Break up the tech giants before it's too late
Artificial intelligence and big data mean Facebook, Google and Amazon cannot be beaten. We need to correct this market failure
Social media — or more specifically: Facebook — is under a microscope at the moment.
From Mark Zuckerberg’s US Senate hearing last year, to the UK DCMS select committee’s report on disinformation and fake news last month, via countless op-eds and white papers, the Menlo Park firm has been on a 24-hour damage control tour ever since its role in the earth-shattering election of Donald Trump was revealed.
And rightly so. Any one of Facebook’s myriad scandals would independently warrant close scrutiny and forfeiture of the benefit of doubt:
Despite this torrent of impropriety though, Facebook remains the sixth most valuable company in the world by market capitalisation, while its Instagram subsidiary is still the fastest-growing social network on the globe. Herein lies a deeper problem. Facebook’s vulnerability to manipulation, hate speech and disinformation is shared with other platforms like YouTube and Twitter, but the even more acute issue of accountability is common to all of big tech, whose global influence is now arguably larger than most nation-states: Facebook, Google, Apple and Amazon’s combined worth is greater than France’s GDP.
Instagram’s continued success illustrates part of the problem — the unparalleled ability to acquire competitors before they’re off the starting line. But M&A are not the only, or even the primary reason for the monopoly-like dearth of competition. The same features that raise concerns about the rapid propagation of fake news or the fragmentation of the public sphere are present across the tech sector, creating a new kind of dominance the market is incapable of correcting. Radical intervention is needed.
When the web began, it ushered in a new era of media distribution, operating at the speed of bits rather than atoms, but the direction of that distribution was mostly unchanged from print or television, with content traveling, one way, from the publisher to the user. We call this period web 1.0. This changed with web 2.0 allowing content to flow in both directions between publishers and users, enabling networked communities and marking the dawn of social media and other digital “platforms”.
For a while now, we’ve gradually been moving into the web 3.0 period. Here the flow of content is not only back and forth between producers and users, it is mediated through artificial intelligence in the form of recommendation algorithms. These form feedback loops with users’ interaction with content influencing the way that content is distributed.
Some of the web platforms came to dominance by deploying these features early. Google was founded in 1998, before web 2.0 had properly emerged, but its ‘PageRank’ algorithm implemented some of the same principles, utilising networks of links between web pages as a representative of quality. Facebook eclipsed all other web 2.0 social networks by introducing an early example of web 3.0: the algorithmically curated News Feed.
While the mechanisms and use cases differ, both Google and Facebook fundamentally operate through algorithms that serve users with content; algorithms that are being constantly trained based on what those users then do with the content. They are incrementally improved with every clicked search result or shared post. The users’ actual use of the service — their data — becomes incorporated into the service. I call this user-incorporated media, and the same approach is being applied beyond digital content, to hailing a taxi through Uber or booking a room with Airbnb.
Data cyclical effects and superlative efficacy
Much is made of “network effects” and the way digital platforms benefit from them to reach dominant scale. It is undoubtedly true that, much like the telephone networks before them, digital platforms become more valuable the more people use them. That also means they become more indispensable the more that usage becomes expected or assumed, to the point that it can actually be an inconvenience to other people if you don’t use them. However, user-incorporated media exploit an altogether more novel kind of network effect.
Because users’ data is incorporated into the service, additional users lead to a better service, not just because there are now more people in the network (like the telephone), but because the service is actively learning from its use. Machine learning is fuelled by data, and since data is generated by usage, the more users a platform has and the more they use it, the more data it generates, the more those algorithms learn and the better the platform becomes, meaning more users, more usage, and so on.
These “data cyclical effects”, enabled by artificial intelligence, create virtuous spirals, accelerating as they generate and process increasing amounts of data, allowing the algorithms to improve non-linearly. Jonas Andersson Schwarz describes this as an “entirely novel form of synergy” that creates “superlative efficacy”.
Optimising to increase usage is actively encouraged by the business model of many digital platforms like Facebook and Google who rely on ad revenue, effectively selling their users’ attention. This was dubbed “surveillance capitalism” by Harvard Business School’s Shoshana Zuboff and Andersson Schwarz believes it represents a new industrial organising principle, “like Fordism and Taylorism before it”, that’s being embraced widely. This is because data has value beyond merely targeting consumers; data is knowledge that, with the right artificial intelligence, can be implemented at previously unseen scales, stretching beyond the digital realm.
Apple can improve phone batteries by tailoring them to the detailed usage patterns each iPhone records. Netflix can commission a new season of a forgotten show based on the volume of viewers in its back catalogue. Alphabet are best-placed to launch a self-driving car because of the extraordinary amount of traffic data accrued from each car navigating with Google Maps. Many startups are hosted on AWS where Amazon can see the amount of traffic they’re attracting, gleaning which are worth acquiring, and at what cost.
User-incorporated media, their associated data cyclical effects, and the superlative efficacy they enable, create a profound first-mover advantage. Competing with the leading platforms would take enormous datasets to train comparable algorithms. Since they are generating more usage data — and therefore improving — all the time with no sign of diminishing returns, it may well be impossible to disrupt their dominance.
Break them up
So what’s to be done? The DCMS report states we need to “use regulation to restore democratic accountability” to platforms like Facebook. Indeed, they join a growing chorus calling for the establishment of an independent regulator to oversee this new class of company that is neither a ‘publisher’ nor a neutral ‘platform’.
But for regulation to work, it needs a sufficiently large stick to beat with. Even the record $2.7 billion fine the EU imposed on Google for anti-competitive practices was a drop in parent company Alphabet’s bucket, amounting to just 3% of its annual revenue. And even if regulation were successful in forcing the big tech companies to take more responsibility for its users’ data and policing its platforms, it’s not clear how much this would do to tackle the monumental accrual of power and market dominance they’ve achieved, not seen since the Gilded Age.
There is another option. Lina Khan’s influential Yale Law Journal article “Amazon’s Antitrust Paradox” has helped spark a rethinking of antitrust, returning to the Progressive Era principles that led to the break up of Standard Oil and the railroad monopolies. She argues that the Borkian consumer welfare standard, with its reliance on pricing, is unfit for the platform economy. Amazon is beloved by consumers for its low prices, while Google and Facebook don’t charge users at all.
Low prices are not the same as low costs though, some of which are obscured, e.g. the value of personal information, while others are entirely hidden, namely the opportunity and innovation that is smothered in the crib by what the Open Markets Institute calls the “platform monopolies”. Tim Wu writes in his new book, The Curse of Bigness, that the market concentration of big tech is having a chilling effect. Startups are near a 40-year low in the US, and the business model for many new ventures is simply to be acquired by a Silicon Valley giant. This does not bode well as history shows how large corporations tend to obstruct important technological shifts, as AT&T did with the internet.
A recent letter to the Federal Trade Commission, signed by nine advocacy groups, hints at a solution: it suggests breaking WhatsApp and Instagram away from Facebook. NYU Stern’s Scott Galloway proposes a similar approach for the other tech giants, moving from an ecosystem of four stakeholders to ten.
This alone is not a panacea, nor a replacement for rigorous regulation and enforcement of user protections and the integrity of the information space. But part of the reason these issues have arisen is the diminution of market competition, and due to these platforms’ status as de facto modern day infrastructure, the only way to inject competitiveness is through trust busting. The alternative is to allow the forces of data cyclical effects and superlative efficacy to make big tech so indispensable that the only option may be to nationalise them in the public interest. ⬢