The real challenge here is finding new things. When search is so completely broken, how do you find valid new things to watch?
Out of curiosity, do you have watch history enabled/disabled?
I found the feed with it enabled is much better than disabled and I have it finetuned to be more in line with the niches I care about.
I am also very proactive with marking channels or content I don't prefer with the "Not interested" or "Don't recommend channel" as well as going through and pruning any content I don't want from my watch history directly.
It's not perfect but it's orders of magnitude better than my logged out or watch history disabled account (though not sure if they have since updated it to not show anything at all)
It’s sad that we replaced manual curation with recommendation algorithms because they were doing the hard job of surfacing relevant content for us and the end result is that you’re spending more effort and intellectual energy into steering the algorithms in the right direction than we ever did before
Manual curation would be waste of time. Why should I spend hours looking through slop when I can get algorithm do it reasonably well.
Youtube actually works reasonably well, as long as you remove from history content you disliked, use do not recommend channel and possibly not interested option. It might not be highest of quality, but at least content stream then is passable.
It's the same with financial content.
I follow N YouTubers who produce genuinely useful and insightful videos on the state of the market and personal finance etc, but the algorithm just recommends people with stock tips and scaremongering junk
Probably there not enough content. Just the same repackaged in human or machine produced slop.
People are addicted to YouTube but I think the key to the healthy watching habits would be restricting the screen time.
Someone knowing about things he's interested in has few problems separating the new and informative content and if he had, say, two hours a week for watching he'd probably enjoy what he sees.
Two hours is just an estimate I came up with, it can be an two hours a month or hour a day. The important thing is that YT just doesn't create enough real new information and after that it is just slop and brain rot, regardless of your habits and filters.
Not only that but also the same short videos keep repeating. I have tons of great suggestions from Youtube on the main feed from creators I like, yet the Shorts feed is almost 90% garbage and AI slop.
It's deliberate, I'm sure. People say they want the vegetables, but then go on to watch hours of fast food / Shorts. Clearly the algorithm knows.
do you have an account about which "you care somehow", like banning/disabling specific channels explicitly? Ive found out: If I tell youtube what I want by sometimes thumb-up/down and explicitly clicking on "no more vids from this channel", then the results a) much better and most of the trash is removed.
Though, new AI-slop channels popup some days - esp. new/fresh channels, but these you can ban anyway since most new channels are AI slop. or "XY ... official" or "XY .. best of" or "XY .. clips" or "XY... fails"
I honestly haven't tried putting effort into the platform for a while now. I have very targeted things I watch (links to the channels) and go direct instead of navigating the homepage at all. I think my upper confidence bound on the value of the YT homepage and improving it is pretty low so it will be a while until I give it a try again.
Do you a favour an try it :-)
If you somehow curate your YT account, the shit wont show up anymore (only new shit not banned already) and you will see one thing: A curated youtube feed is actually rather empty :-D
On some days, I really have to actually look precisely if there is something new at all: Im a subscriber of 30-35 technical/business channels - its very rare that i could "spend a whole day watching YT", because there is just not enough new stuff.
Before curating my account, I also had the impression that "its endless and not any sort of structure" - but it has.
HN’s front page isn’t 20% AI-generated—sorry, “cleaned up from my notes”—but it’s a lot closer than it was a few weeks ago.
The problem is that it’s working; HN voters seem to love the stuff.
That ain't exactly a study as people think of, something scientist do and that get published in a journal, peer reviewed, approved, etc.
It's just kapwing employee checking YouTube channel views and reporting it, same for the feed, so I can't say that the 21-33% number can be trusted.
Now the fact that YouTube has ai slop isn't new, but my bet is that many of these 'subscriber' are in fact bots used to inflate their numbers.
And on a more personal touch, I suggest you install YouTube unhook extension to your computer AND the ones of your relatives. The one less tech savvy that are the more susceptible to fall for this. I surprised my father once watching these kind of crap, he couldn't understand it wasnt even human made. Now I know he's safe from at least that.
Breaking this down by country must be the least interesting analysis possible. I'm much more interested in a breakdown by video genre and the type of person watching.
Youtube is horrible, specially the force-fed "shorts". Its 95% AI slop with the same generic graphics and voice-over. From the actual videos most are also AI generated, making me quit the video as soon as i see it.
YT has hit rock bottom. Just sad.
My shorts are mostly of the creators I follow or adjecent creators. I do get the occasional AI slop, but it looks more like the algorithm is testing someone new, to learn what they are, than the algorithm feeding me slop.
