DeArrow – Solving clickbait on YouTube
dearrow.ajay.app
dearrow.ajay.app
...and not because the creators want to make clickbait titles and OTT thumbnails, but because that's what everyone else clicks on, and is therefore what The Algorithm boosts, so that's what they all need to chase.
A third-party database that provides descriptive titles does not have this incentive, but instead benefits from being qualitatively distinct from the OG YouTube experience. (Otherwise, why would people use the extension?)
But yeah for a random "guilty pleasure" garbage video I use incognito too, it's just faster.
It’s like adding an item to your cart, and then having an ad for it at the top of every search result.
Another thing I hate is if you watch a video from a playlist (or the “podcasts” tab on a channel), and go into your watch history to continue the video, you can’t. It just shows a link to the entire playlist
When I open YouTube I’m always distracted by the recommendations on the initial screen, and then I never get around to the Watch Later list. So it’s nice when Watch Later videos show up on the initial screen.
I wish I could tell it to consistently show me one or two Watch Later videos on the initial screen rather than doing so seemingly at random.
I would be 100% A-OK with that if it had its own section. Like there's a "News" section if I scroll down a few rows. Put "You wanted to watch these later" down there as well. Absolutely do not put them in suggestions though.
[1]: https://www.wired.com/story/tiktok-platforms-cory-doctorow/
I see the fact that the YT algorithm is tuned to encourage click-baiting as a symptom of the business goals at Youtube.
It is not "self-mutilating their content" if it is required to survive on the platform.
Sure that's a bit difficult to quantify. Like for example, there was a period where discovering new channels that were right up my ally was something that happened once or twice a week. YouTube would just recommend a really interesting video from a channel I have never seen before, then I'd discover a lot of content that I really enjoy.
Recently, I needed to cruise Reddit and other random sites for recommendation. Then maybe every few months I'd come across a channel that's 6 or 7 years old that's exactly the stuff I watch all the time, and some how I never saw it before despite searching for very similar terms before and finding nothing I was interested in.
For example, I just discovered a channel called asianometry. He's been making videos that are exactly the kind of stuff I wanna listen to when I'm walking the dogs, but some how I came across it by a complete accident through Reddit.
At least in my experience, that list seems to be running the old algorithm that actually pulls videos related to what you’re currently watching, and doesn’t utilize the awful “personalized” algorithm that just recommends that same three videos over and over again.
The worst part is that the existence of that feed in some versions of YouTube means that they’re totally capable of offering the classic 2014-era recommendations that people really liked across the board, they just...don’t. For some reason.
How do folks who compare YT with itself justify the explosive growth YT has experienced in that timeframe?
In the last 10 years, YouTube has expanded their partner program across the globe and reduced the requirements to become eligible to make money. They struck profit sharing deals with many traditional media companies many countries too that drove more people from those countries to the platform, which then drove views for organic users, which made being a "YouTuber" a viable career option in many new markets, which drove engagement, etc.
There is also no surprise that during the last 10 years there was an absolute explosion in the number of 1-14 year olds getting hooked on the platform. That cohort has almost infinite free time. Videos targeting 1-5 year olds literally have hundreds of millions of views. There are channels that regularly do 200M views per video just in tantalizing colors, music and shapes for that cohort alone. Then there is the 8-12 cohort, and the 13-17 cohort. YouTube only reports some of the 13+ numbers because of laws probably, but it's an everyday sight for me to see very young kids in trance staring at "YouTube Kids" videos while their parents do grocery or deal with guests, etc.
And finally, there is no YouTube competition. Facebook tried for a while, but they didn't make a dent. Twitch is for streaming, not video content. TikTok is the only competition YouTube sort of has, and it's still a different format/appeal. YouTube killed short content format a long time ago because they thought people leave YouTube too quickly if they load a page to play 20 second video. TikTok showed them that there is at least 1 billion people who are into that if the UX is right. But other than that, there is no YouTube alternative/competition because it's almost impossible to operate such thing profitably. In the last 10 years more and more people have come online. Video is a very important format, and YouTube has an absolute monopoly on it. Google can turn profit on YouTube now, while also controlling the entire online video market. No one else finds the ROI equation worth it because there is a very little chance they would dethrone YouTube, and it's very expensive and not at all profitable being a very distant #2.
