While interesting... Federation is not about equal distribution so measuring its success based on that doesn't really make sense. Federations are self-governing ponds which interact with others....
While interesting... Federation is not about equal distribution so measuring its success based on that doesn't really make sense.
Federations are self-governing ponds which interact with others. You are meant to trust people because any other way would make life really hard. The advantage is that you can talk to people who trust other people to maintain the servers for them while you still get who you trust with your server, whether that's you, a friend or just someone who seems trustworthy.
I disagree with this article on a lot of points. First things first - it's not that in a decentralized system you expect everyone to host their own servers, you, instead, expect an ability to talk...
I disagree with this article on a lot of points. First things first - it's not that in a decentralized system you expect everyone to host their own servers, you, instead, expect an ability to talk to other servers. Web was built on standards meant to cross-communicate with each other - HTTP, email, XMPP - it's about having an option to disconnect from a majority - populated server and branch out, yet keep the connectivity. This is a real democracy, much more real than any other option - because even though you would always get the major servers there is leverage for any given end user over them - all you have to do is move elsewhere for the situation to change. It is especially true for chats and communication platforms, where, as the article truthfully states, we are locked up with our friends and family - you can't break away from the tyranny of Google Hangouts, for example, even if you don't want them to touch your data with a ten feet pole after project dragonfly. They used to have GTalk, that used to be XMPP based, and it could happily communicate across the XMPP network, but they killed the cross-communication. This story repeated itself with AOL, ICQ and a lot of other IMs. Even in the article the author acknowledges the inherit benefit of standardization:
True, Mastodon is de facto centralised, but despite the size of the largest instances, it retains the ability to federate with other Mastodon instances. Further, Mastodon is able to federate with other free software friendly networks via a pair of common protocols, creating the familial fabric of the “fediverse”. Centralization and federation can certainly co-exist in harmony to improve efficiency while retaining user choice.
There is no democracy without leverage, and I would never ever trust a company to police my communications with other people.
That said, when we ultimately do centralise, we must ensure that the central organization belongs to us, the users, not to special interests. On this point, both corporate and decentralised services alike fail. Yes, corporate interests are oligarchies, bowing to money, but so too is the decentralised technocracy, bowing to arcane technical know-how. In a truly democratic digital space, participation must be accessible to everyone, not just those with money or expertise. If my grandmother cannot participate in the administration of the technology she uses, her best interests may not be adequately represented by the technology. Whether she is beholden to a corporation or to a system administrator, that is not digital freedom.
Well, you can take my island of technocracy from my cold dead hands! This view ignores the center idea - unless you are setting your own shit up, someone does that for you. Your grandmother still has to rely on people to maintain this "informational democracy" network, and these people would have the power over her. This is an unfortunate reality of life, and you know what can change that? P2P decentralization! Self-configuring networks, all made more and more real by the progress made with block chain and spread of IPv6. Ultimately we want to remove any human interaction with machine logic running the network, because machines can't be corrupt and you can inspect the source code to your heart content.
In any democracy, centralised power is wielded to protect freedom, but in decentralised anarchy, no power is wielded at all. Anarchy fantasizes that freedom protects itself; inevitably, life in anarchy is life in chains, for any freedom gained is temporary as the system collapses under the tragedy of the commons and the paradox of tolerance.
Ah yes, freedom is slavery, we need other people to police us, woudln't anyone think of the children and yada-yada. In a decentralized anarchy you get an equal voice to each other, there is no power grabs, hence no possible collapse. And what tragedy of commons has to do with digital world exactly, especially one where you do have an option to have your private server?
So, in conclusion, a lot of somewhat fair criticism of current state of affairs in decentralized systems, nothing new, really, since all the problems are being addressed as we speak, but with a dash of shilling of a magical solution without any details on how that magical solution is supposed to work.
What point are you trying to make with this? Machines are exactly as corrupt as their programmer.
Ultimately we want to remove any human interaction with machine logic running the network, because machines can't be corrupt and you can inspect the source code to your heart content.
What point are you trying to make with this? Machines are exactly as corrupt as their programmer.
Currently most social media websites have proprietary "algorithms" that are notorious for being so complex and not transparent, that even companies running them can't understand fully how they...
I'd argue that the machine is more corrupt since it's very possible to create biases which the programmer never intended to have.