Now Facebook/insta shorts, they are somehow just trash. But maybe that is because I don’t follow any creators on those platforms.
At a societal level [if that's what we should be worried about], it's not what percentage of results are slop. It's what percentage of people believe the slop. Arguably, that seems to be decreasing proportionally. What I mean is: Slop is on a parabolic, hockey stick upward, and trust is on a logarithmic decline. So, good? - let it blow itself out? Bad information is worse than no information, and evolutionary pressures have many ways of proving that.
Skimmed through the article, some interesting numbers but not a single statistic is per capita (or per million, whatever). How do I understand the scale of the phenomenon without the per capita figures? Sorry but seems a bit useless.
AI is OK. I used to ignore AI stuffs on YouTube but in recent months the channel “Napoleons Hill Notes” helped me tremendously. It is an AI voice reading channel.
Napoleon Hill, the well-known crook? https://gizmodo.com/the-untold-story-of-napoleon-hill-the-gr...
Not to say that some of his work is not useful though.
That's fair. I don't even know whether they are authentic NP notes (never read his books), but the channel is the only thing that helped me to be very close (a few days/a couple of weeks away) to complete a non-trivial learning project out of school -- first time ever in my life, how weird that sounds. I guess as long as it works it's fine.
I saw one Warren Buffett video where the voice was Buffet's and he said "in this video I will show you...". Most videos featuring Buffet's voice are snippets from actual interviews etc, but this one threw me off. I stopped it immediately. Sigh.
It has really become horrible in the last some days, probably weeks now. It is not just fake-videos AI generated, without Google having any decency to mark it, as it wastes my time and the time of others - but there are now also videos where some videos are real (I know because some of these videos came from years ago), mixed in with AI fake crap. Now, I am able to spot many AI videos, but I bet many other people simply don't have the knowledge. That is also a generational problem, where people have a harder and harder time to separate real from fiction. But I have no desire to waste my time with fake, so Google now has started to kill youtube. I still have use cases of youtube without this problem, e. g. good music (here, whether it is AI generated or not, makes no real difference IF the music is good; but most AI music is crap anyway but I can not listen to it and only focus on good music), but this is getting more and more of a dead end here.
Google already killed its search engine and other things. It is continuing on its path to now kill Youtube. And, mind you - Youtube already had problems before AI. Many content creators felt violated and abused by Google. I really think we should end Google as a company - it is not doing the world any good now. It changed completely; the old Google is permanently gone. Nobody needs the AI slop infected money-milking-via-ads machine.
Also, Google further ruined its already by-now-total-crap search engine, with crap videos nobody really cares about in 99% of the cases. Or the "others searched for xyz" - what the heck do I care what others did? If I want to find something, I don't want google to distract with excuses. Google abuses people here. It is an EVIL company now. These are not "accidents" - this is deliberately aimed at wasting people's time. I want compensation money for Google wasting my time here. This has been different in the past, so it is 100% Google's fault. No more excuses here.
Google, you are the guilty party.
"Deepfake Yanis Varoufakis Videos Are Flooding YouTube": https://www.youtube.com/watch?v=v1ZewbOd2JQ
There are already a lot of impersonating AI-Slop videos appearing, not just faking Yanis Varoufakis, but also many other political commentators. It's hard to find the real videos by now.
I tried to flag these videos, but the process of doing so is so cumbersome that I finally abstained from it. AI-Slop is slowly destroying everything: books, youtube, education, in the end everything that is data driven... Where it could be useful, e.g. high quality video translation, it fails utterly.
I noticed that too and it’s kinda scary. Soon we will have the opposite of canceling, where the target will be deepfaked to say everything and its opposite to nullify their signal to noise ratio.
Yeah, I hate that also. I just subscribed to Yanis Varoufakis‘s official feed and installed a Safari extension to block everything on YT not in my subscriptions. I think this will work out: use the YouTube app occasionally to discover new material to subscribe to, but usually access YT on Safari using this extensikn.
I'd be more concerned about just slop, whether AI or human created. And the fact that Youtube content is overwhelmingly slop - regardless of creator type - is not news at all.
There's good stuff too if you are a bit selective.
Reminder that you don’t have to play by YouTube rules. Unsubscribe from everything and disable video history, that also stops the slop recommendation engine.
YouTube search still works, you’ll still come across interesting YouTube links and if there are channels you want to subscribe to there is a way to get them in your RSS reader. You can also get the feed to be full-length videos only, ignoring all the noise coming from shorts.
> YouTube search still work
Does it? Last time I tried YouTube search it was useless. It showed two or three relevant results, then a bunch of shorts, then a long list of totally unrelated "you might also like" results.