> why would YT even care to stay relevant for the niche needs of some individuals, when it has a massive incentive to provide the world with a better way to watch videos?
Why would YouTube has massive incentive to provide the world with a better way to watch videos? YouTube’s incentive is to increase ad watch time. That’s all. Whatever shape that takes. If it means pushing videos targeted at getting kids to click on them, they’ll do that. If it meant improving relevancy for niche needs of some individuals, they’ll do that too. When you’re only chasing metrics like that, things usually go in cycles.
The numbers tell you longer videos == more watch time == more ads == more ad watch time. Your goal becomes clear, push more longer video content. Force people to record 10 minute video to show you how to remove a laptop screws, 9 minute talking about the laptop and what screws are, and 1 minute showing what stickers or labels you need to remove to get the screws out.
Have TikTok come and eat a significant market share of “video watch time” on the internet, which causes advertisements dollars to now split their budget between the 2 of you, so you correct by introducing or pushing short format again.
Assume that fun, simple, entertaining videos have a bigger market == more views == more ad watch time, so push those more and push more in-depth “niche needs of some individuals” aside. Few years down the line, discover that some content is “ever green” and some isn’t, and “ever green” content has a significant market so autocorrect to push that.
YouTube, and most companies, have given up on trying to have consistent logic or statement or vision or mission or plan or any of that nonsense stuff. Just have a model for things, we want to maximize X, what do we need to do to Y. With X being profit only constrained by legal requirements.
YouTube will come around again. It has before, and it will again. It’s just at the moment it’s focusing on short term viral entertainment videos, and trying to figure out the right recommendation algorithm for that.
So my experience is not ad related whatsoever.
Also I'm not reading most of what you're writing here, a Gish gallop is probably not your wisest move on HN.
Maybe if the user didn’t need to create a brand new account to make their point, I’d be more receptive.
Your one or two sentence insinuations instead of statements which can be argued against lay ground for long responses because you, sir or ma'am, are using a dishonest tactic (two actually, add ad-hominem for the new account jab) and I feel that your 'others in HN' would agree with me.
Continually asking easy, loaded questions then disregarding the responses and attacking anyway is a low form of rhetoric.
Nor did I, "disregard the responses." I pointed out missed concepts, which to me were gaping and exposed a fundamental lack of understanding of the problem (how do you not know about YT Shorts when talking about TikTok eating YT's lunch?).
Finally, for what I wrote to be an ad hominem, I would have needed to claim the author's argument was, specifically, wrong as a result of who they were. I never actually said anyone was "wrong" per se; my disagreements with the arguments have been on substance (I mention YT Premium as a counterpoint, for example). I also made observations related to things other than the user's argument, but I never claimed they were wrong as a result of those observations, only that their credibility was injured.
Sorry but if I call a fish a cigarette it is not a disagreement over belief. In this case 'gish gallop' has a definition:
The Gish gallop /ˈɡɪʃ ˈɡæləp/ is a rhetorical technique in which a person in a debate attempts to overwhelm their opponent by providing an excessive number of arguments with no regard for the accuracy or strength of those arguments. Gish galloping prioritizes the quantity of the galloper's arguments at the expense of their quality. -wikipedia
Please tell me how this applies to the response you admit to not reading.
> Nor did I, "disregard the responses."
You specifically said you didn't read at least one of them.
> Finally, for what I wrote to be an ad hominem, I would have needed to claim the author's argument was, specifically, wrong as a result of who they were.
You wrote that you would take them seriously if they weren't a new account. If that isn't 'wrong as a result of who they were' then what would you call it?
You have quoted the fact, which is the definition of the Gish gallop. Your opinion is that it does not apply here, and I disagree. You do not hold authority over how to interpret English, so you cannot therefore declare factually that I am wrong, only that you disagree.
As for taking something seriously or not, that’s got nothing to do with right or wrong; you need credibility to be considered, and without credibility the “rightness” of your argument never makes it to evaluation. Happens all the time in the legal world [1], and certainly not at all related to the innate properties of the person who makes the argument.
Admit your mistake and correct your behavior going forward like a decent human.
I'd argue Youtube got bad when they transitioned from a reverse chronological subscription feed to a primarily algorithm driven front page, which is at least a decade ago.