Currently most social media websites have proprietary "algorithms" that are notorious for being so complex and not transparent, that even companies running them can't understand fully how they work. YouTube being the best example, since the algorithm tries to just maximize viewing time and nothing else. But I'd argue a good P2P decentralized platform should never imply complex machine learning algorithms to recommend stuff or promote it. People are perfectly capable of doing it themselves, in fact most YouTubers and people viewing them are annoyed by how the algorithm often hides videos form their subscriptions or promotes trash - quality content is not the goal there. I fully agree that that practice should be abandoned in favor of fully deterministic behavior and that means no magic numbers, no neural networks and hence no possible hidden biases in the code.
Let me unpack that. They do understand. The issue is more so what the effect is. If I give more importance to watch time, what will happen? That's what's not known to them. We both know that's...
Let me unpack that.
Currently most social media websites have proprietary "algorithms" that are notorious for being so complex and not transparent, that even companies running them can't understand fully how they work.
They do understand. The issue is more so what the effect is. If I give more importance to watch time, what will happen? That's what's not known to them.
YouTube being the best example, since the algorithm tries to just maximize viewing time and nothing else.
We both know that's false, it does factor a lot of variables.
But I'd argue a good P2P decentralized platform should never imply complex machine learning algorithms to recommend stuff or promote it.
Then you'll have people gaming the system. Have you forgotten the past?
People are perfectly capable of doing it themselves, in fact most YouTubers and people viewing them are annoyed by how the algorithm often hides videos form their subscriptions or promotes trash - quality content is not the goal there.
They aren't. Algorithms are entirely needed. There's more content on Youtube than it'll ever be possible to watch if you put hundreds of people watching 24/7 without sleep. You need something to passthrough it to make the service actually usable.
People are mad at subscriptions because they are devalued the timeline of uploads is fucked but not because of other algorithms.
Let me reiterate, they can't predict the behavior, they can't retrospectively tell someone why something happened to some video. In my book it means they don't understand it, and such is the...
They do understand.
Let me reiterate, they can't predict the behavior, they can't retrospectively tell someone why something happened to some video. In my book it means they don't understand it, and such is the destiny of any machine learning system - the more iterations it gets through the less understandable it is.
We both know that's false, it does factor a lot of variables.
I've been hyperbolic there, yes
Then you'll have people gaming the system. Have you forgotten the past?
I haven't and it was just fine. Quality stuff on Newgrounds and youtube got shared around, that's how the content used to promote itself and grow organically.
You need something to passthrough it to make the service actually usable.
But even though these days this whole thing is pretty useless we used to have actual categorization, tags, and simple algorithms (notice how there are no air-quotes this time on this word, since that is an actual algorithm and not a machine learning system) that build graphs based on what people watch that also watched that one video. That's how LastFM used to work, and it worked well enough. I can say for sure that before all that machine learning stuff YouTube incentivized really interesting well produced content without any complex machine learning stuff, and noways it's just minimum effort videos that don't often get shared around because of how mundane they are. The machine learning "algorithms" breed addictions but being on the internet for a really long time, somehow most quality content comes from websites without any machine learning, but rather with people curating the content and sharing it. Tildes is one of these websites.
If you need another site to find the things you want then the site is broken. Not only that but compared to now, Youtube had barely 0.00000000000000000000000001% of the content it has today. Oh...
I haven't and it was just fine. Quality stuff on Newgrounds and youtube got shared around, that's how the content used to promote itself and grow organically.
If you need another site to find the things you want then the site is broken. Not only that but compared to now, Youtube had barely 0.00000000000000000000000001% of the content it has today. Oh and it still used algorithms just like you acknowledged later ;)
But even though these days this whole thing is pretty useless we used to have actual categorization, tags, and simple algorithms (notice how there are no air-quotes this time on this word, since that is an actual algorithm and not a machine learning system) that build graphs based on what people watch that also watched that one video.
Categories still exist, tags still exist and the simple algorithms were gamed to death in the early days of Youtube. As for the graphs based on what people watch - they still exist, they're just much more complicated as to not be gamed.
I can say for sure that before all that machine learning stuff YouTube incentivized really interesting well produced content without any complex machine learning stuff, and noways it's just minimum effort videos that don't often get shared around because of how mundane they are.
Well, considering the watch time bias of their algorithm, it is harder for them to do so but also, that content is a really small fraction of all content.
The machine learning "algorithms" breed addictions but being on the internet for a really long time,
Well, that's simply because of what they prioritize in the algorithm, it's not some diabolic scheme, it just so happens that if you give people what they want all the time then they get addicted.
somehow most quality content comes from websites without any machine learning, but rather with people curating the content and sharing it.
