In our current generative AI paradigm, so-called hallucinations are typically seen as a kind of nuisance that will eventually be swept away as the technology improves. There are several reasons to question this assumption. One of them is that the very phenomenon is the result of deliberate business decisions by corporations invested in delivering diverse sentence structures through deep learning and generative pretrained transformers (GPTs). This article urges a fresh view on “hallucinations” by arguing that, rather than being errors in any conventional sense, “hallucinations” are evidence of a probabilistic system incapable of dealing with questions of knowledge. These systems are epistemologically indifferent. Yet, by presenting as errors to users of generative AI, “hallucinations” can function as practical reminders of and indexes to the limits of this kind of machine learning. Viewed this way, “hallucinations” remind us that every time you get something reasonable-seeming from a system such as OpenAI’s ChatGPT, you might as well have been given something quite outrageous; from the machine’s perspective it’s all the same.

Deadbots—AI systems designed to simulate the dead—clarify how generative AI reshapes temporal sense‑making. Operating in the “digital limit situation” of loss and finitude, they neither preserve memories nor store archives; they sever material traces and outsource the work of remembrance to automated interaction, ultimately fostering forgetting. The talk frames deadbots as a convergence of two traditions. From cybernetics, they inherit the “empty archive,” where feedback replaces retention and provenance is erased during model training. From technological spiritualism, they draw on practices that use technical mediation to confer authenticity, echoing nineteenth‑century séance boards and mid‑twentieth‑century Electronic Voice Phenomena. In both cases, technology gains authority through its apparent objectivity and opacity, inviting speculation about contact with the absent. Yet deadbots diverge from their spiritualist lineage by eliminating the interpretive labor once required to sustain such connections. The user’s role is reduced to passive consumption of a corporate service, while the system’s probabilistic token prediction turns remembrance into chance encounters. Consequently, deadbots function as engines of presentism—the endpoint of an “automation of memory” that dissolves the past into ever‑renewed simulations.

How were the problems and promises of technology addressed in the heightened public awareness of environmental issues during the 1970s? ARARAT provides insight here. It was an exhibition arranged at Moderna Museet in Stockholm in 1976. Its goal was to inform visitors about the interconnections between humans, society, and the environment and, on the basis of such knowledge, empower people to seek less wasteful ways of life.To this end, the exhibition invitedvisitors to become actively involved in experiments and get acquainted with everyday technologies that were less expensive to make, easier to repair, and more transparent in terms of production and life cycle. Previous research on ARARAT has focused on its relevance for current practices in art and architecture. The presentarticle, by contrast, aims to situate ARARAT in contexts that are both more general and more specific than previous work has been able to show.The study argues, first, that ARARAT can be understood as pioneering the field of practical knowledge (“praktisk kunskap”) before it was more formally established around 1980 in Sweden; second, that the ARARAT undertaking amounted to a new kind of popular education(“folkbildning”) with its combination of science, politics, and hands-on experimentation; and, third, that the ARARAT project was a demonstration of the exhibition format as an environing mediumwhereby it actively took part in changing the environment by empowering the population in an era with significant collective challenges.

With generative artificial intelligence being applied in everyday operations, e.g. as an alternative to web search, the question of models as sites of knowledge must be considered. As a source of information, an AI model is neither an ordered archive nor a database but a statistical engine. In this lecture, I discuss the relevance of error in the production of knowledge in cybernetic and AI systems and how it relates to specific uses of the past. While error-correction is crucial to the operation of language models, false outcomes e.g. hallucinations, can hardly be considered errors or mistakes in a epistemological sense because the concepts of truth and falsity are beyond the model architecture. This situation can be compared to an earlier cybernetic principle of negative feedback as a method to regulate and control a system. Such an approach has been suggested as being the opposite of a traditional archive effectively producing a very instrumental use of the past: input to steer toward a more desired outcome. Nevertheless, there have been approaches in social and human sciences which drew on cybernetic ideas for how to make use of accumulated knowledge. Comparing cybernetic principles with current AI regimes – especially their conceptions of errors – this lecture asks: what uses of history (broadly conceived) are made possible by these paradigms respectively?

For more than half a century, the use of artificial intelligence in legal domains has been a topic of interest for scholars and organizations. In this talk, I look at some major trends in these undertakings, asking: How did they conceive of the role of technology in the area of legal work? From the standpoint of how these projects sought to implement AI, how did they imagine the function of the law in society? Then, turning to our current situation and looking forward, I discuss some potential trade-offs which lie before us. If previous attempts to make legal matters computable ran up against the limits of propositional logic, the bluntness of formalization et cetera – what can we expect from deep learning? More specifically, is there a conflict between prediction as a goal in AI and explainability as a requirement in the practice of law? Or – might the efficiency of predictive statistics encourage a new self-image of the legal domain: from one concerned with precision, transparency, and accountability to one which traffics in data-driven forecasting, valorized by the old dream of prevention?