Even if it didn't immediately turn bad at that point, that's when they committed to the enshittification (though I'm sure one could make a case for an even earlier point).
Taylor Red is a country and western band consisting of identical redhead triplets, who haven't played a concert in years. They apparently run a content farm and a bizarre one at that. Every single one of their videos is a Short with the same title, "wait for it... :scared emoji: #shorts" and they do crazy nonverbal stunts on every one.
As near as this investigator guy could figure, this content farm is designed to have an international reach (therefore, no dialogue in English) and also to appeal to children (even the youngest nonverbal ones) with bright colors and amazing optical illusions.
Yet Taylor Red continues to hold themselves out as a country band. Who knows how this works? Gaming the algorithm FTW!
What really gets my goat is when I flip back to the browser to see something I have already previously downvoted since there is a "thumbs down" highlight. Whatever algorithm is being used, it doesn't have the ability to not play something that has previously been disliked.
Why I find this amazing is because if you told me to write a "Discover Mix" algorithm, the first or last step of that is going to be the only thing I really know. Filter out the manual dislike clicks.
But music? Most people don’t listen or engage with music they dislike, and just skip to the next track. They can’t even serve additional ads that way, at least as far as I know. Unless they’re trying to inflate some made-up “total streams started” metric or something, it doesn’t make any sense to avoid filtering out disliked tracks. It’s purely a net negative for them.
My fav is Countdown at Kusini by Manu Dibango. The soundtrack to a 1976 Nigerian action film that I absolutely love. Totally from left field.
Most recommendations though are just absolutely terrible. Either stuff I would never listen to or David Bowie "Let's Dance" because I am a Bowie fan. The most trivial,useless, brain dead recommendation possible.
At this point I just assume Google is far less competent than what it it seems like they should be.
This is what I notice as well. Before the end of Play Music, I'd uploaded around 30k tracks to my library. I stream music fairly often. So it knows enough about artists I enjoy and can suggest others I don't know.
But when it picks something from an era/genre that should be at least adjacent (let's say early 90's alt rock) it skews so heavily toward the same overplayed tracks I grew sick of hearing on the radio decades ago.
Nirvana? Teen Spirit/Come as You Are
Jane's Addiction? Jane Says
Soundgarden? Black Hole Sun
Then as often as it throws out something modern or classic that I'd never really heard and enjoy discovering, it goes into weird tangents with stuff I dislike and always thumbs-down. If I always downvote an artist, the algo should be able to infer that I don't like the artist. From there, it should pick up the pattern. It never does, and I still get Kid Rock, Dave Matthews, Nickelback, and Linkin Park in there constantly.
I installed a channel blocker[1] extension to block channels from ever showing up.
[1] https://addons.mozilla.org/en-US/firefox/addon/youtube-clean...
Wipe your search (and watch) history - assuming you meant YouTube Search (which is getting worse by the day but I digress); Google Search at least for me doesn’t affect my YouTube recommendations at all.
It seems to me that if you spend all your time telling YouTube its recommendations are bad, it can only respond by giving you worse recommendations. Given your expressed lack of interest in everything it has recommended thus far, have you considered that when you search for something, that might become the strongest signal of interest that the recommendation engine can work with?
not really these days if you ask me. yt has been terrible at suggestions for years. it came to a point where i went on yt almost anxious in anticipation of all those thumbnails featuring open mouthed morons staring at me.
When I search something that I suspect will affect my feed in a bad way, I'll open it up in the container.
The same can be done with a private tab, this just works a bit better.
Not interested
Tell us why!
I’ve already watched this video
Submit
But odten OVER HALF of my YouTube feed is shit that I’ve already watched from channels that I follow, and The Algorithm Almighty won’t take the fucking hint to stop showing me the same shit no matter how many times I click furiously at the incantation.
It’s almost like they really wanti to emphasize the “submit” part.
It's just the same as how the creator's previous work, SponsorBlock, can only put a bandaid on things, because ad incentives lead to creators padding out their videos to 10 minutes, causing the videos to be needlessly long and rambling. Not to mention videos where the entire content itself is an advertisement.
I disagree. There are many creators that produce content that I consider high quality that use clickbait thumbnails and titles. If they don't, they have a disadvantage because of the metrics the YouTube recommendation algorithm uses.
Other examples are Rick Beato and LTT.