Reddit has all of that. Tildes uses filters for the most part but some simple algorithms are used to determine comment relevance.
Sorry, but are you claiming YouTube's (and pretty much any other of the major platforms being discussed here) focus on viewing time and engagement and the kinds of content that maximize those...
Well, that's simply because of what they prioritize in the algorithm, it's not some diabolic scheme, it just so happens that if you give people what they want all the time then they get addicted.
Sorry, but are you claiming YouTube's (and pretty much any other of the major platforms being discussed here) focus on viewing time and engagement and the kinds of content that maximize those things grew naturally out of giving people what they want? You don't think there's a difference or tension between the interests of users and the interests of these platforms? That the addictiveness is an accident born from catering to users?
I think I'm pretty slim, don't hurt me like that :'( But jokes aside, Sure, we can put a tinfoil hat and say that but I doubt developers thought "Let's be evil and get them addicted", they're...
I think I'm pretty slim, don't hurt me like that :'(
But jokes aside,
Sure, we can put a tinfoil hat and say that but I doubt developers thought "Let's be evil and get them addicted", they're humans like you and me. It's most likely out of advertising interest to keep us on the site rather than an altruistic one, I'll give you that, but the addiction effect was most likely not their intention.
Of course it's not evil for the sake of being evil, it's evil for the sake of their wallets. Why would that not be their intention? If your goal is to make your platform as profitable is possible,...
Of course it's not evil for the sake of being evil, it's evil for the sake of their wallets. Why would that not be their intention? If your goal is to make your platform as profitable is possible, the most profitable thing to do is to make the platforms addictive, and the platform is addictive...connecting the dots there isn't hard.
A company doing what's in their own best interest to get where they are now isn't tinfoil hat-y, it's just how they work.
I don't know of many people who would willingly work for such a company which openly stated that as their intention. Even if only as a cover-up, my theory seems more plausible. But there's no...
Why would that not be their intention?
I don't know of many people who would willingly work for such a company which openly stated that as their intention. Even if only as a cover-up, my theory seems more plausible.
A company doing what's in their own best interest to get where they are now isn't tinfoil hat-y, it's just how they work.
But there's no proof of it, it is tinfoil hat-y, this might as well be the definition of a conspiracy theory.
There's a lot of people employed at Philip Morris marketing and selling cigarettes as well as at my nearest casino. I don't think people are as universally opposed to taking advantage of human...
There's a lot of people employed at Philip Morris marketing and selling cigarettes as well as at my nearest casino. I don't think people are as universally opposed to taking advantage of human addictions as you think they are. At least, not in a way that stops them from doing these things. Most of us don't choose employment based on morals...most/all businesses are shitty and objectionable in some way, especially the big successful ones with the most employees. It's not much of a deterrent.
I don't know what to tell you if the simplest explanation that entirely lines up with what they are rewarded and optimized for doing while also strongly mirroring the practices and strategies of other industries that have done similar things (gambling, mtx-fueled gaming, etc.), I really don't know what to tell you. By the metrics a company is measured by, doing the things they're doing rewards them the most.
Or do you need the owner of your local casino to tell you their slot machines are purposely preying on addiction to believe it? Do you think those are an accident?
Oh, you must not live in Brazil. Sorry, man, but there are a lot of people who love to take advantage of others and some of them boast about it. Our culture is getting better, but people often...
I don't know of many people who would willingly work for such a company which openly stated that as their intention.
Oh, you must not live in Brazil. Sorry, man, but there are a lot of people who love to take advantage of others and some of them boast about it.
Our culture is getting better, but people often brag when they can fool someone, when they cut lines, when they cheat.
People will do anything for profit. Cigarettes are still being sold. Casinos still exist. And piramid schemes.
I've made a distinction between a simple understandable algorithm that can be predicted and "algorithm", which is a machine learning thing that youtube calls that. Of course by this time we all...
Oh and it still used algorithms just like you acknowledged later ;)
I've made a distinction between a simple understandable algorithm that can be predicted and "algorithm", which is a machine learning thing that youtube calls that. Of course by this time we all know it's closer to a neural network and this monstrosity is constantly learning and changing. There is no shame in using fair deterministic algorithms.
the simple algorithms were gamed to death in the early days of Youtube.