How is YouTube going to deal with all the storage of these videos?
Maybe they will avoid the storage cost by generating this slop on the fly; who would notice?
AI compute costs far outweigh storage costs.
Works as intended. My YouTube feed is 0% AI slop.
Same as mine. I just don't use YouTube at all and search for information on alternative sources, preferably in text form.
Could you elaborate? What have you done to achieve a 0% slop ratio?
Turn off watch history and there’s no feed whatsoever, they started to refuse to show anything on the home page with watch history off since a couple years ago. Yes that includes even Shorts, you’re not allowed to doomscroll without watch history.
> Recommendations are off
> Your watch history is off, and we rely on watch history to tailor your feed. You can change your setting at any time, or try searching for videos instead. Learn more.
Same. This is the way. If you use Firefox, install the "unhook" extension to remove all the junk (https://addons.mozilla.org/en-US/firefox/addon/youtube-recom...)
UnTrap for iOS works practically the same.
> they started to refuse to show anything on the home page with watch history off since a couple years ago.
I remember when that happened. It was hilarious to me because the way they phrased it felt like they think of it as a punishment or incentive, but not having anything on the homepage is incredibly refreshing, exactly what I wanted. YouTube’s homepage (at least in my country) is just garbage videos no one should be watching.
Your recommendations depend on what you watch. If you don't watch AI slop, you will not receive AI slop recommendations. This is what works for me.
> This is what works for me.
Than you are the lucky one.
E.g.: Just for entertainment I recently searched for information on a new LLM model. The result was filled with AI garbage. I stumbled upon videos with AI generated content, presented by an AI generated voice and an AI generated human. And if it is not something like that, YouTube at least lies to me about the language of the video and tries to make me listen to an AI voice which is equally vile.
There are times where I constantly have to close videos after a couple of seconds. It is unbearable. At least they should mark these videos clearly. At best they should allow me to filter them out but of course they won't do that or else they couldn't make the experience on their platform even worse than it already is, what seems to be their real goal.
That's a skill issue on your part. Do not use YouTube to search for information. If you need to find information on a new LLM model, look for official documentation from the model's maker. This applies to everything.
> That's a skill issue on your part.
No, it's not. I search for videos and i get AI garbage as a result. It shouldn't matter what i search for or if you think that it is dumb to search for it. It shouldn't return garbage in the first place unless I explicitly ask for it.
> Do not use YouTube to search for information. If you need to find information on a new LLM model, look for official documentation from the model's maker.
That is what I do but I like the entertainment factor of some YouTubers. I know that I do not get the best information from them but sometimes I'm entertained. That's why I'm using YouTube after all. Entertainment (and I'm not entertained by AI).
> This applies to everything.
So you are saying: In this glorious new AI world I need a search diploma just to evade the endless AI slop YouTube is filled with nowadays. And than there are topics that are taboo to search for at all. From what angle do I have to look at this to discover the good part?
I'm guessing you're just bad at identifying slop. It's EVERYWHERE on YouTube and I find it hard to believe it's recommending slop to everyone but you.
I just didn't watch AI slop when it popped up and it doesn't pop up.
The only place I see slop on YT is ads.
I avoid watching any AI content, and I still periodically have music channels where every video has view counts in the low hundreds get recommended, with a description implying it's AI.
Do you define "AI slop" as "easily identifiable as AI"?
My wife and I pay for YouTube and the prevalence of AI Slop definitely reduces the value we get for our money.
YouTube has done a fairly good job over the years of providing useful recommendations but lately I always immediately ‘back out’ of any AI generated videos or materials with robot-voiced narrations. I keep hoping that the algorithms quickly learn to not show me this stuff, but in the meantime I have had to change my viewing habits to following specific channels - and this reduces ‘discoverability.’
YouTube promotes slop and "brainrot" irrespective of AI use, this through the client being tuned for clickbait title presentation, forcing back shorts after every relaunch and promoting clickbait select single frame thumbnails.
AI slop is a significant improvement over countless videos in the past few years that were little more than slide decks with bad TTS. The current equivalents are better written, better spoken, better visuals, etc.
If YouTube slop is in that range, Instagram must be 50-60% and Facebook 70-80%.
Same is true for Pinterest. Try to find human generated interior design images on Pinterest. It’s flooded with AI generated content.
I thought it was simply the way all youtube videos are now made :) my feed has little but those, i need to search for something specific to see anything else.
Matches my experience
And ~90% of google searches now gives you 10 videos as answers to any query rather than just the web pages that have been turned into videos because youtube makes them more money than regular ads. Assholes.