Our notion of what it means to communicate – where does it come from? This chapter visits a seminal work facing this question head on. In Speaking into the Air (1999), composed on the doorstep to the new millennium, historian and philosopher of media John Durham Peters suggests that our understanding of communication rests on an unfortunate view of mediated interpersonal exchange as something inherently broken. The presumed defective state of our interactions across distances, he argues, is based on the misguided idea that a harmonious and flawless union of souls is both possible and desired. To demonstrate his contention, Peters locates an original separation between the dialogues of Socrates – forcing participation, pushing towards agreement – and the dissemination of Jesus as told in the Gospels – spreading the seeds or words for those who have ears to hear. He then carries this ancient bifurcation on an idiosyncratic route between pillars of the Western canon of intellectual history, among them, Augustine, Aquinas, Bacon, Locke, Hegel, Marx, Kierkegaard, and Haraway. As Peters interrogates their positions on angels, money, love, law, ethics, labour et cetera, he takes these topics to be, ultimately, problems of communication whereby he builds on scholars like McLuhan and Kittler while also laying a foundation for continued extensions of concepts of media. Peter’s work productively shows the deep affinities between spiritual practices and technical endeavours in establishing contact with absent entities as in the cases of telegraphy and telepathy. It reminds its readers that techniques such as phonography, telephony, and photography were received with existential unease and wild speculation regarding the proper and probable locations of mind and matter. All carried fantasies of disembodied paths to the dead and the distant. Critiquing the idea of communication as an enduring shortcoming, Peters’ proposal is that we seek a less ambitious notion of what it means to connect. One that gives up on forced perfection and the desperate technological fixes employed to attain it and, instead, stands in awe of the fact that we are able to reach one another at all: humans and, perhaps, animals, computers, and aliens too. In making this case, Peters draws on the American pragmatist tradition and finds with William James and Ralph Waldo Emerson an attitude towards an exchange with others based on “making-do”; communication as a form of work that is ongoing, embodied, trusting, and always attentive to the profound otherness of fellow creatures and environments.

Artikeln argumenterar att Donald Trumps valseger 2024 utgör en strategisk triumf för Silicon Valleys “tech‑brorsor” – Elon Musk, Peter Thiel, Marc Andreessen m.fl. – vars långsiktiga mål är att ersätta politik med teknik. Författaren väver ihop observationer från Berkeley, historien om Ted Kaczynski och den moderna AI‑ekonomins extrema energihunger för att visa hur ett teknikcentrerat tänkesätt allt mer dominerar USA:s maktstruktur. Han beskriver hur riskkapitalet backar kandidater som J.D. Vance, hur AI‑jättar planerar egna kärnkraftverk (t.o.m. återöppningen av Three Mile Island) och hur Andreessens “tekno‑optimistiska manifest” speglar Kaczynskis idé om en livsavgörande fiende – nu riktad mot institutioner, reglering och demokrati. 2024 framställs som året då Silicon Valley “vann valet”, med löftet att teknik ensam ska lösa samhällsproblemen – men först efter att politiken avskaffats.

The presentation argues that the design and system development approaches articulated through the Dialogue Seminar (“Dialogseminariet”), including their philosophical underpinnings, are only more relevant in today’s AI landscape. The Dialogue Seminar was a collaboration between the Royal Dramatic Theater, Center for Working Life and KTH Royal Institute of Technology, all in Sweden, from circa 1985 onwards (which can be broadly characterized as human-centered, participatory, and non-rationalist. It makes the argument that the specific manner in which the labor, knowledge, and skills of the human was articulated in relation to technology in general and AI in particular in the Dialogue Seminar holds the power to productively inform our current challenges in deciding what agency is, where it should lie, and why it matters. In making this argument, the presentation specifically turns to the concept of error. Errors and mistakes, it is argued, were employed by the Seminar – in lectures, syllabi, books, articles, conference proceedings – to demonstrate the unique nature of human capabilities; we fail, as it were, intelligently, as opposed to the rigidness of machines that err predictably, but also incoherently which is a testament to the profound difference in quality between ourselves and our tools. The talk shows that such “error studies” are highly useful in scrutinizing the alleged faculties and capacities of generative AI. 

Systems of machine learning (ML) are increasingly being used as ”archives”, replacing the conventional search engine interface to depositories of collected knowledge. Yet, because these systems were never intended to be used for information retrieval, their widespread use in this capacity hints toward a new archival regime. To understand the historical circumstances which have shaped how current ML systems retain and serve back data to users, this chapter revisits the fundamental training methods of cybernetics and connectionism: negative feedback and backpropagation, respectively. By turning to these mechanisms, the investigation highlights the importance of error in the history of artificial intelligence, yet not the sorts of error that have typically been recognized and embraced by humanistic scholarship – deviations, wanderings, nonconformities – but, rather, types of error more often studied in psychology and engineering.

In doing so, it makes the argument that these error-correction techniques – applied in the effort to optimize a learning outcome in the system – effectively treats the past as a series of errors whose only value is to steer a machine or model toward a predefined goal. Critically, this line of inquiry enables the chapter to ask about the effects that standardized approaches to make predictions have on how the past is conceived. The suggestion is made that by implementing systems based on cultures of prediction to act as “archives”, we arrive at an order where information is organized according to the principle of statistical likelihood. Whatever falls outside of the most probable pattern, tends to disappear as error. This is a core property of machine learning. If digital computers introduced a problematic conflation between memory and storage (this was intensified with database technology), artificial intelligence forces us to consider this dynamic again. After all, large language models were designed with inspiration from neural networks, not records management. What is desirable for a computational learning task is quite different from what we expect in a database or an archive. The chapter argues that this ongoing shift in architecture for how data about the world are accessed shapes our informational landscape and, as a consequence, our ability to form an image of the past.