Just like the modern ones are gamed to death by abusing any nook and cranny. The fact that people, even ones working for youtube can't predict the algorithm or explain why it does what it does some times it doesn't mean that people wouldn't find ways to fuck with it just by throwing stuff at the wall. We essentially have a battle of 2 neural networks, and human collective neural network is always the most sophisticated and witty so far. Clickbait is as rampant as it always been.
Well, that's simply because of what they prioritize in the algorithm, it's not some diabolic scheme, it just so happens that if you give people what they want all the time then they get addicted.
This might appear so, easy connection - you view what you like, you'll get addicted, but there is a much deeper story here. Most people you'll ask today about the content quality would tell you that it had been on the decline ever since Alphabet took over YouTube - we used to have shorter, quality content and now we get zero to none effort content but on a daily basis and much longer. It's pretty evident if you'll look at YouTube animation community - it's barely holding together. I can recommend this podcast if you want real life numbers and comparison of how animation was and how it is now. These days people have YouTube on background all the time like they used to do with TVs, this clearly translates into watch time and ad revenue, but you have to ask yourself - building a new platform, especially an open one that will propagate itself over blockchain - is it watch time and ad revenue that matters, or content quality? Especially if we can tie it all into some form of cryptocurrency used to run the backend that might be earned by creators.
Reddit has all of that. Tildes uses filters for the most part but some simple algorithms are used to determine comment relevance.
See, simple algorithms, they can do the job adequately. And I would much prefer things based on user interaction rather than indirect metrics, that may or may not translate well into measurable content quality.
The thing is, it's not a fight. They're helping each other out. They both gain from that particular design. Yet we're at a point where Youtube is more popular than ever. I'd say most people are...
We essentially have a battle of 2 neural networks, and human collective neural network is always the most sophisticated and witty so far. Clickbait is as rampant as it always been.
The thing is, it's not a fight. They're helping each other out. They both gain from that particular design.
Most people you'll ask today about the content quality would tell you that it had been on the decline ever since Alphabet took over YouTube - we used to have shorter, quality content and now we get zero to none effort content but on a daily basis and much longer.
Yet we're at a point where Youtube is more popular than ever. I'd say most people are quite happy with the results. Those who want educational content are ultimately a minority of that pond.
but you have to ask yourself - building a new platform, especially an open one that will propagate itself over blockchain - is it watch time and ad revenue that matters, or content quality? Especially if we can tie it all into some form of cryptocurrency used to run the backend that might be earned by creators.
You can't just value "content quality", it's not that simple.. surely you know this, right?
As for blockchain, god please don't use that technology for a video platform, it doesn't make sense at all. Especially at scale, it'll be impossible for anyone to keep up with the amount of data so the blockchain aspect won't mean anything. Furthermore, people can't delete on blockchain technology and that's a very big issue for privacy and accountability.
See, simple algorithms, they can do the job adequately. And I would much prefer things based on user interaction rather than indirect metrics, that may or may not translate well into measurable content quality.
If we had to sort all of Youtube's content by humans like we did Tildes, we would need thousands of people working on the clock to do it and even then, it wouldn't really work. There's billions of hours being uploaded every day. There's more content than you can even imagine.
Not if the code is open source, no. The whole foundation of blockchain technology rests on the idea that once the code is written and the network starts propagating this code is not subject to...
What point are you trying to make with this? Machines are exactly as corrupt as their programmer.
Not if the code is open source, no. The whole foundation of blockchain technology rests on the idea that once the code is written and the network starts propagating this code is not subject to major changes. Think of it as law that enacts itself. So the corruption is done in secrecy - right now Deimos can change any post ever written on tildes, and we wouldn't know that happened unless someone archived the original in time and noticed that it changed. He has this power, and if he uses it or not doesn't matter - it's an option. But on a P2P platform based on block chain technology nobody will have that power, and no other powers. The shadowy backroom practices are omnipotent on currently popular social network, the shadowbanning alone is one of the most evil practices and it's applied arbitrarily all across social media. But if the whole network is broken into encrypted chunks and nobody has the access to the back end split across millions of computers worldwide you exclude that human element altogether, and anyone can inspect the code that actually runs there.
I think the article eventually finds an interesting point with the information democracy, but it treats imperfect decentralization rather unfairly. While email is practically fairly centralized,...
I think the article eventually finds an interesting point with the information democracy, but it treats imperfect decentralization rather unfairly.
