Just as YouTube enabled one-to-many video producers to replace the networked media production hegemony, we are now entering the era of many-to-many video production and distribution, where ultra-niche programming is co-created by communities. The next leap forward will come through a combination of machine learning to automate production, video templating to automate post-production and crowdsourcing content to decentralise the filming.
Originally delivered as a Lightning Talk at the 2050Emergent stage of Spark Festival Sydney 2019.
Good evening, and welcome to television
Broadcasting has been around since the 1920s and the introduction of AM radio. A revelation, it was a one-to-many model of content distribution.
In Australia, commercial broadcast television didn't start until September 1956. The new medium was introduced by advertising executive Bruce Gyngell on channel 9 with the words "Good evening, and welcome to television".
Channel 7 started in Melbourne a month later and by the time of the Melbourne Olympics, Australia already had four channels, while the UK had just the BBC Television Service at the time.
Although lagging behind European countries, whose government-funded television channels started either side of the Second World War, Australia was able to observe the trend in the United States towards a multitude of channels, owned by commercial entities and targeting slightly different demographics.
Broadcasters, although initially feared as a potential failing industry by governments, went onto become hugely powerful entities, reaching into the homes of most of the developed world within 20 years. Television channels controlled the content creation and distribution.
As Leonard Cohen sagely observed in Tower of Song:
"The rich have got their channels in the bedrooms of the poor
And there's a mighty judgment coming, but I may be wrong"
Video killed the radio star
image via GIPHY
Broadcasting and broadcasters carried on growing pretty much unchanged between the 1960s and the 1980s. Then cable arrived. Rather than rely on the restrictive bandwidth of radio transmission, physical cable could be used to pipe television content directly into people's homes.
A sudden panoply of channels became available on a subscription basis. For the first time, niche interests could be served up without fear of alienating mass audiences. Cable specialty channels, starting with channels oriented to show movies and large sporting or performance events, diversified further, and "narrowcasting" became common.
It also reduced the barrier to entry for broadcasters. Stations such as MTV (formerly Music Television) sprung up serving up a daily diet of free record-company produced music videos. The marketing departments of the entertainment industry essentially funded the content creation for the new channel. And teenagers loved it. But FM radio stations hated it and tried to get it shut down.
Big Brother is watching you
Today you'd be hard pressed to find any music on MTV's channels. The network found even cheaper content than advertiser-supplied promo videos was reality television. Why pay stars and actors, when you could shoot your audience being themselves? Real World from 1990 focused on the lives of a real family and spawned an entire new genre of shows.
Big Brother was one of the biggest reality television shows, where audience members vied for the chance to become television stars simply by being themselves.
But reality television, while participatory, was still controlled by network executives. Formats were conceived, series commissioned and talent recruited. The industry controlled the distribution via television cables, satellites and transmission towers.
Television 10 minutes at a time
And then came YouTube to spoil all the television networks' fun. Except they didn't realise it at
first. YouTube was a weird corner of the internet where people posted up short videos of funny things. This was augmented by people duplicating old VHS tapes and uploading 10 minutes at a time. It was democratic in a way, but super clunky to upload and largely the preserve of the fans of the content, who would go that extra mile to make their favourite shows, clips or compilations available to a wider audience.
It was the start of hobbyist videographers to show off their talents, usually with the aim of getting a job working for a television channel.
Now everyone is a microblogger
YouTube led to YouTubers, the first one to many form of broadcasting. Initially either a solo video blogger, a couple travelling or a topic expert, the genre of self-produced video is set to overtake professionally produced content this year. Disney and other studios now control the largest YouTubers in the toy sector, while multichannel networks working with paid influencers have replaced the traditional talent loop of television channels and their stars.
Beyond that, the method of distribution has changed. Today over 75% of all content consumed on Facebook is video content. Big brands and advertisers in the platform already schedule series and what used to be called shows into their content marketing strategies.
Everyone is sharing, everyone is producing content and everyone is consuming content. We are approaching narrowcasting.
So why not a TV channel for all?
Video bloggers and social media influencers are driving the shift from one-to-many broadcasting to many-to-many narrowcasting. Fundamentally, if over two billion people around the world carry a high definition video camera with them at all times, then why can't all those cameras be used to capture and produce content?
If there are 620 million Facebook groups, for example, already self-selecting by niche interests, why shouldn't each have its own television channel. The idea isn't science fiction, with Facebook aiming to be 100% video by 2021, group admins are currently scrambling to equip themselves with the tools to produce video rapidly, at scale but at very low cost.
Decentralise and automate
This is where decentralised video production comes in. Vloggi has been a pioneer in this movement. Our system relies on piecing together atomised pieces of content sourced from multiple locations and users. These are then packaged together into television-style video segments, together with music, captions, intros and credit sequences. Dynamic data is merged with video assets to automatically produce professionally-looking programmes with no editing required.
An example is below, with the Travelogue template made available by our partners Fanmade Films, making it possible for anyone to produce slick mini travel documentaries from their travels. Each may have an audience of fewer than 10, but those video assets then also become available for anyone else to re-use under license to create their own content. So by providing tools to make it enjoyable for users to create content, we are building a library of stock video that can be automatically re-sequenced in future projects.
The ghost in the machine
But how to automate production further? The answer lies in machine learning. Video is currently very poor at being indexed. Machine learning or artificial intelligence in the jargon, has just about cracked the spoken word (think of Alexa or Siri) and has long since mastered the written word, but it still struggles on images. Feed a hundred images into Google, for example, and the success rate of an intelligent description can be as low as 25. Of course, the machine needs to learn through repeated ingestion of new images, hence why you are always being asked whether you can see traffic lights or buses in Captcha verification images.
But video is far trickier, with an average 60 frames per second, running the same image identification algorithms is possible, but would consume so much computing power as to render it unsustainable.
Machine learning is also very poor at motion context. In the example above, from the Vloggi labs, you'll notice that our machine learning algorithm assumes the military flypast is a flock of birds until the very last moment when the Dassault/Dornier Alpha Jets' profile is revealed against the sky.
A human watching the same clip would recognise the distinct flying gait of aircraft quicker. Anyone reading the caption, however, would know instantly. This is where, in our system, we triangulate geolocation with a human-annotated caption to add context to the computation. In our next iteration, we'll be using search engine data based on the timestamp to further deepen the data associated with each 10-second video clip in our library.
Narrowcasting in 2050
Our vision for narrowcasting is already almost upon us. Apps like Cameo already allow you to buy micro pieces of a celebrity's time for ultra-niche programming (an audience of one). Indeed, micropayments is the final piece of the narrowcasting puzzle. People create video content for their friends to enjoy on Instagram, SnapChat or TikTok, but if that content was repurposed, they'd want some compensation. In the Vloggi system, we pay content creators if their video clips are used outside their initial projects. Small sums, but for a prolific videographer, a new income stream can open up.
Our vision is to deliver minutely tailored video content to every individual based on their needs. So a marketer will request "60 second highlights video of happy Japanese tourists on the the Harbour Bridge" (a real scenario we are working on) and get a video within seconds. But beyond that, disabled travellers in a powered wheelchair will only see video content made by people in a similar situation with real reviews when they research travel destinations (see Access All Areas below). Those into needlecraft will only see content made by fellow needlers and so on.
So for the narrowcaster of 2050, think of a dynamic audience that grows and contracts depending on the video available. Every second of video you shoot as you go about your life could be available to be re-used to automatically create television shows for people like you. That is the future we're building at Vloggi.