And the way he addresses this when being asked about it on his streams is just adding insult to injury. It's kind of sad that such a good source of news and general information in that field is being diluted with so much (more or less blatantly) sponsored content.
> It's not just the Linus Tech Tips or the Mr Beasts of the world doing algorithm optimized titles and thumbnails now. Even really great channels like Tom Scott, CGP Grey, and Kurzgesagt are falling for the trap now.
> And I don't blame them! That's what's needed to survive on the platform these days. It's a race to the bottom.
> DeArrow hopes to solve this. DeArrow puts everyone on a level playing field. There is no race to the bottom since no one is allowed to be algorithmically optimized. It pauses the race, and let's the true best videos rise to the top.
While a niche extension of course can't remove all bad incentives, it still brings out the videos that actually are good themselves, ignoring the branding they were put in.
This is incorrect, lots of educational channels use clickbait because otherwise they simply won't be as successful and would not be able to sustain themselves. Veritasium does a video on exactly this: https://www.youtube.com/watch?v=S2xHZPH5Sng
That explains a lot of the experience recently, I was thinking about my old subscriptions and why they haven't made a poke in a while.
I wonder can I make my browser redirect youtube.com to actually open https://www.youtube.com/feed/subscriptions ?
Looks like a straightforward browser extension implementation too for a content script to inject this meta tag for select urls
https://www.w3docs.com/snippets/html/how-to-redirect-a-web-p...
https://chrome.google.com/webstore/detail/enhancer-for-youtu...
What they _don't_ want you to do is just watch a few channels that you subscribe to and leave the website, so they force all this other stuff you didn't ask for in hopes that you stay (which works, hence why they keep doing it).
Or like me: I learned to always click the subscriptions button when I'm on the start page or done watching a video. Had quite a few moments where I saw somebody complain about the home page and my first thought was "did they remove the subscriptions page?"
Edit: To clarify, when I say "clickbait" here, I am using it as a neutral term referring to any materials used to convince someone to click on a link. I understand that it is kind of a loaded term and it is not used anywhere in the extension itself because of this.
The title of this post seems to be taken from the satire title on my YouTube video demo. If you install the plugin and view the video from there, there is a better title. It was meant as a demo of the extension.
Edit: this video shows the old and new styles: https://youtu.be/CcyTzhrN6bM
However, I'm not sure a random screenshot, or any still frame from the middle of a video is a better replacement for a custom thumbnail.
The only benefit I see of something like that is to explicitly reduce irritation of stupid thumbnails, not to improve your ability to find content you care for. And if you get emotionally worked up enough because of stupid youtube thumbnails that you must change them to something that's obviously just as random, you might need more meditation. It's just youtube, it's not worth it.
Linus Tech Tips famously ran an experiment several years ago (and I believe have rerun it a few times since then) and determined that, even though they hated doing it, videos which used clickbait titles and thumbnails got consistently got 20% higher engagement than ones that didn’t use clickbait.[2] I’ve also heard anecdotal numbers even higher than that, and there’s a knock-on effect as YouTube tends to promote channels with consistently high performance way more frequently (this is why big Game Theory uploads instantly and consistently show up in everyone’s recommendations).
This isn’t a complaint about your tool at all, don’t misunderstand me. I just wanted to point out that most of these big channels are their creator’s primary business, and it’s gotten very difficult to be successful at YouTube without resorting to algorithm tricks like these.
[1]: https://support.google.com/youtube/answer/3382248
[2]: https://youtube.com/watch?v=DzRGBAUz5mA
Edit: Ah, yeah, I missed your “It's no one's fault. It's a system that creates a race to the bottom.” text on the page. You already probably know all of this.
I will add an addendum in response to “Many have even started going through their entire backlog, changing old titles and thumbnails to be more attention grabbing and vague.” because I actually know how that got started.
About a year or two ago, the late great Minecraft streamer Technoblade ran into an issue where it turned out that one of the community-made thumbnails on one of his videos had used traced art. He quickly sourced a replacement piece and updated the thumbnail, but then quickly noticed something interesting happen: YouTube treated the thumbnail update almost like a new upload, and that video showed back up in people’s recommendations, accordingly with a massive bump in views. He tried this with a few other videos, sometimes changing the title as well, and got similar results.