While email is practically fairly centralized, it still retains many upsides of federation: I don't actually need to use Gmail, and if I switch to something else, I can still communicate with my friends. In the end the article acknowledges this, but I really think that this is very important.
Compare this to say, WhatsApp: I feel really uncomfortable using it, but I can't switch away because it's practically the only way to communicate with some older relatives. If it federated with XMPP or Matrix, this would not be problem even if nobody else switched with me. And the same with social media: if everything would federate, we would not feel the need to stay on platforms we find unethical or otherwise not good for us.
Federation beats the network effect, and that alone makes it worth it.
This is absolutely the main reason to push for federation, and the main reason I only really communicate over federated protocols. Sure, I do also happen to run my own email, XMPP and nextcloud...
Federation beats the network effect, and that alone makes it worth it.
This is absolutely the main reason to push for federation, and the main reason I only really communicate over federated protocols. Sure, I do also happen to run my own email, XMPP and nextcloud servers, but there are "centralised" offerings for each of those services which means anyone else can still communicate with me without having to be a sysadmin.
I think it's missing the point to say that just because most Mastodon users use one of the 3 most popular instances, that decentralization/federation is failing. (Similarly with email.) To me,...
I think it's missing the point to say that just because most Mastodon users use one of the 3 most popular instances, that decentralization/federation is failing. (Similarly with email.)
To me, federation is about having the freedom to switch providers. If a provider does something egregious or merely something I don't prefer, I have the option to switch providers without burning all of my bridges. This discourages the providers from doing anything too egregious in the first place. Conversely, it also allows providers some room to experiment or be opinionated. Individual providers can easily defend having a radically different UI or blanket policies (like no alt-righters) with the explanation that if you don't like it, you can easily find a technically similar provider somewhere else. The provider doesn't have to deal with so many complaints that they're universally excluding their technology from being used for a use-case.
There is a fair amount of bridge burning when switching instances. Migration, as far as I can tell, is very limited. The only thing you can directly transfer in the migration is the list of people...
I have the option to switch providers without burning all of my bridges
There is a fair amount of bridge burning when switching instances. Migration, as far as I can tell, is very limited. The only thing you can directly transfer in the migration is the list of people you are following/blocking/muting, and you can get an archive dump of your toots. Your followers don't come with you; they will have to manually refollow your new account. Your toots don't come with you either. Once you've built up history and/or a sizable following, switching providers is going to be painful.
There's a github issue for real migration that's been open since Nov 2016.
This is no less true for email; you still have to tell everyone you know about your new email address, in the same way that Mastodon users have to refollow your new account by hand. It's not...
This is no less true for email; you still have to tell everyone you know about your new email address, in the same way that Mastodon users have to refollow your new account by hand. It's not perfect, but it's better than nothing.
Great article. Federation seems good on paper but is cumbersome on reality. Centralization is way more practical for the user. Information democracy is the way to go.
Great article. Federation seems good on paper but is cumbersome on reality. Centralization is way more practical for the user. Information democracy is the way to go.
I would suggest that the de-facto centralized, but over a federated protocol setup the article decries is actually the best of both worlds. It makes it easy to use the service, as there are 1-3...
I would suggest that the de-facto centralized, but over a federated protocol setup the article decries is actually the best of both worlds. It makes it easy to use the service, as there are 1-3 'obvious' hosts of the service which your average user can hop on. And yet, if you decide that you don't like it's administration for whatever reason, you have a choice in switching to another.
Federation is a rejection of all and any control over your data; it simply gives you the ability to leave and know the person your trusting. edit: Thou shall commit adultery
Federation is a rejection of all and any control over your data; it simply gives you the ability to leave and know the person your trusting.
That's a bold claim, I'd be interested to know why you think that considering that anyone can host a server and control their data in (most) federated systems.
Federation is a rejection of all and any control over your data
That's a bold claim, I'd be interested to know why you think that considering that anyone can host a server and control their data in (most) federated systems.
While interesting... Federation is not about equal distribution so measuring its success based on that doesn't really make sense.
Federations are self-governing ponds which interact with others. You are meant to trust people because any other way would make life really hard. The advantage is that you can talk to people who trust other people to maintain the servers for them while you still get who you trust with your server, whether that's you, a friend or just someone who seems trustworthy.
