Den 15 oktober 2026 ägde fakultetens konferens ”AI and the Humanities” framgångsrikt rum. Jag publicerar här ett utdrag ur min inledning:
We hear about it every single day. Since the fateful release of ChatGPT 3.5 in 2022, AI has been the topic of a relentless barrage of articles, news clips, memes and social media commentary. It never lets up, and it leaves me fundamentally torn.
One the one hand, I think of this as possibly the most successful public relations campaign in history, orchestrated by a few major tech corporations intent on securing returns on their astronomical investments. Like shrewd drug dealers, they flood the market at bargain prices first, only to raise them later when the addiction has set in. And the addiction is setting in. Why else do we keep on talking about AI when the planet is burning and inequalities are deepening? Why else should we be pandering to big tech by keeping the hype cycle spinning?
On the other hand, the pace of technological change is astounding and the technology itself is fascinating in its own right. Although the concept of neural networks is as old as computing technology itself, what has happened these last 15 years is, first of all, the exponential upscaling of computing power and speed. Even when producing the simplest statements in response to a prompt, LLMs perform probabilistic calculations with tagged tokens (parts of words) not by the millions but by the billions and trillions. With no obvious end to upscaling in sight, and with the continuous refinement of algorithmic computation, we are now surrounded in our digital lives by a technology producing human-like linguistic statements at an unimaginable, non-human scale. At the same time, this behemoth remains dependent on an all-too familiar set of preconditions: massive extraction of raw materials, biased training data and a vast supply of mostly underpaid human labour. It is also here, in these material conditions, that the technology confronts its obvious limitations.
AI, we need to remember, is a cluster of technologies with a long history of its own. Cold War-era machíne translation was not just simpler but crucially different from consumer-oriented generative AI today. Machine learning used for restricted purposes is again distinct from the apparent – I would say spurious – generality of public gen AI tools. From gen AI chatbots there is yet another leap to agentic AI. For the time being, it is gen AI that keeps us talking, enthralled and appalled alike. Its impact – especially within higher education and research – is undeniable and its use across our sector has become diversified and even trivialised. This is a situation that requires hard thinking. Understandably, much of the response among universities has been reactive or defensive. Sit-down exams are the new black. But there are any number of fundamental questions that require more creative responses. What do learning and understanding actually mean in the age of generative AI? How will, and how should, AI shape our research methodologies? What happens to core academic skills such as writing, language competence and domain skills, when translations, written texts and literature reviews can be produced in a matter of minutes? Will academic publication itself become obsolete, the better AI services get at churning out research articles?
What has struck me these last few years is how many of the questions provoked by AI converge upon knowledge domains proper to the humanities: language, ethics, history, the culture of reading, critical theory – and so on. When I hear computer scientists speak of the “end of programming” since a chatbot only needs to be prompted with “natural language”, my observation is that natural language is one of the key tools and study objects of the humanities. A recent book by Leif Weatherby, Language Machines, demonstrates this fundamental relevance of humanities. I quote here from his opening statements:
Language has entered the age of its algorithmic reproducibility. The human ability to speak can no longer define our singularity. […] AI has accidentally divorced language from cognition, raising more questions than it answers. The isolation of language as a sign-system from human cognition is an unprecedented experiment about how meaning is made. Language algorithms are a kind of hypothesis about how two systems of representation—computation and language—converge. That experiment, along with its hypothesis, calls for a theory to make sense of it, which must be a theory of meaning in the absence of intelligence. (Weatherby 2025, 1)
Whether or not Weatherby’s theoretical argument holds water is another matter. What is crystal clear from his book, however, is that generations of scholarship on language and literature can be brought to bear on contemporary developments.
Choosing a different tack, when yet another AI scare is announced, my reflection is how primed we are, culturally and iconographically, by a long tradition of science fiction and fantastical literature (Terminator, 2001, Frankenstein), to imagine AI as a thinking being that most likely is evil. Such a cultural and historical demystification of our own responses is precisely what the humanities enable.
When dealing with these issues, however, we need to steer clear of rank amateurism in relation to the technology itself. In a recent review, the American literary scholar Brian Lennon makes some cutting remarks about his own colleagues:
Today, technical ignorance of computing has two principal consequences in general, and specifically among ostensible professionals and experts in nontechnical domains of knowledge. First, it facilitates the projection of the shallowest and least gainful of futurisms […] enticing nontechnical scholars of any serious subject at all to make absolute fools of themselves. (They oblige in large numbers.) Second, it permits Big Tech to ideologically exploit such callow futurism, indeed to laugh behind their hands while shielding themselves and their own offspring from its consequences […] To be blunt about it, the silliness must stop. (Lennon, Comparative Literature Studies 63.3, 2026, 522)
One objective of “AI and the Humanities” is to make ourselves better prepared and to counter-act such silliness. Another is to facilitate unanticipated points of contact across disciplinary domains. Today’s conference is an initiative by the Faculty of Humanities at Stockholm University in recognition of the complexity of the questions we need to ask about AI from different disciplinary standpoints, and also in relation to research funding and research strategies. The conference does not have just one specific expected outcome, but is an opportunity for all participants to be inspired and reflect on the relevance of these issues from their distinct institutional location. An intellectually appealing aspect of our current situation is the lack of consensus in relation to AI. We have real questions to tackle – what could be better? In this context, I should also mention that a Stockholm University centre for AI in science, society and the arts will be launched early next year. Today’s conference feeds into the preparatory work for that centre.