Writing an afterword is a privileged position, somewhat akin to hindsight, from which one can see the rich diversity of essays in this special issue as a whole. Together, they form a larger picture beyond what any one by itself could achieve. The vision that emerges shows the myriad entanglements that weave together humans being shaped by computational media, even as algorithms are molded and shaped through human embodied practices. Some emphasize putting humans back into the mix. Harry Halpin, for example, urges us to think about collective intelligence rather than artificial intelligence, while Lewis et al. point toward indigenous knowledges as demonstrating “abundant intelligences” that extend throughout the natural world. The lesson here is well worth thinking about as a general template for how communities can assume control over the kinds of AI created, the purposes they are designed to fulfill, and the human interests that they can support and help to enact.

It may seem self-evident that humans shape algorithms, but the devil is in the details, as the saying goes. It should be no surprise that humans, with embodied knowledges and embeddedness in the world, possess far richer contexts than are available to algorithms. Carboni et al. demonstrate how algorithmic decisions are moderated and influenced by the work of nurses in a clinical setting where violence is a possibility. They point to the importance of the “doubt” the nurses introduce as an important contribution to collective decision-making. Van Voorst emphasizes the “collaborative tinkering” between humans and algorithms as a form of collective decision-making in the context of public health, a conclusion largely in accord with Carboni et al.’s analysis. Cabitza et al., coining the neologism “cybork” (cyborg plus work), emphasize that algorithms frequently depend on human labor. For example, Bennett et al. explore the social and power dynamics when humans clean data so that algorithms can operate more efficiently (and more correctly). They, along with Cabitza et al., argue that this work should not be occluded, as if the algorithms were completely autonomous; rather, algorithmic processing should be seen as a form of collective networking and human–machine collaboration.

The recursive feedback loops between AI and humans imply that algorithms influence human behavior simultaneously with humans shaping algorithms. Lynch et al. demonstrate this on a microscale when they discuss how the introduction of robot tour guides into two Nevada museums helps to create the distinctive social spaces that visitors encounter. Kalthoff and Link discuss the fabrication of humanoid robots designed to create social dynamics through their interactions with humans, such as the Kismet robot designed by Cynthia Breazeal. Richard Groß makes a broader claim, arguing that Large Language Models, using stochastic procedures, are shaping social reality in their image (a claim that Louise Amoore and her colleagues have made by focusing on the principal characteristics of LLM technologies such as latency and tokenization, creating a “world model” that is rapidly propagating through the infrastructures of developed societies).Footnote 1

In addition to essays demonstrating the recursive loops between AI and humans, there are others that argue for changing the frameworks, metaphors, and narratives that construct AI–human relations. Halpin and Lewis et al., already mentioned, delineate new visions that depart from received views by fundamentally altering who and what is in control of AI development. Bennett et al. echo these concerns in calling for Responsible Artificial Intelligence (RAI), where alignment with human values is primary. Bory et al. look critically at AI narratives, arguing that those with more modest visions (“weak” rather than “strong”) have more potential to lead to futures beneficial to humans. Nugent critiques the pervasive analogy between biological and technological evolution, arguing that it works by suppressing crucial differences that show the complexities of biological processes.

Finally, I want to single out Fabian Offert et al.’s essay on protein folding as predicted by AlphaFold and other Transformer technologies. This essay is of special interest to literary and cultural critics, because it deconstructs the widespread belief that LLMs produce “natural” language, a criticism that I have also made. As Offert et al. show, through tokenization, word embedding, and positional embedding, the linguistic qualities of language are deconstructed, reconstructed as numbers in a high-dimensional vector space, and then reconstituted through stochastic procedures that differ radically from how natural language is used and understood by humans. These procedures, they suggest, create a new episteme that creates and reinforces an “intuitive” knowledge of sequence, hence its applicability to protein folding. Analyses like this open new possibilities for literary criticism, cultural critique, and critical code studies.

In conclusion, readers will find a plethora of diverse approaches, specific demonstrations, and game-changing ideas in these collected essays. They point to the ongoing necessity for critique and analysis by humanists, cultural critics, and others as we are confronted with powerful technologies developing at breakneck speeds. We cannot afford to ignore AI, and these essays help to chart the paths that can help us understand them, participate in their development, and have a say in how they are and will be used.