I disagree with this article on a lot of points. First things first - it's not that in a decentralized system you expect everyone to host their own servers, you, instead, expect an ability to talk to other servers. Web was built on standards meant to cross-communicate with each other - HTTP, email, XMPP - it's about having an option to disconnect from a majority - populated server and branch out, yet keep the connectivity. This is a real democracy, much more real than any other option - because even though you would always get the major servers there is leverage for any given end user over them - all you have to do is move elsewhere for the situation to change. It is especially true for chats and communication platforms, where, as the article truthfully states, we are locked up with our friends and family - you can't break away from the tyranny of Google Hangouts, for example, even if you don't want them to touch your data with a ten feet pole after project dragonfly. They used to have GTalk, that used to be XMPP based, and it could happily communicate across the XMPP network, but they killed the cross-communication. This story repeated itself with AOL, ICQ and a lot of other IMs. Even in the article the author acknowledges the inherit benefit of standardization:
There is no democracy without leverage, and I would never ever trust a company to police my communications with other people.
Well, you can take my island of technocracy from my cold dead hands! This view ignores the center idea - unless you are setting your own shit up, someone does that for you. Your grandmother still has to rely on people to maintain this "informational democracy" network, and these people would have the power over her. This is an unfortunate reality of life, and you know what can change that? P2P decentralization! Self-configuring networks, all made more and more real by the progress made with block chain and spread of IPv6. Ultimately we want to remove any human interaction with machine logic running the network, because machines can't be corrupt and you can inspect the source code to your heart content.
Ah yes, freedom is slavery, we need other people to police us, woudln't anyone think of the children and yada-yada. In a decentralized anarchy you get an equal voice to each other, there is no power grabs, hence no possible collapse. And what tragedy of commons has to do with digital world exactly, especially one where you do have an option to have your private server?
So, in conclusion, a lot of somewhat fair criticism of current state of affairs in decentralized systems, nothing new, really, since all the problems are being addressed as we speak, but with a dash of shilling of a magical solution without any details on how that magical solution is supposed to work.
What point are you trying to make with this? Machines are exactly as corrupt as their programmer.
I'd argue that the machine is more corrupt since it's very possible to create biases which the programmer never intended to have.
Currently most social media websites have proprietary "algorithms" that are notorious for being so complex and not transparent, that even companies running them can't understand fully how they work. YouTube being the best example, since the algorithm tries to just maximize viewing time and nothing else. But I'd argue a good P2P decentralized platform should never imply complex machine learning algorithms to recommend stuff or promote it. People are perfectly capable of doing it themselves, in fact most YouTubers and people viewing them are annoyed by how the algorithm often hides videos form their subscriptions or promotes trash - quality content is not the goal there. I fully agree that that practice should be abandoned in favor of fully deterministic behavior and that means no magic numbers, no neural networks and hence no possible hidden biases in the code.
Let me unpack that.
They do understand. The issue is more so what the effect is. If I give more importance to watch time, what will happen? That's what's not known to them.
We both know that's false, it does factor a lot of variables.
Then you'll have people gaming the system. Have you forgotten the past?
They aren't. Algorithms are entirely needed. There's more content on Youtube than it'll ever be possible to watch if you put hundreds of people watching 24/7 without sleep. You need something to passthrough it to make the service actually usable.
People are mad at subscriptions because they are devalued the timeline of uploads is fucked but not because of other algorithms.
Let me reiterate, they can't predict the behavior, they can't retrospectively tell someone why something happened to some video. In my book it means they don't understand it, and such is the destiny of any machine learning system - the more iterations it gets through the less understandable it is.
I've been hyperbolic there, yes
I haven't and it was just fine. Quality stuff on Newgrounds and youtube got shared around, that's how the content used to promote itself and grow organically.
But even though these days this whole thing is pretty useless we used to have actual categorization, tags, and simple algorithms (notice how there are no air-quotes this time on this word, since that is an actual algorithm and not a machine learning system) that build graphs based on what people watch that also watched that one video. That's how LastFM used to work, and it worked well enough. I can say for sure that before all that machine learning stuff YouTube incentivized really interesting well produced content without any complex machine learning stuff, and noways it's just minimum effort videos that don't often get shared around because of how mundane they are. The machine learning "algorithms" breed addictions but being on the internet for a really long time, somehow most quality content comes from websites without any machine learning, but rather with people curating the content and sharing it. Tildes is one of these websites.
If you need another site to find the things you want then the site is broken. Not only that but compared to now, Youtube had barely 0.00000000000000000000000001% of the content it has today. Oh and it still used algorithms just like you acknowledged later ;)
Categories still exist, tags still exist and the simple algorithms were gamed to death in the early days of Youtube. As for the graphs based on what people watch - they still exist, they're just much more complicated as to not be gamed.
