The aim of AI is not to imitate humans. The aim is to offer humans the tools, technologies, and techniques to better realize human goals that support and enhance human well-being in a sustainable environment for all. AI is thus a family of techniques where algorithms uncover associations of predictive usefulness from data. Some of these techniques include deep learning which relies on multiple layers of representation of data that produces complex relationships between inputs and outputs.

Machine learning algorithms improve predictions by learning from more and more data. However, the kind of AI that can be described as “imaginaries” has not been achieved as yet. As we probe Ethical AI, Explainable AI, or Responsible AI, new shadows and silhouettes of the assemblage are disclosed on the horizon. Here, AI is seen as being part of a socio-technical system that must bear responsibility and ensure trust.

In the futures we are building, AI assemblages and humans are seen as constitutive of the social world through their dynamic interplay. While perspectives and definitions of AI vary widely at this time, one thing is clear and that is the very nature of any assemblage is that it never has any definitive beginning or ending. The void is not the lost part – it is the essential part because the incompleteness – the continuing character – is what makes an assemblage in the first instance.

Unobtrusive tools like voice-controlled smart speakers are models of how AI-driven technologies vine human users into spreading extensions and intertwinements with the broader AI assemblage. After humdrum commands like “set the volume to level 5”, followed by a response from the smart speaker, the volume is adjusted. Such commands are illustrative of the ordinary blunt and fleeting mode of AI-human interaction.

This bland and economical design of the AI assemblage renders the AI-human interaction as a “blankness”. An emptiness that conceals on the surface, a disembodied voice that represents the AI-human interaction inside a dumbfounding complex set of information processing layers across the matrix.

Beneath this transitory interaction is a vast matrix of capacities, interlaced chains of resource extraction, human work, and algorithmic processes across networks of data lakes, logistics, distribution, optimization, and prediction. But because gratification is instantaneous, the human user is not troubled by the work of the matrix.

In any given household, different people will have preferences for different Apps and generative AI platforms. Some may use AI-go which provides a wide range of pre-trained models for industrial artificial intelligence applications, or AI Watermarking. What this means is that each person in the home is intertwined into far-flung networks of different sites of social practice.

The home attracts an AI assemblage of technological components. Each house in a borough selects and attracts a self-selected collection of AI technologies and tools that have a social, political, and ideological undercarriage that reformats the home and society. Humans are interpellated into AI models in the shape of an assemblage of different preconceptions, subjectivities, and biases.

AI remains hidden in tiny black box devices scattered across the rooms of homes. But its reach stretches much further than the house’s walls, leveraging the formation of discordant elements that, taken together, materialize as the greater multifaceted socio-political assemblage of what we accept in our homes to be “AI”. In all of our micro-interactions with AI systems, we overlook that its real power and complexity lie somewhere else.

The AI natives in each home evolve multiple identities by being simultaneously a consumer, a resource, a worker, and a product. What differentiates AI from other forms of consumer technologies is its ability to draw upon the history of human learning for optimizing and training the AI assemblage.

Smart devices link the world together in a neural lace that is a multidimensional assemblage of many entities that constitute, and are constituted by, one another. AI requires a vast planetary network, fuelled by the continuation of extractive operations of industrial capitalism that involves materials like lithium in Bolivia, human work, data gathered by sensors sprinkled like confetti everywhere, and data refineries that produce distillates by conjoining unrelated data sets to create embryonic patterns.  All of this has a price tag.

Calculating the magnitude of the resources used to produce instant gratification is indeed concerning. AI assemblages and smart devices in particular presently lace the world together with invisible threads of e-commerce, technology, platforms, power, and politics.  It is an enterprise that relies on the ingestion, analysis, and optimization of vast caches of images produced by humans, unstructured data, and videos.

AI is a social actor. It is obligatory to construe AI in terms of social agency. It does things with, and to, the life worlds we scaffold around ourselves. It is not passive. A symbiosis between technological components and human beings is developing. This new interdependence entails understanding how digital tools and technologies mediate our experiences, opinions, and behaviours and equally, how human reasoned agency and capabilities unsettle the use of technological artefacts. To unmask what AI does – its agency – that only becomes visible through its socio-cultural assemblage, we must demystify it, and remove it from the realm of the inscrutable and the elusive.