Other YouTubers began to copy this strategy, and it was ultimately discovered that updating the title or thumbnail caused Google to “re-crunch” the video in terms of the algorithm, and applying the most popular current title and thumbnail styles were led to a serious increase in recommendations and views, even if the video was several years old by that point. Many YouTubers began to abuse this, updating their videos with “modern” metadata stylings, and since then it’s become standard practice for big YouTubers and has led to a serious increase in growth for them.
Once again, it’s an algorithm problem and not an individual YouTuber problem, but that’s what started the most recent trend of editing the metadata on old videos rather than just applying it to new uploads.
They claim that they only create clickbait because the platform requires it, but that argument is undercut a bit by the fact that they use the same playbook on floatplane.
They are not being honest when they say they have to do it for clicks. They have to do it to monetize attention when it is most important to their brand. After that they have sacrificed any long-term value. This is also why they have to create new content all the time at all costs. It is unsustainable and why youtube creators burn out. I hope they invested their money well before that happens.
> Clickbait is a sensationalized headline that encourages you to click a link to an article, image, or video. Instead of presenting objective facts, clickbait headlines often appeal to your emotions and curiosity. Once you click, the website hosting the link earns revenue from advertisers, but the actual content is usually of questionable quality and accuracy. Websites use clickbait to draw in as many clicks as possible, thus increasing their ad revenue.[0]
In Tom Scott's case, the video thumbnails look flashy but are accurate and representative of the video.
I think a good example of clickbait vs non-clickbait according to this definition would be Jazza vs NerdForge. I stopped watching Jazza because he would put great looking art in the thumbnail slightly blurred out, and the actual video would be about everything except that artwork. Whereas Nerdforge will show a huge piece of beautiful art, and that's what the whole video is about.
Just because someone presents their work well doesn't mean it's clickbait. And this is an age old problem, it's the reason we have the idiom "don't judge a book by its cover". I appreciate the effort people like Tom Scott, Sebastian Lague, and Nerdforge put into their thumbnails, because a picture is worth 1000 words and they use their thumbnails to paint that picture.
[0]: https://edu.gcfglobal.org/en/thenow/what-is-clickbait/1/#
I'm not sure I agree with this analysis
I agree that CGP Grey isn't the best example though, his thumbnails seem to just broadly be those little cartoons as they always were
https://www.socialvideoplaza.com/en/articles/what-is-clickba...
This means if the thumb/title is a bit sensationalist, but the content is still good, it could stay up. This leads to the current situation where slightly annoying titles and thumbnails are used, but the really egrious stuff gets filtered.
Not only is the quality extremely high, users self-regulate: a user submission is first spread to a few random users to see if they approve of it, then more, until it spread throughout. So, bad submissions will go out the window. Are there some submissions like highlights that are just "good enough" ? Yes. Does it need to be perfect ? Absolutely not. Does it need to run through a black box of an AI model that's going to upload 5% of the time absolute crap ? Never.
User submissions will always be better than an LLM (which isn't even appropriate for the context of a video.)
That statement will not age well.
Consider the extraordinary case of AlphaGo Zero. Despite beginning with a random initialization and without any human game data to train on, it mastered Go and chess solely through game feedback, reaching superhuman levels. The potency of feedback is nothing short of magical.
Shifting our focus to humans, what's the nature of our major breakthroughs? More often than not, we serendipitously encounter them. They don't typically arise from deduction but from meticulous observation of how our existing theories align with reality. Essentially, we observe and integrate feedback.
Involvement in a larger system - be it the world, society, the internet, or even a dialogue session with a human - is how AIs can transcend the mere regurgitation of the training set. With every interaction, they receive a nugget of new data in the prompt and feedback following the response.
The model doesn't have to be a blackbox either. Plenty of open-source models can summarize a video faster than any human can.
beyond the error rate, the problem with using an LLM vs user-generated titles is that LLM use costs money, and we're not quite at the point of running high-quality LLMs on generic hardware yet. also, realistically titles aren't the main problem that needs solving here
also, do not underestimate petty people: sponsorblock works just fine with user-generated data
finally, the video demo clearly shows that if there isn't a user-uploaded alternative thumbnail, DeArrow picks a frame from the video to use, as is handily suggested by GP who hasn't read the article
That's my method. And you probably could say "people sometimes can't see the reasons" - that's another issue, and there are different solutions to this, but it is a problem.