Well, considering the watch time bias of their algorithm, it is harder for them to do so but also, that content is a really small fraction of all content.
Well, that's simply because of what they prioritize in the algorithm, it's not some diabolic scheme, it just so happens that if you give people what they want all the time then they get addicted.
Reddit has all of that. Tildes uses filters for the most part but some simple algorithms are used to determine comment relevance.
Sorry, but are you claiming YouTube's (and pretty much any other of the major platforms being discussed here) focus on viewing time and engagement and the kinds of content that maximize those things grew naturally out of giving people what they want? You don't think there's a difference or tension between the interests of users and the interests of these platforms? That the addictiveness is an accident born from catering to users?
That's a pretty bold thing to say.
I think I'm pretty slim, don't hurt me like that :'(
But jokes aside,
Sure, we can put a tinfoil hat and say that but I doubt developers thought "Let's be evil and get them addicted", they're humans like you and me. It's most likely out of advertising interest to keep us on the site rather than an altruistic one, I'll give you that, but the addiction effect was most likely not their intention.
Of course it's not evil for the sake of being evil, it's evil for the sake of their wallets. Why would that not be their intention? If your goal is to make your platform as profitable is possible, the most profitable thing to do is to make the platforms addictive, and the platform is addictive...connecting the dots there isn't hard.
A company doing what's in their own best interest to get where they are now isn't tinfoil hat-y, it's just how they work.
I don't know of many people who would willingly work for such a company which openly stated that as their intention. Even if only as a cover-up, my theory seems more plausible.
But there's no proof of it, it is tinfoil hat-y, this might as well be the definition of a conspiracy theory.
There's a lot of people employed at Philip Morris marketing and selling cigarettes as well as at my nearest casino. I don't think people are as universally opposed to taking advantage of human addictions as you think they are. At least, not in a way that stops them from doing these things. Most of us don't choose employment based on morals...most/all businesses are shitty and objectionable in some way, especially the big successful ones with the most employees. It's not much of a deterrent.
I don't know what to tell you if the simplest explanation that entirely lines up with what they are rewarded and optimized for doing while also strongly mirroring the practices and strategies of other industries that have done similar things (gambling, mtx-fueled gaming, etc.), I really don't know what to tell you. By the metrics a company is measured by, doing the things they're doing rewards them the most.
Or do you need the owner of your local casino to tell you their slot machines are purposely preying on addiction to believe it? Do you think those are an accident?
Oh, you must not live in Brazil. Sorry, man, but there are a lot of people who love to take advantage of others and some of them boast about it.
Our culture is getting better, but people often brag when they can fool someone, when they cut lines, when they cheat.
People will do anything for profit. Cigarettes are still being sold. Casinos still exist. And piramid schemes.
I've made a distinction between a simple understandable algorithm that can be predicted and "algorithm", which is a machine learning thing that youtube calls that. Of course by this time we all know it's closer to a neural network and this monstrosity is constantly learning and changing. There is no shame in using fair deterministic algorithms.
Just like the modern ones are gamed to death by abusing any nook and cranny. The fact that people, even ones working for youtube can't predict the algorithm or explain why it does what it does some times it doesn't mean that people wouldn't find ways to fuck with it just by throwing stuff at the wall. We essentially have a battle of 2 neural networks, and human collective neural network is always the most sophisticated and witty so far. Clickbait is as rampant as it always been.
This might appear so, easy connection - you view what you like, you'll get addicted, but there is a much deeper story here. Most people you'll ask today about the content quality would tell you that it had been on the decline ever since Alphabet took over YouTube - we used to have shorter, quality content and now we get zero to none effort content but on a daily basis and much longer. It's pretty evident if you'll look at YouTube animation community - it's barely holding together. I can recommend this podcast if you want real life numbers and comparison of how animation was and how it is now. These days people have YouTube on background all the time like they used to do with TVs, this clearly translates into watch time and ad revenue, but you have to ask yourself - building a new platform, especially an open one that will propagate itself over blockchain - is it watch time and ad revenue that matters, or content quality? Especially if we can tie it all into some form of cryptocurrency used to run the backend that might be earned by creators.
See, simple algorithms, they can do the job adequately. And I would much prefer things based on user interaction rather than indirect metrics, that may or may not translate well into measurable content quality.
