Was asked to give a few remarks at the 2020 Dahrendorf Colloquium organized by Timothy Garton Ash at St Antony’s College in Oxford today. Really interesting discussion with six of us kicking off discussions of six big questions relevant for the Dahrendorf Programme for the Study of Freedom, with Rana Mitter talking on China, David Priestland on Russia, Kate Sullivan de Estrada on India in terms of geopolitics, and then Paul Collier on the future of capitalism, Dieter Helm on Climate change and energy policy, and myself on the internet and AI.
I had 5 minutes, so necessarily short. Also, meant to tee up a discussion that goes far beyond my own core expertise. A cleaned-up transcript below with a few links added.
Empires by innovation? A few thoughts on the internet, AI, and Europe/the US
I’ve been asked to speak about the internet and AI with a focus on differences or convergences between US and European approaches
I’ll talk about the Internet first, AI second, then US and Europe
First the internet.
The internet increasingly is turning into only partially overlapping and interoperable “splinternets” as states assert control and regulate more and more.
Wendy Hall and Kieran O’Hara suggest we can think of as at least four different internets, Silicon Valley’s “libertarian internet” (they call it “open internet”), Brussels’ “bourgeois internet”, Beijing’s “authoritarian internet”, and Washington DC’s “commercial internet”.
But even as many states are asserting more control, leading to greater regulatory divergence, outside of those countries where they are in practice prevented from operating, we see greater practical convergence, as a growing part of the practical infrastructure of free expression, at least at the application and operating system-level, is operated by a limited number of US-based, for-profit private platform companies that people increasingly rely on for a large part of their news and information needs (and much more).
Consider just two datapoints from the UK
First, our media environments are already digital-first. According to eMarketer, in 2019, adults in the UK spend on average 23 percent of their time with media watching television, 15 percent listening to radio, and 55 percent online across smartphones, tablets, and personal computers.
Second, whereas domestic media dominate offline media (in early 2019, we found that the BBC accounted for 63% of all radio listening in the UK, and 31% of all linear scheduled television viewing), US-based platforms loom large online. In early 2019, the BBC accounted for just 1.5% of all time spent with digital media. By comparison, Google’s various products and services made up 22% of all time spent with digital media, and Facebook’s 14%.
This shift is in part the result of choice, people have access to printed newspapers, broadcast, and countless websites, and in practice choose to spend much of their time online, and primarily with a limited number of platforms, who over time build up dominant positions for their core services, even as they face competition more widely for attention and advertising.
So in terms of individual behaviour, we have practical convergence, as people seek out and use the same set of platforms in the US and Europe, even as we see institutional differences and regulatory divergences, as the implications of this move are, and are seen as, somewhat different.
All the major platforms are based in the US, most responsive to US stakeholders, whether private companies or political actors, suffused with (some) US values, and often design first for US market and regulatory context, then roll out elsewhere afterwards. Overseas, I think they are seen by many as fundamentally US entities.
In Europe, they are in practice embraced by hundreds of millions of people – an embrace sometimes accompanied by considerable reservations about their data collections practices etc., but an embrace nonetheless – even as many European activists and NGOs, many European private companies, and many European policymakers worry about the consequences of the rise of what we might call rival, competing, commercial “empires by innovation”, with a nod to Geir Lundestad’s notion that the expansion of US power in Western Europe during the early years of the Cold War was the expansion of “Empire by Invitation”. (The empire analogy is imperfect – the platforms do not rely on physical violence and are not exactly monarchical in their governance, so far from the authoritarian bloodbath of the British Empire – but perhaps generative in drawing attention to how they extend power relations across spaces where they have no prior or given sovereignty, legal or otherwise, and where, in one or more of the domains, they gain some measure of extensive hegemony over those spaces that help them extract or accrue value.)
What does this mean for AI?
I’m not a technical expert but I’m happy to point to a few institutional dynamics.
I think changes in how we use the internet has led to a situation where large US-based companies play an increasingly important role in structuring European public debate, news, and information, completely integrated into their technical and data-extraction structures, and where Europe in turn is these companies’ second-most important market currently, accounting for a large part of their revenues, in many cases something like a quarter.
This means, first, that when these large technology companies increasingly rely on machine learning and other forms of AI as part of their various ranking algorithms, these (for-profit) AI technologies increasingly shape public debate, news, and information in Europe – in 2019, for example, we found, looking across 38 markets, including more than 20 in Europe, that just 29% of internet news users say going direct to news sites or apps is their main way of accessing news online, compared with 53% who say that various forms of algorithmic selection, including search engines, social media, and news aggregators, as their main way of accessing news online.