Kagi search. I've never looked back. It's what Google used to be and should have stayed.
HN is the ONLY place I've seen Kagi recommended as useful. So much so I bought it for a year and had to move back to google. I was missing so many search results.
Maybe I didn't configure Kagi enough? But I don't want to configure a search engine.
This is interesting to me. I've done basically no configuring (except frequent use of a "site:reddit.com" lens) and it's been great. Now when I see Google or Bing results it's jarring how bad they seem.
For search I'm increasing just using Claude + web search tool, it's better than Google has been in years. I have my own local ollama-powered RAG setup as well which isn't as good but may serve alright if/when Anthropic starts with ad insertions.
>I was missing so many search results.
Have you considered the difference between quantity and quality?
I find doing more focused searches with bang patterns work well for me, I.E searching wikipedia, manpages, mdn etc directly. Sometimes coupled with LLM to more quickly understand the syntax because docs tend to be sparse with examples.
I still miss something for quickly finding random facts like "how long to boil an egg" but for programming the above works quite well.
>I still miss something for quickly finding random facts like "how long to boil an egg"
All you're missing on Kagi is a question mark:
https://bayimg.com/LaOpNAAbJ
I evaluated Kagi a couple of months ago, but its results were swamped with AI slop, too. These days, it's mostly DuckDuckGo for me.
DuckDuckGo has also been my goto for years, but it is also getting swamped with SEO-rigged spam sites.
2 days ago I was making multiple searches looking for the websites of specific hotels and spa facilities, and none of them showed up in the first 4 pages of results, even when searching by exact name.
Out of desperation I switched back to Google, and, surprisingly, it was willing to give me what I was looking for on the first page. (But not as first result, of course..)
> DuckDuckGo has also been my goto for years, but it is also getting swamped with SEO-rigged spam sites.
True, but DDG seems, at the very least, no more afflicted than other engines. If I'm supposed to pay for a search engine, I expect something better than DDG.
Kagi systematically doesn’t return any good answer whatsoever. It’s such an awful quality that I can’t imagine anyone seriously promoting it, therefore I bet these comments about Kagi are advertising posted by bots. It’s not possible otherwise.
I find it to be more reliably useful than the other non-Google alternatives I’ve tried. I find its “PDFs Only” filter to be awfully handy too. Proudly not a bot, but I freely admit that, because I pay for it and our incentives seem aligned, I’m biased toward giving it the benefit of the doubt.
Always open to new horizons, though—are there non-Google search products that you find to consistently work better?
Definitely not a bot (you can check my comment history) and I can recommend Kagi.
I share the same sentiment. I know many people praise Kagi, and I respect the effort behind it too. I tried it for three months and realized it was not for me. Google works just fine for my needs.
Haha… no bot here. Been using Kagi for years now. Not sure what you‘re searching for. My own tests, admittedly early, found no instances, where Google gave better results.
Happy user of kagi for several years. This is the opposite of my experience. Your comment strikes me as dishonest.
I am also very happy with Kagi's search result and suspect that someone is just trolling.
Holding an opinion like “everyone apart from me is a shill or an idiot” is not an attractive quality in a human.
> And ~90% of google searches now gives you 10 videos as answers to any query rather than just the web pages that have been turned into videos because youtube makes them more money than regular ads. Assholes.
Which also reinforces already declining literacy rates.
These changes are wrecking the independent web. It's crazy that people don't see LLMs as the monopolistic land grab that they are.
The most insidious aspect is that tech CEOs and influencers keep parroting how "AI" will bring global prosperity and empower the masses. When it's clear to anyone with an ounce of reason that the exact opposite is happening. The amount of hypocrisy and manipulation is sickening.
I feel like the search result of YouTube videos on Google search is much worse than the result on YouTube itself. It's strange because Google develop both Google and YouTube search. It's like reverse Reddit, where the website's search is so unusable you have to search on Google like "xxx reddit".
True that. It’s annoying especially when you really need answer quickly. I often then use temporary chat to ask ChatGPT.
Gemini seems heavily tuned to return YouTube sources too
Yeah honestly it really makes me want to use Claude
Gemini can be a powerful model but it's just infested with the stink of Google
Example?
Any google search I do.
"how to configure arducopter gps"
9 crappy videos before they figure maybe they should link to the documentation after all.
To give them some credit, at least that search did not also come with a crappy AI summary that is broken, not applicable and wastes valuable space and bandwidth.
edit: corrected count, I missed one. And a very prominent 'see all' link before the actual one.