Edit: live as you want, but I believe that the best way to you live life is to first control your life.
This is another way to say that it is a market of attention with individual actors acting (nearly) independently but with similar convergence.
In a way, that makes it everyone that is a "buyer's" fault. Which unfortunately means this will not appeal to most people (doesn't mean it isn't useful). But we've converged to selecting click bait videos over non-clickbait.
Cliques are a weird thing. People have them memorized and will roll their eyes at you if you repeat them, but this is the epitome of "judging a book by its cover." Repeat it all you want, but there has to be a hook. And anyone properly marketing their work intentionally wants you to judge the work by the cover.
As a bigger issue, I think YouTube fails for "explorers" and concentrates too much on "exploiters." Referring to exploration vs exploitation strategies. ML models will typically do this. Especially because most people are "exploiters."
So I guess I'm also in the “why and how do other people put up with so much crap” camp :)
General patterns:
- "I <verb>ed the <superlative> thing in <place>"
- <thing> vs <other thing>
- Boobs/cleavage in thumbnail
- Stupid face in thumbnail
- Large amounts of money and/or other numbers
I'm going to try the plugin.
- NEVER do this (thumbnail of doing something clearly correctly)
- 12 things you NEED to know before you [buy/do something trivial]
https://i.ytimg.com/vi/nzlSBBVpbS4/maxresdefault.jpg
I actually like 'Pitch Meeting' videos - but their thumbnails are the worst.
Having it be a chrome extension only fixes the problem for a computer, not everyone in a household.
And realistically, it would be nice to extend way beyond YouTube, rewriting all headlines on all articles with a community sourced patch list you can scroll through and upvote the best variant of each title. And then rewrite the articles to remove repetition. Combine articles that say the same thing.
I can almost imagine a web browser, that when it invests a page, uses what it can take from the page to update an internal store of that story with new information.
or if you still want to see them, but with the normal player: https://uscript.mchang.xyz/yt/yt-shorts-redirect.user.js
! YT Homepage - Hide the Shorts section
youtube.com##[is-shorts]
! YT Menu - Hide the Shorts button
www.youtube.com###guide [title="Shorts"], .ytd-mini-guide-entry-renderer[title="Shorts"]
! YT Search - Hide Shorts
www.youtube.com##ytd-search ytd-video-renderer [overlay-style="SHORTS"]:upward(ytd-video-renderer)
! YT Search and Channels - Hide the Shorts sections
www.youtube.com##ytd-reel-shelf-renderer
! YT Channels - Hide the Shorts tab
www.youtube.com##ytd-browse[page-subtype="channels"] [role="tab"]:nth-of-type(3):has-text(Shorts)
! YT Subscriptions - Hide Shorts - Grid View
www.youtube.com##ytd-browse[page-subtype="subscriptions"] ytd-grid-video-renderer [overlay-style="SHORTS"]:upward(ytd-grid-video-renderer)
! YT Subscriptions - Hide Shorts - List View
www.youtube.com##ytd-browse[page-subtype="subscriptions"] ytd-video-renderer [overlay-style="SHORTS"]:upward(ytd-item-section-renderer)
! YT Sidebar - Hide Shorts
www.youtube.com###related ytd-compact-video-renderer [overlay-style="SHORTS"]:upward(ytd-compact-video-renderer)
It also contains rules to block consent dialogs, suggested "mixes" and plenty of other annoyances.
Tap the video. It's even easier to pause that full videos.
>no way to scroll
If you mean seek in BM the video. You can just drag the playhead like a normal video. If you mean scroll to another video you can swipe up.
I only want to see if such a video is coming from a channel I explicitly subscribe to (friends usually).
Personally I BlockTube anyone I see using a soyface/arrow because it is indicative of what kind of uncritical audience they want. I can easily wait a year and get a summary of all those videos in 1 minute total from a respectable intellect. There is no FOMO.
This is for the small niche of the public that wants and cares about sane upfront media.
Unless there was something like sampling, similar to how shazam recognising music from a few seconds works.
Could someone only get a portion of the content and, semi accurately generate a summary?
https://addons.mozilla.org/en-US/firefox/addon/clickbait-rem...