'The responsibility is with Google and YouTube, to regain control of their machines.' Photograph: Alamy
There are upsides and downsides to trusting machines to make important decisions. The story of the Mars rover, Curiosity, is a shining example of the former. Nasa's communications expert compared conversing with the 352-million-miles-away-from-home machine to awaiting for a call from an errant teenager. But there is a dark side too, and it largely springs from the assumption that our digital automata have automatically inherited our sense of decency.
Companies pay for online programmes such as Google Adwords where they design ads and set keywords for their advertising campaigns. Google then matches these campaigns to online content, automatically inserting ads into pages on YouTube, which Google owns, and other affiliated websites. Matching text with text is fairly straightforward, but making comparisons in digital media is more complicated, and on occasion ads are displayed because of YouTube's Content ID system.
Content ID depends on copyright holders uploading examples of their products to YouTube from which a comprehensive library of digital fingerprints is compiled. The system then compares uploaded videos against this library, flagging close matches as likely infringements of copyright. In theory, Content ID allows music labels and other copyright holders either to have copyright-infringing content removed, or reclaim some of their losses from illegal distribution by receiving a cut from the advertising revenue.
Additionally, the revenue sharing part of Google Adsense allows YouTube members to also benefit from advertising: Ray William Johnson is the top ranking user on YouTube, earning more than $1m of ad revenue in 12 months.
In practice, the whole system's performance is less than ideal and shows the symptoms of lacking human discretion. In the three years since Content ID was introduced, it has often misidentified content or mismatched keywords, but the most odious outcome from this blind process is the association of product advertising with unrelated and undesirable content.
In a previous article, I wrote about how corporate ads were appearing alongside hate content and neo-Nazi propaganda. Google's excuse is that their quality control is overwhelmed by the volume of uploads, with 72 hours of video uploaded every minute, resulting in nearly 12 years of content uploaded every day. Their solution is to rely upon users of the website to flag content considered inappropriate, as a form of self-regulatory crowd-sourced control.
But relying on users to report videos is not a sustainable way of moderating content and advertising on YouTube. It is out of control: the very same mechanism is also associating ads with other disturbing videos, often showing suicides, murders, and the aftermaths of violent deaths. The multimillion pound dating agency "Cupid", voted Scotland's fastest growing technology company in 2011, were unknowingly paying for their ads to be displayed alongside pornography, "snuff" movies and footage of suicides and accidental deaths. Inexplicably, ads for Zaggora Hotpants, a brand of bioceramic anti-cellulite slimming shorts, were being displayed alongside pro-BNP videos, soft porn and films of actual, often violent, deaths.
One YouTube channel called "Minus Thrash Show" presenting videos of gory accidents and real suicides was recently closed down, but immediately reappeared, sidestepping its expulsion with a subtle renaming and has re-uploaded all its video content. It now broadcasts as "Minus Thresh Show" and has advertising associated with several of its videos: the UK government's national citizen service, pitched as a "a life-changing experience for 16 and 17 year olds", has its ads displayed next to this channel's footage of the aftermath of a traffic accident where the top of a man's skull has been crushed upon impact with the road surface. A decapitated body lies nearby, its head is several metres away. The only explanation provided by the Content ID system for bringing together these apparently disparate elements is a positive match in the video with a song by a "blackened death metal" band from Gdansk.
Two questions arise from this situation: first, is it acceptable that this material is hosted by YouTube and should they and their revenue partners be profiting from its existence? Google announced revenues of $12.21bn (£7.8bn) for the quarter ending 30 June 2012, an increase of 35% compared to the second quarter of 2011, $1.2bn of which comes directly from the UK. That they only paid £6m in taxes against a £2.6bn UK revenue last year is a separate issue. Second, should companies be paying for advertising with Content ID matching that is currently juxtaposing their brands with execrable content?
I share the belief with Robert Levine, the former executive editor of Billboard magazine that, "some people will scream that YouTube is a public space and therefore you can't give them any private liability because it's a freedom of speech issue. But it's not a public space – it's more like a mall than a park. It's a multibillion dollar company and it has to have some kind of corporate responsibility".
While great scientific breakthroughs are hitting the headlines, with the Higgs boson and Curiosity, there are also impressive advances in artificial intelligence. Autonomous moral agents and the moral Turing test will one day produce a machine that can at least emulate decency, and it could be the solution to this problem. Until that point, the responsibility is with Google and YouTube, to regain control of their machines and their greed, and not prioritise money over morals.
• This article was amended on 16 August 2012. It originally referred to "Adsense's Content ID system" rather than YouTube's, and to Google Adsense matching campaigns to online content, rather than Google. It has also been updated with newer statistics on the amount of video uploaded to YouTube every minute (it has increased from 60 hours a minute to 72).
No hay comentarios:
Publicar un comentario