I searched for that exact phrase out of curiosity because I haven’t used Google in years… and it’s even worse than you say.
I got an AI summary that takes up half the screen, which doesn’t even give the right answer, then 5 YouTube videos with thumbnails and extra crud (most not even related to the question but just mention “ArduPilot” somewhere on the title), then half a page of “Other people searched for..”
Then about 3 “screens” down the page I get the ArduPilot homepage and then a bunch of embedded Reddit/Facebook discussions about ArduPilot in general, none about setting up GPS.
Google has *completely* lost the plot.
There’s a Web tab which you can click to show only the web results, none of the AI, videos, other people, etc.
I presume you want that to be the default, but I’m sure you understand that Google can’t grant your wish, because it’s subject to market forces.
For me, in Ireland, tacking "-ai" on the end of Google searches disables the hallucination engine. For now at least.
It seems to differ extremely by account or region?
I have to say AI summaries work pretty well for me on Google and my overall subjective sense is the results have been getting better the past 1-2 years.
Here's what I get for your exact search and it looks pretty reasonable?
https://postimg.cc/6y8p3XH7
My search results were the opposite: 1. an AI summary 2. a link to the documentation 3. 4 videos that seem legit 4. a bunch more web links
So you're not talking about AI generated videos but just regular videos by hobbyists that you think are crap?
I knew you'd come back with a comment like this one. You never argue in good faith. No, I wasn't talking about AI generated videos because then I would have written a different comment. Bye now.
I searched for this phrase in chrome, first with quotes, and next without.
The very first response I got with quotes was this HN post.
The very first response I got without quotes took me to the docs here: https://ardupilot.org/copter/docs/common-positioning-landing...
I sometimes wonder how much other browsers, countries, or languages play into the search results for a particular user.
Every time I read someone complaining about Google search results, I try to use another search engine for awhile, just to see if I'm missing something. Every time I do this, the alternatives are awful and I shake my head and go back to Google search.
Can I ask why you’re still using Google to search at this point?
There are ppl that depend on me for tech support. What search engine should I install on their phones?
duckduckgo: https://noai.duckduckgo.com/?q=how+to+configure+arducopter+g...
Optionally with a custom CSS rule to block the starting video block: [data-layout="videos"]
Regular DDG doesn't seem much better than Google: https://duckduckgo.com/?q=how+to+configure+arducopter+gps
AI summary and videos dominating it.
AI summaries are actually rather good, in the general case. I know they take away traffic from source sites, but they work well for search users.
Videos, on the other hand, are a cancer.
> AI summaries are actually rather good, in the general case.
The google ones for me seem to be talking about something unrelated to what I'm looking for a significant percent of the time because it interprets it as a much more popular concept that looks like my query if you squint very hard. Other times it pulls up seo-spam tier answers that are plainly wrong.
The best free and mainstream option would be duckduckgo at the moment in my opinion.
I hate to say it, but Zuckerberg is right about where online content is going. The future is just an endless feed of personalized short form AI brain candy.
My own "research" has me in complete agreement. Venues like Etsy are now 33% Ai slop now so why should Youtube be any different?
Yeah don’t use the feed. Use subscriptions- old school chronological list of things the creators you follow published.
I do worry that yt will intentionally break that functionality in their quest for maximum enshitification though
Earlier: https://news.ycombinator.com/item?id=46403805
This is definitely true for my mom's feed because she watches so many random pet videos.
Mine is like ~1%. I'm actually surprised about how good quality the feed algo has been, and for especially showing small creators and older videos too.
Each person has their own filter-bubble, I guess.
Search engines are definitely worse. DDG shows so much AI shit websites now for specific search queries. Google is almost as bad. I'm having to double quote almost everything because the quality and fuzziness of matching is much worse now. I miss the OR operator and groups parens that were possible way back.
This site appears to be an AI slop generator though?
The current utilization of generative video is almost universally horrible, which this article does suggest, so I'm not too surprised there are players trying to differentiate themselves. Slop for thee, not for me.
How can I block AI generated videos (especially AI generated scripts) in my browser? Youtube has to give the viewer an option or the website is going suffer greatly from AI slop...
Why is youtube in general making it so hard to block content or channels? and now they made it harder to clean up the user's own viewing history, it just doesn't make sense...
Do you believe we can automatically detect AI generated videos, without having too many false positives and without spending huge amounts of resources playing the cat and mouse game?
It's not possible and it's not economically viable. Best we could do is some sort of signature to prove that the source of a video is trusted by the proof issuer.
Because quantity > quality, as long as Google keeps making money from your time, engagement and attention. Easy.