The thing is, it's not a fight. They're helping each other out. They both gain from that particular design.
Yet we're at a point where Youtube is more popular than ever. I'd say most people are quite happy with the results. Those who want educational content are ultimately a minority of that pond.
You can't just value "content quality", it's not that simple.. surely you know this, right?
As for blockchain, god please don't use that technology for a video platform, it doesn't make sense at all. Especially at scale, it'll be impossible for anyone to keep up with the amount of data so the blockchain aspect won't mean anything. Furthermore, people can't delete on blockchain technology and that's a very big issue for privacy and accountability.
If we had to sort all of Youtube's content by humans like we did Tildes, we would need thousands of people working on the clock to do it and even then, it wouldn't really work. There's billions of hours being uploaded every day. There's more content than you can even imagine.
Not if the code is open source, no. The whole foundation of blockchain technology rests on the idea that once the code is written and the network starts propagating this code is not subject to major changes. Think of it as law that enacts itself. So the corruption is done in secrecy - right now Deimos can change any post ever written on tildes, and we wouldn't know that happened unless someone archived the original in time and noticed that it changed. He has this power, and if he uses it or not doesn't matter - it's an option. But on a P2P platform based on block chain technology nobody will have that power, and no other powers. The shadowy backroom practices are omnipotent on currently popular social network, the shadowbanning alone is one of the most evil practices and it's applied arbitrarily all across social media. But if the whole network is broken into encrypted chunks and nobody has the access to the back end split across millions of computers worldwide you exclude that human element altogether, and anyone can inspect the code that actually runs there.
I think the article eventually finds an interesting point with the information democracy, but it treats imperfect decentralization rather unfairly.
While email is practically fairly centralized, it still retains many upsides of federation: I don't actually need to use Gmail, and if I switch to something else, I can still communicate with my friends. In the end the article acknowledges this, but I really think that this is very important.
Compare this to say, WhatsApp: I feel really uncomfortable using it, but I can't switch away because it's practically the only way to communicate with some older relatives. If it federated with XMPP or Matrix, this would not be problem even if nobody else switched with me. And the same with social media: if everything would federate, we would not feel the need to stay on platforms we find unethical or otherwise not good for us.
Federation beats the network effect, and that alone makes it worth it.
This is absolutely the main reason to push for federation, and the main reason I only really communicate over federated protocols. Sure, I do also happen to run my own email, XMPP and nextcloud servers, but there are "centralised" offerings for each of those services which means anyone else can still communicate with me without having to be a sysadmin.
I think it's missing the point to say that just because most Mastodon users use one of the 3 most popular instances, that decentralization/federation is failing. (Similarly with email.)
To me, federation is about having the freedom to switch providers. If a provider does something egregious or merely something I don't prefer, I have the option to switch providers without burning all of my bridges. This discourages the providers from doing anything too egregious in the first place. Conversely, it also allows providers some room to experiment or be opinionated. Individual providers can easily defend having a radically different UI or blanket policies (like no alt-righters) with the explanation that if you don't like it, you can easily find a technically similar provider somewhere else. The provider doesn't have to deal with so many complaints that they're universally excluding their technology from being used for a use-case.
There is a fair amount of bridge burning when switching instances. Migration, as far as I can tell, is very limited. The only thing you can directly transfer in the migration is the list of people you are following/blocking/muting, and you can get an archive dump of your toots. Your followers don't come with you; they will have to manually refollow your new account. Your toots don't come with you either. Once you've built up history and/or a sizable following, switching providers is going to be painful.
There's a github issue for real migration that's been open since Nov 2016.
This is no less true for email; you still have to tell everyone you know about your new email address, in the same way that Mastodon users have to refollow your new account by hand. It's not perfect, but it's better than nothing.
Great article. Federation seems good on paper but is cumbersome on reality. Centralization is way more practical for the user. Information democracy is the way to go.
I would suggest that the de-facto centralized, but over a federated protocol setup the article decries is actually the best of both worlds. It makes it easy to use the service, as there are 1-3 'obvious' hosts of the service which your average user can hop on. And yet, if you decide that you don't like it's administration for whatever reason, you have a choice in switching to another.
Federation is a rejection of all and any control over your data; it simply gives you the ability to leave and know the person your trusting.
edit: Thou shall commit adultery
That's a bold claim, I'd be interested to know why you think that considering that anyone can host a server and control their data in (most) federated systems.
Wow, that was my worst typo; I meant isn't