Slashdot Mirror
YouTube, the Great Radicalizer (nytimes.com)
Zeynep Tufekci, writing for the New York Times: Before long, I was being directed to videos of a leftish conspiratorial cast, including arguments about the existence of secret government agencies and allegations that the United States government was behind the attacks of Sept. 11. As with the Trump videos, YouTube was recommending content that was more and more extreme than the mainstream political fare I had started with. Intrigued, I experimented with nonpolitical topics. The same basic pattern emerged. Videos about vegetarianism led to videos about veganism. Videos about jogging led to videos about running ultramarathons. It seems as if you are never "hard core" enough for YouTube's recommendation algorithm. It promotes, recommends and disseminates videos in a manner that appears to constantly up the stakes. Given its billion or so users, YouTube may be one of the most powerful radicalizing instruments of the 21st century.
This is not because a cabal of YouTube engineers is plotting to drive the world off a cliff. A more likely explanation has to do with the nexus of artificial intelligence and Google's business model. (YouTube is owned by Google.) For all its lofty rhetoric, Google is an advertising broker, selling our attention to companies that will pay for it. The longer people stay on YouTube, the more money Google makes. What keeps people glued to YouTube? Its algorithm seems to have concluded that people are drawn to content that is more extreme than what they started with -- or to incendiary content in general. Is this suspicion correct? Good data is hard to come by; Google is loath to share information with independent researchers. But we now have the first inklings of confirmation, thanks in part to a former Google engineer named Guillaume Chaslot. Mr. Chaslot worked on the recommender algorithm while at YouTube. He grew alarmed at the tactics used to increase the time people spent on the site. Google fired him in 2013, citing his job performance. He maintains the real reason was that he pushed too hard for changes in how the company handles such issues.
This is not because a cabal of YouTube engineers is plotting to drive the world off a cliff. A more likely explanation has to do with the nexus of artificial intelligence and Google's business model. (YouTube is owned by Google.) For all its lofty rhetoric, Google is an advertising broker, selling our attention to companies that will pay for it. The longer people stay on YouTube, the more money Google makes. What keeps people glued to YouTube? Its algorithm seems to have concluded that people are drawn to content that is more extreme than what they started with -- or to incendiary content in general. Is this suspicion correct? Good data is hard to come by; Google is loath to share information with independent researchers. But we now have the first inklings of confirmation, thanks in part to a former Google engineer named Guillaume Chaslot. Mr. Chaslot worked on the recommender algorithm while at YouTube. He grew alarmed at the tactics used to increase the time people spent on the site. Google fired him in 2013, citing his job performance. He maintains the real reason was that he pushed too hard for changes in how the company handles such issues.
For any given video, it will recommend a range of other similar videos which by definition must be a bit more radical or a bit less radical. If you keep clicking the more radical ones, of course you will slowly gravitate up the radical tree. How could it be otherwise? I don't consider it radicalising, it's just providing information. For this topic, this is how radical you can go, this is how far you can take it. Anyone interested in a topic enough to keep watching videos is sooner or later going to want to know, how far can I take this? And YouTube has the answer for you.
This is all retarded. I binge watch some Good Mythical Morning, surprise surprise it's all GMM in my feed. Same if it's retrotech whatever, that dominates the feed. The subscription list is the only savior really. Youtube is garage and only the most braindead of users will be swayed by it's reinforcing effects thanks to it's terrible algorithms.
On the Oregon Cost born and raised, On the beach is where I spent most of my days
Have you ever shopped for something only to have google spam you with ads for weeks on end for the actual thing you bought for weeks after you've already bought it.
Considering how often this happens, it would be helpful if the AdSenseWords feedback form had an option, "I already bought this item and am not currently shopping for another one."
YouTube is NOT garage, comrade. YouTube is video sharing site. No place to park in YouTube.
Sincerely,
Your totally American pal, Brad. #Veteran #JesusIsLord #2ndAmendment #NRA #MAGA
You are welcome on my lawn.
I believe this is the result of the goals used to develop this AI. The evolution of an early neural network is value free - what works wins. The goals set by a commercial organisation will orient the growth towards profit, using the power of clickbait. The videos that are most engaging/radical to YOU will be recommended to YOU. What incentive is there to grow the brain for social good (whatever that means)? Should society impose such guidance? This is a big problem and key to the long term impact of AI. It is the same problem that we have grappled with for millennia, and may bring with it the same painful lessons of war and politics. We play god.
Oh that's awesome, I vote for the "I already bought everything and I'm not ever shopping for anything else in my life" option. I just use adblock software, that's more effective.
Certainly matches my own experiences on YouTube, though I think it's not purely the EVIL of the google that's driving it. I'm convinced that there are also trolls who are loving the chaos and who are strategically promoting their videos to be linked from opposition videos. Less annoying but similar to the original article are extremists who are also involved in strategic promotion of their videos to viewers of other videos that they regard as sympathetic.
However, as gawdawful as the EVIL google and the most EVIL YouTube have become, I'm convinced that Facebook and Twitter are worse. Much worse.
