The Evening Standard listed Demis Hassabis as 2014’s third most influential Londoner. After completing his PhD in neuroscience sometime in 2011, Hassabis, erstwhile computer game designer, world gaming champion, chess master and artificial intelligence expert, had started a company called DeepMind. In 2014, Google paid £400 million for DeepMind. DeepMind has no products, but it had managed to construct neural networks that learned to play early 1980s Atari computer games (Mnih et al. 2013). This game-playing performance motivated Google’s purchase. As is well known, Google has had a long standing commitment to data mining, machine learning and artificial intelligence in general in their bid to “organise the world’s information”. Reportedly, Hassabis’ DeepMind work at Google is already shaping search engine results (Levy 2015).

All of this might seem to lie a long way from roads, water and sanitation, energy or transport infrastructures of modernity. Or does it? Two things might motivate us to reassess that sense of distance between infrastructural modernity and the digital revolutions, one of whose cutting edges Hassabis personifies.

(1) Hassabis came after George Osborne (UK Chancellor of the Exchequer) and Boris Johnson (Mayor of London) in the list of influential Londoners. If Hassabis, the computer-scientific polymath, comes after the Johnson, the city mayor responsible for transport, waste, etc. in the global city of London, and after Osborne, the government minister responsible for national finance and economic management, then there might be an infrastructural imaginary that connects them. What could the cutting edge of digital infrastructure, that takes the form of predictive modelling, tell us about the materiality, the temporality, the forms of subjectivation, power relations or publics associated with infrastructure more generally? Might we say that after the infrastructures of State (UK), urbanisation (London), come the promise of mindful infrastructures? By ‘mindful’ I mean attempts to lend coherence to something that otherwise threatens to appear distracted, splintered, fragmented, or somehow aleatory. It might be worth thinking about, for instance, how Hassabis’ work on personality and the hippocampus might inform the operations of search engines (Hassabis et al. 2013). Something like a problematisation - ‘a kind of general historical and social situation-saturated with power relations … and imbued with the relational “play of truth and falsehood,” a diacritic marking a subclass of situations, as well as a nexus of responses to that situation’ - might be shaping there (Rabinow 2003, 19)

(2) Google’s purchase of Hassabis’ DeepMind is not an isolated case of the melding of artificial intelligence, pattern recognition or machine learning techniques with information infrastructures. We could turn to IBM’s Watson, an enterprise artificial intelligence that first shot to prominence by winning the US TV quiz show ‘Jeopardy’ in 2011. In the years since, Watson has been networked, re-tooled and infrastructuralised as a ‘cognitive computing’ assemblage whose Ecosystem has, with the help of $1 billion investment and a footloose global team of several thousand technical and marketing personnel, rapidly penetrated elite medical institutions such as the Mayo Clinic or the Sloan Kettering Memorial Cancer Institute, and has wedded itself to insurance, higher education, the health industry, various scientific fields (notably genomics) and cooking, amongst other things. The financialised leveraging of these approaches is also well evidenced in the case of HP Autonomy. Autonomy, one of the UK’s largest and most profitable software businesses, was bought by HP for £7.4 billion ($10 billion) in 2011 on the basis of its Bayesian pattern recognition techniques (also branded as ‘meaning-based computing’, themselves allied with researchers in computer science at Cambridge). Despite the fact that within months, HP ‘wrote down’ the value of its acquisition by around $8 billion, alleging accounting improprieties, HP has integrated the key components of ‘meaning-based computing’ into its many faceted utility, healthcare, transport, call-centre management, data management, operations management, patient records, legal support, security and intelligence products.

