Introducing a series on regimes of relevance
Over the past few days, I have published, without prior announcement, a series of five texts in this public notebook. I did so in silence by design: I preferred the sequence to be complete before presenting it, so as to offer the full trajectory rather than five disparate notifications. It is gathered here.
The series originated from a minor scene. A journalist granted an artificial intelligence agent access to their entire digital life and requested that it organise their birthday. The system performed a competent task and classified the most important person in their life as a "close friend". It possessed all the data; it had no conception of what mattered. That discrepancy—between what we deem significant and what a system finds relevant—appeared to me as a window into something more profound than a product failure.
The question that permeates the five texts is this: comprehending a work, a person, or a world does not depend solely on sharing a code or possessing more data, but on concurring on what type of difference is significant. I term this a regime of relevance. Misunderstanding—between human cultures and between humans and machines—may be described with precision as a misalignment of those regimes. The conclusion I reach is that the distance between the human mode of selecting the world and the machine mode is not one of degree, but of nature.
The series progresses from the conceptual to the technical; each text stands alone, yet together they articulate a narrative that none conveys in its entirety. This is the trajectory:
- What Matters and What Is Relevant. Initial thesis: aesthetic subjectivity as a method of selecting the world, and misunderstanding as a misalignment of relevances.
- Where the Machine Looks. Empirical section: when a human and an AI observe the same image, do they observe the same locations? The answer is measurable and is no.
- The self-effect. What causes a work not merely to please us, but to matter to us. Self-relevance as the core of intense aesthetic judgement, and the boundary that a system without a self cannot traverse.
- Processes Without Meaning. Generative art as a territory where human and machine share a commonality: the process, if not the meaning. From Conway to Tanimoto.
- The Box That Does Not Read Itself. Conclusion: why a machine's regime of relevance remains opaque even when one has access to all its internal mechanisms.
The sequence remains, for now, concluded in its initial trajectory. Its natural continuation is research; in time, perhaps a more formal work will articulate it. Yet its place of origin is this notebook, and here it shall remain open to any who wish to intervene—from the perspectives of aesthetics, cognitive science, AI studies, or artistic practice.
Thank you for reading.
Juan A. Esteban