And yet all of these problems could be greatly reduced by the use of EPR (Earned Public Reputation) to gently filter in favor of nice folks. The trolls and other villains can be nudged back under invisible rocks to amuse themselves and the play with the few people who enjoy that form of slumming. I have much better uses for my time.
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
It doesn't matter if you started off with an innocuous video or an extreme one, the suggested videos will lead to the same place eventually. YouTube's algorithm places popular videos in a genre high up in the recommended videos results. Likely because a crowd attracts a crowd. More extreme videos are always more popular by numbers, because clickbait.
This is a side effect of crowdsourcing their ratings by going on views, and perhaps secondarily how many people 'liked' or 'disliked' a certain video.
In any given genre (books, TV, movies, YouTube) the most popular item by numbers is the item that has the broadest appeal. The one with the broadest appeal is usually on the lower end of the intelligence scale. As with cinema, intelligent, artful pieces are typically relegated to small audiences, with the occasional oscar-bait breakout.
So if you put together a system whereby the most popular videos are suggested first, the feedback loop described in the article will happen. The only way out of that is to hand curate the algorithm. And that's the very thing that NO large scale tech company wants to do. The moment they stop automating everything possible is the moment scaling becomes expensive, and they no longer reap their huge margins - a license to print money as long as they can keep it going.
What pisses me off with those retarded algorithms is that they ignore the obvious in favor for always being a day late and a dollar short!
Hey guys if I haven't looked at anything other than a BATTERY for a particular laptop? Odds are really good I'm not looking for a new laptop or I wouldn't be looking to replace the battery, would I? Likewise if I'm watching vids on CPUs or a particular game? Maybe, just maybe, you should show me ads for the products I'm actually watching videos on instead of continuing to show me ads for something I looked for a PART FOR weeks ago?
The really sad part? When they actually used "dumb ads" where they just based the ads on what you were actually watching? I actually bought products because they showed me deals on things I actually cared about, now I honestly cannot remember the last time they got a sale because they never seem to get WTF I'm actually trying to get, like showing me new cars when I'm looking for spark plugs or trying to sell me a POS laptop when I'm looking at games that would never run on said POS laptop they just don't seem to have a damned clue WTH they are doing when it comes to ads anymore.
ACs don't waste your time replying, your posts are never seen by me.
if you want to be noticed the easiest way is to do an extreme ultra-marathon with a thumbnail of yourself being carried away on a stretcher.
Since I started using Video Blocker add-on that kind of click-baiting only gets that entire channel blocklisted.
You know that thing where you watch a video on a topic and lunatics start appearing in your recommended list? Now they show up only once.
Mit der Dummheit kämpfen Götter selbst vergebens
It's hardly new to the internet :
it's been discussed for quite some time that TV channels, specially their News, tend to over-dramatize news and cast dramatic negative light on the world.
Basically, when creating emotion in viewer, their attention increases.
This works even more with negative emotions (fear, etc.).
This attracts more eyeballs to your channels,
giving you more opportunity to sell these attentive eyeballs to advertisers and thus increase your revenue stream.
In Europe, this is more prominent on private channels (mostly paid by ads) than on public channels (partially paid by taxes).
The thing is that, in practice, it has been proven to have an actual effect :
in Europe, watching TV and watching news even more so, has been linked to causing an increased feel of "insecurity, danger, etc."
This is despite the situation in Europe being much better and safer than before.
Criminality rate is globally decreasing, but TV reporting thereof is on the rise.
The neural-network AI used by youtube to process recommendation has simply rediscovered on its own the same results as what was already found on TV :
increase the emotional response of viewer by showing more extreme videos, you attract their attention and thus can sell more ads.
The AI doesn't even really have an actual concept of "emotional response" and "increased attention". The AI only notices that after recommending some video, revenue increase.
If video B is shown after video A : more retention, increased ads revenu. If then video C is shown after video B : again more.
AI the remembers to use the chain A -> B -> C, because that's what increases the parameter it is optimising for : eyeball time sellable to advertisers.
But because of what we know from what was studied on old school TV, that will eventually mean showing more and more extreme videos, because that's what has been proven to work on human brains. This AI has simply "independently rediscovered this fact".
----
For shit and giggles, let's gice a source that's on Youtube itself (and even parodies the usual style of conspiratorial videos). Sorry it's in French, but it has English CC.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
For any given video, it will recommend a range of other similar videos which by definition must be a bit more radical or a bit less radical. If you keep clicking the more radical ones, of course you will slowly gravitate up the radical tree. How could it be otherwise?
It goes a tiny bit further :
- it's been already studied and proven (on oldschool TV) that more extreme content (specially more frightening) increases viewer engagement.
- and engaged viewers will bring more revenue by selling their eyeballs to advertisers.
- (this happens even more on private channels than on public TV).
- thus TV channels, specially news casts, tend to gravitate toward more
The AI neural net behind Youtube recommendation just simply "independently rediscover" what's been studied regarding old school TV.