DeepMind, Watson and Autonomy have something in common. They are all highly ‘mindful’ in their promise to re-integrate the dispersed, forgotten or contradictory experiences of infrastructure.More literally, they abound in references to cognition, meaning, perception, sense data, hearing, speaking, seeing, remembering, deciding, and surprisingly, imagining and fantasy. They are all polymorphous figures of infrastructural reorganisation around the ideal of something like pattern recognition or indeed conscious awareness. There modelling practices are no longer the statistical rendering of number in the hands of government, science or commerce, as in William Gosset’s construction of Student’s t-test for monitoring Guinness stout quality in Dublin in the early 20th century (Porter 2008; Hacking 1990), nor the intricate interweaving of classification systems and things that grew into a forest of operational standards during the course of the last century(Bowker and Star 1999). DeepMind, Watson and Autonomy each seriously address the relation between humans and infrastructures, not so much as a matter of imagining (Mackenzie 2003), practice, configuration (Suchman 2006), or repair (???) but as a matter of competitive cognitive challenge. Watson, DeepMind and Autonomy present problems of seeing, hearing, checking and comparing as no longer the province of human operators, experts, professionals or workers seeking to navigate and finesse the constraints, limitations, breakdowns and vicissitudes of infrastructures, but as challenges set for an often almost Cyclopean cognition to reorganise and optimise in ongoing competitive experimentation. Here the initial training of DeepMind to play a half-dozen Atari computer games (from the early 1980s) or Watson to win Jeopardy directly point at this orientation towards challenges amidst competition, disconnectedness and disparity.

Would it be fair to say that these kinds of cognitive infrastructures, with their appetite for data and their ambition to reorganise other infrastructures (e.g. Google’s Self-Driving Car or its acquisition of the home thermostat company Nest; etc.), emanate from the drooling techno-cosmological imaginaries of Silicon Valley engineers and the like? Yes and no. On the one hand, the centres of calculation that Google or HP make and manage are effectively global assemblages (Ong and Collier 2005), with specific administrative, commercial, engineering and scientific apparatuses and regimes of value. Undoubtedly, they powerfully subduct (Mackenzie 2012) existing infrastructures. On the other hand, the increasing ‘mindfulness’ of the infrastructures under construction at IBM, Google and the like predicate a certain re-concatenation of the world, no longer in the mobile train of experience of people moving through streets, houses, factories and offices, but instead in the relations mindfully discerned in streams of data. But the mode of existence of contemporary infrastructures might be changing shape as a certain kind of ‘mindfulness’ is brought to bear on the scattered, partial and disaggregate worlds, themselves partly a product of the splintering of infrastructural modernity (???), as well as the device-specific intensities of ‘knowing capitalism’ (Thrift 2005).


Bowker, Geoffrey C, and Susan Leigh Star. 1999. Sorting Things Out. Classification and Its Consequences. Cambridge MA: MIT Press.

Hacking, Ian. 1990. The Taming of Chance. Cambridge University Press.

Hassabis, Demis, R. Nathan Spreng, Andrei A. Rusu, Clifford A. Robbins, Raymond A. Mar, and Daniel L. Schacter. 2013. “Imagine All the People: How the Brain Creates and Uses Personality Models to Predict Behavior.” Cerebral Cortex: bht042.

Levy, Steven. 2015. “The Deep Mind of Demis Hassabis — Backchannel. Medium.” January 16.

Mackenzie, Adrian. 2003. “These Things Called Systems. Collective Imaginings and Infrastructural Software.” Social Studies of Science 33 (3): 385–387.

———. 2012. “Sets.” In Devices and the Happening of the Social, edited by Celia Lury and Nina Wakeford, 219–231. Routledge.

Mnih, Volodymyr, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin Riedmiller. 2013. “Playing Atari with Deep Reinforcement Learning.” ArXiv:1312.5602 [Cs] (December 19).

Ong, Aihwa, and Stephen J Collier. 2005. Global Assemblages : Technology, Politics, and Ethics as Anthropological Problems. Malden, MA: Blackwell Publishing.

Porter, Theodore M. 2008. “LOCATING THE DOMAIN OF CALCULATION.” Journal of Cultural Economy 1 (1): 39–50. doi:10.1080/17530350801913627.

Rabinow, Paul. 2003. Anthropos Today. Reflections on Modern Equipment. Princeton; Oxford: Princeton University Press.

Suchman, Lucy. 2006. Human and Machine Reconfigurations: Plans and Situated Actions. 2nd ed. Cambridge University Press.

Thrift, N. J. 2005. Knowing Capitalism. London: SAGE Publications.