2018-06-20

Youtube and the Closing Internet - a Discoverability Crisis

After getting kicked off Youtube, the Blender foundation is now experimenting with hosting their own videos on a peer-to-peer network. The reason for the ban seems to be that Blender didn't enable ads on their content, so hosting their content for "free" was a money-losing proposition for Youtube. While it is debatable whether Youtube actually loses money by hosting popular videos ad free (considering how they still show ads on the video page itself, profit from the network effect and viewer retention that comes with the video, and are able to inject heavily monetized videos into the viewer's follow-up queue) or whether Youtube should maybe just force ads on every creator instead of just banning their channel out of the blue one day, none of these address the root problem.

Monoplatform Dilemma

The obvious answer to cancerous growths like Youtube are decentralized platforms - software and communication protocols that put users in control over publishing and consuming their own content. This idea is what the internet was initially built on. Email is (in theory) still a decentralized solution for sending messages. The original concept of the web itself is nothing if not decentralized, a collection of independent servers sending content to users and to each other over an agnostic network. However, these core design assumptions have since been eroded by technological and political realities. Networks are no longer agnostic. To even operate a public server, you need the support of several infrastructure companies. Operating a website is fraught with legal liabilities and potentially puts you in the crosshair of powerful adversaries both in your own jurisdiction as well as internationally.

Most crucially, just operating a web site is not enough. You need users to consume and interact with your content or platform. In the olden days, users came to and stayed on websites due to word of mouth, a holdover from the BBS days. Then came search engines, they were the first sign of big gate keepers entering the arena to control what users can and cannot see online. In modern times, search engines are also no longer the primary way to publish and consume content. Instead, we are using completely integrated and monolithic platforms. They host our content, and decide who sees it.

Given the obvious conflicts of interest arising from running these monoplatforms, it is no surprise that content creators and consumers are firmly at the very bottom of this new hierarchy that includes media conglomerates, monoplatform providers, their associated megacorp customers, and multiple governments. They have no say in how these platforms work, and they have no real agency to migrate away either.

Discoverability Crisis

While it is easy to see why the decentralized facilities of the web degenerated and consolidated over time, it may be less obvious why the modern attempts at decentralized content distribution are not taking off. Users may be latently dissatisfied, even temporarily outraged at some acute injustice, but no change in online behavior will likely occur.

Decentralized platforms exist, but they tend to be walled gardens themselves. Climbing those walls takes significant effort and technical expertise, and what lies on the other side is bound to disappoint: a wasteland bare of life and content at best, a hellscape frequented by violent extremists and criminals at worst.

Yet, the most pressing deficiency of free platforms and federated sites lies elsewhere. The core value proposition which the monoplatforms offer is missing: discoverability. There is no real alternative to the thoroughly optimized attention and suggestion algorithms that are driving the monoplatforms in a decentralized world. Part of the problem lies in how platforms like Youtube and Facebook tune their software: with unimaginable amounts of behavioral data on their users. No federated platform or small site can replicate that, even in principle.

When you're on Blender's new Peertube site, you have reached a dead end. There are no suggested links, no gentle reminders about other content you have missed this week, no engagement signals from your friends or like-minded peers. You might as well have clicked on a blog article. Peertube solves the problem of hosting the content very well, but it does not solve the nagging issue of a deficient user experience.

Algorithmically, this is an unsolved problem. A naive approach would be to alleviate the dead end issue by featuring content liked by the content creator you are currently visiting, and some decentralized sites do that. However, clout is a hierarchical phenomenon that prevents this from being useful. Sites like Blender are at the top of the clout pyramid and are unlikely to feature smaller channels, whereas smaller channels are very likely to feature larger ones. The graph network arising from this goes in one direction only and tends to have these dead ends. They may be highly relevant dead ends, but still. However, if you allow featuring links to travel down the graph, you suddenly have a big spam problem on your hands (one so big it could potentially make your network completely useless over night).

The more free a platform is, the easier it is for bad actors to manipulate. Even networks weighted on trust and clout mechanisms cannot make this problem going away. Decentralized platforms are pinned between a rock and a hard place, between being barren dead end wastelands and traumatizing hellscapes consisting entirely of shitposts.

Personal agents run by users, tasked with looking at and filtering content, AI systems with only one allegiance (to the actual user) seem like they would be a mandatory component for a solution, but it is less obvious how to make such a thing accessible or even desirable to an apathetic end user.

A Runaway Scenario

This is an issue the internet at large has. It is not only due to Youtube and its status in the web world, it is a sickness that lies at the core of Facebook, the greater Google, and all the other tech/media giants. It is a disease that is spread daily by all internet users, a contagion continuously hardened and improved upon collectively by an incidentally symbiotic ecosystem of legislation, policy, megacorp lobbyism, and pragmatic user behavior.

The internet is caught in a feedback loop that eliminates smaller non-corporate sites and protocols in favor of large quasi-monopolistic platforms who own and control all the content. It seems ironic that free user choice has been a significantly contributing factor to this phenomenon: the more people use a platform, the greater the value proposition of using that platform becomes. This effect is by no means limited to actual social networks. The data gathered by large platforms on user behavior, as well as the voluntary production of free content for those platforms play a critical role in crafting user experiences that compel even if you are not personally linked to the other users. User behavior in of itself, and secondary effects due to platform mechanisms designed to maximize the ability to advertise to these users, both continue to consolidate the grasp of megacorporations over the global information network.

At the same time, lawmakers are working hard on eliminating independent sites and restricting user-generated content. There are two constituencies driving these legislative endeavors. First, post-911 terrorism hysteria, aided by a general paranoia about crime and political unrest, has reached a point of societal habitualization where complete content control is now merely a self-evident and mostly unchallenged goal we strive towards. Much like the speed of light, this goal can never actually be reached, but we can certainly expend an ever-increasing amount of our civilization's energy to get arbitrarily close to it. The second constituency driving direct and indirect government support for big monoplatforms is the content industry, which currently enjoys a privileged position in many western countries: placed on a special rung above ordinary companies and citizens, they are in a position to deputize the legal system with increasing scope. As such, the amount of power this industry wields not only moves legislature to further increase that power over time, but also has a direct effect on the viability of small platforms and can to a large extent dictate the content policies of big platforms.

In cybernetics, a positive feedback loop is a reaction that amplifies itself going forward, a runaway effect that occurs when balancing factors are no longer able to bring the system back into homeostasis. There is not only one positive feedback loop currently driving large monoplatform adoption and small site atrophy, it is a whole network of runaway effects all going in the same direction.

When this reaction is complete, the internet and society at large will be substantially changed. It is not difficult to see how far this change has already progressed. At the same time it is very unclear what a workable alternative to this end state would look like, but the time to try altering this outcome is rapidly running out.