(while being probably even less aware of it : during A/B tests the algorithm only notice that video on list A tend to increase viewer retention compared to list B and thus maximize ads exposition and revenu stream. it just happens that the videos on this list A are the most extreme due to what we already know of human psychology and past TV studies. The algorithm will eventually automatically build a chain of recommendation of increasing extremeness, because that's what works better for the result it tries to maximise)
The sad thing is that this has been also proven to increase the feeling of insecure.
So, yes, initially the youtube algorithm will show up a variety of similarily themed video recommendations, some of which "must be a bit more radical or a bit less radical". But eventually some of these recommendation will prove more popular and youtube will learn to show them more. Due to how human psyche works, those more successful videos *will be* a little bit more frequently the more radical ones. And thus youtube learns to show the radical more a bit more often (without even having the notion of what "radical" is, only that they are successful). And again, sadly due to how human psyche works, it will have a negative impact on viewers.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Blaming YouTube is like blaming the phone company for telemarketer calls.
Congratulations, you have missed the point quite spectacularly.
It's not that these videos exist on youtube, it's that they are PROMOTED HARDER by the algo.
I do not want your cheap brainburning drugs. They are useless for work. And I am a working man today.
Looks like something a russian bot would say. Well I guess that confirms what everyone already suspected. PopeRatzo is really a russian bot in disguise.
Om, nomnomnom...
Right, but you're choosing from the choices they give you. If there's an issue with the choices, then it's fair to blame Google. That's what we're talking about here.
When I hover over a YouTube recommendation, a vertical "..." control appears, which can be clicked upon to pop up a small menu.
Inside this floating drip, drip, drip menu there are three items: Not interested, Add to Watch later, and Add to playlist.
I've been running experiments on Not interested. First I applied it to every video where the thumbnail contained giant boobs. I like boobs, but there's a time and place, but pressed into my nose all day long—under false pretences, more often than not—is not the time and place.
If it really is machine learning under the hood, in theory YouTube would detect this conspicuous pattern. Miraculously, after dismissing many dozens of these, YouTube rarely offers up thumbnail cleavage any longer. But what did it really conclude? That I don't like boobs? That I don't like videos thumbnailed under false pretences? That I don't like the kinds of subject matter typically bannered under "here be the big boobies"?—for which the "fail" genre servers as the conspicuous anchor tenant. Or did it just run out of booby thumbnails in its primary recommendation rotation? From the outside looking in, it's hard to know.
Then I watched a bunch of chess analysis videos after AlphaZero "destroyed" Stockfish. I decided that I really like agadmator's coverage in general, so I watched some of his classics. By this point, 50% of my recommendation column on nearly every YouTube screen was chess videos. So I started to systematically blow these away with my persistent Not interested assault weapon (more of a musket than a semi-automatic, but you go to war with the army you have). It took about a week, and one- to two-hundred repetitions, but now the chess videos arise in my feed no more.
Then I got interested in the Sam Harris interactions with/about Jordan Peterson (who is not an idiot, and not a puppet of the far right, but very well read, articulate, 50% a clone of my own perspective on life, and 50% the exact opposite of my perspective on life; in short, about the most useful resource presently available to me to drive actual personal growth). It wasn't long before I was viewing Harris's "controversial" interview of Charles Murray. (By merely adding that scarequote disclaimer, a certain faction of the Identity Politics Police have already won.)
You can guess what happened to my recommendation feed after that.
Now, this could have been far worse than it was, because I had long been waging a slow campaign of rooftop assassination of any video containing ALLCAPS somewhere in the video title (especially if the main verb, and most especially the snowclone "x DESTROYS y about z"—if you've already mentally replaced z with "Zionism", YouTube has conditioned you well).
Optimally x and y are selected to maximize brass-knuckle pursuit dynamics. We've all seen this trope on WWE. Back when I grew up in the two-channel 1970s, wrestling was one notch above ultimate pain, variety hours such as Lawrence Welk, Tommy Hunter, Rene Simard, or the The Pig and Whistle, so I endured enough wrestling to internalize all the wrestling tropes for life, while desperately checking back to the other channel every three minutes in prayer, I guess, for the kind of programming miracle—surely on par with the virgin birth (whatever that was)—where an entire show is cancelled and replaced mid-episode (I dreamed this dream week after week for what seemed like years and years).
Brass-knuckle pursuit dynamics is where the black hats have both guys in the ring, while the white-hat's partner distractedly sits the imbalance out (bear in mind, this is Canada in the 1970s, where any given NHL bench-clearing brawl clears the bench right down to the lowest equipment manager, so the 250+ lb muscle-bound white hat going Daisy Daydream while his partner gets two-wayed in the ring already strained the c
I think it is simpler than this. When you allow people to produce their own content and associate freely, they tend to cluster together into distinct groups. Hell, I have a farm and several breeds of chickens. The chicks are raised together from 1 day old and even once they begin maturing will instinctively cluster together into their own types in the coop. It's a little depressing to think on overmuch but this is something very deep in us.