It means, second, that, given their popularity and pervasiveness, the biggest US-based platforms not only have more money to invest in developing AI, and a greater ability to recruit and retain tech talent, they also often have greater ability to extract in volume, at velocity, and with variety, much of the data that powers the practical application of most AI technologies in Europe, than many of their European competitors. (Thought there are other kinds of data than the consumer data platforms extract online, including industrial and inter-organizational data, and that the quality of the data platforms extract on the two final Vs of the “5 Vs” of big data, veracity and value—yet to be seen.)
The big US-based platforms are thus the most visible incarnations of what Shoshana Zuboff has called “surveillance capitalism”, oriented not just to the accumulation of profit, but also of data, but they are far from alone, and both European companies and European politicians would actually rather like to grow European surveillance capitalists champions (always, everyone will hasten to add, as the US incumbents do too, in a “privacy-compliant fashion”). This is an interesting situation where we have a technology industry often associated with libertarianism behaving in a way some would say is fundamentally illiberal, especially when pursued by private enterprise and executive power working in concert, as we have seen for examples in collaborations between AT&T and the NSA.
It means third, because the big US-based platforms make so much money in Europe, that they would like to continue to operate here and are thus likely to acquiesce to many forms of regulation and other forms of political intervention.
Third, then, the US and Europe
What the, given the practical convergence, and the central role of a few US companies with global reach, are the differences between between the US and Europe (with China lurking in the background)?
One of them is regulation.
Crudely put, I think the difference can be summarized thus:
- US companies build many of the new technologies that countless people and a multitude of private companies and public authorities choose to use, both in the US and Europe (and China provides more and more of the infrastructure and hardware).
- The US politically has chosen to mostly regulate these technologies by not really regulating them very much, as suggested by Hall and O’Hara’s notions of the libertarian and commercial internets (with a bit of national security lurking in the background, some interesting ideas circulating in left-leaning legal circles, and much sound and fury from Donald Trump, but no new action yet)
- Europe, in contrast, increasingly try to actively regulate these technologies both at the EU level and the member state level, across data protection and privacy, eCommerce, competition, taxation, AI, and responses to misinformation.
- And, at least looking at misinformation, our survey research suggest greater public acceptance of government intervention in Europe than in the US.
- So the US is building technologies, but regulating through non-regulation, and Europe is not really building very much, but regulating. (China, meanwhile, is both building AND actively regulating, and keeping out technology companies from elsewhere unless they fully comply with often draconian rules and regulations.)
In one sense, this makes Europe very powerful – through what Anu Bradford calls the “Brussels effect”, where the EU exercise a unique power to influence global corporations and set the rules of the game by acting in the regulatory space (while the US does little or nothing), with many countries across the world then choosing, effectively between doing nothing (like the US), following Europe, or imitating China.
But in another sense, this power is not easy to exercise in a convoluted European-level political system, with democratic member states’ own domestic political dynamics, and complex national and regional economies, where “Europe” is not one thing, does not necessarily agree on what it values and interests “it” wants to promote, and where even individual actors, whether us as citizens and end users, private companies, or public authorities, often have multiple and not always easily aligned priorities.
As individuals, many of us may have reservations for example about how technology companies collect and use data, even as we use their products and services every day, often quite like to use them, and find them very useful.
As private companies, many European corporations may have reservations about how they are becoming more and more reliant on large US-based technology companies both in terms of reaching their customers and often in terms of their technical back-end, even as they also use their products and services every day, often quite like to use them, and find them very useful.
And among public authorities, European politicians might have reservations about the impact of, say social media and microtargeted advertising, even as they use them to campaign and communicate with the public, often quite like to use them, and find them very useful, and some European public authorities might have reservations about the use of, say, AI-powered facial recognition technologies, even as we know from independent reporting that law enforcement in many European countries have already been using these technologies, would quite like to use them, and would probably find them very useful.
So, in summary—
Even as increased assertiveness at both the state- and EU-level is driving regulatory divergence, there is considerable practical convergence as end users, private companies, and public authorities in Europe often embrace the same US-based tech companies as their American counterparts do –leading to the formation of competing commercial ‘empires by innovation’. (As with the US Cold War “empire by invitation”, this is complicated and in some ways problematic, but also not necessarily the worst actually existing option. Was my native Denmark worse of for joining the American international order? I don’t think so personally, even as I am keenly aware of the many downsides to this order.)
This in turn means that these US-based tech companies are increasingly important in Europe, and able to both apply their various technologies, more or less reliant on AI, and collect great amounts of data, but also that they make so much money that they are likely to acquiesce to forms of regulation that might lead them to withdraw from smaller, less lucrative markets.
This gives Europe power to act – especially in a situation where the US politically has chosen not to act – but how to act, and in the pursuit of what balance of what values and interests, that is the question.