Dr Santiago Rentería

Selected Transmissions from the Abyss

Tue, 17 Mar 2026 00:00:00 +0000

My Twitter account is dead. What remains is an archive, half-record, half-ruin. Two selections are presented accompanied by a lexicon. The first gathers the most creative transmissions: aphorisms, thought experiments, accidental poems. The second collects the most controversial: the statements that burned bridges, alienated colleagues, or simply said the unsayable without the courtesy of softening it.

Read them as you would read signals intercepted from a dead star.

I. Static: Ten Creative Transmissions

In the combed compartments of a city house, covert chambers held peculiar scenes. Each dim lighted room in fragments repeated an event I had lived before, endlessly, yet subtly altered with utmost perfection. Like the verses of a poem written in Gray code.

The most literary tweet in the archive. Dreamlike imagery resolved into a computer science punchline — genuinely haunting.

What is ambient but music aspiring to the condition of perfume.

An exquisite aphorism. Riffs on Walter Pater’s famous line about music and reframes the entire genre of ambient sound in one breath.

Storytelling was the first (analogue) virtual reality experience and literature the first immersive technology.

Deceptively simple, genuinely reorienting — a reframe that holds up philosophically.

Bureaucracies are the dependency hell of social systems.

Perfectly targeted for both programmers and social critics. The metaphor is exact and surprisingly rich.

Do Laplacian machines dream of Gödelian sheep?

A masterclass in compressed reference — Philip K. Dick, Laplace’s demon, and Gödel’s incompleteness theorems folded into a single question.

The problem of stabilizing knowledge in oral cultures led to the invention of myths. Impermanent but robustly heritable and polymorphic storage devices.

Reframes mythology as information science. Dry, technical vocabulary applied to ancient human practice — startling and accurate.

The territory is the map of all maps that are not about themselves.

A brilliant self-referential twist on Borges and Russell’s paradox simultaneously — philosophy and cartography colliding in one devastating line.

Twitter is just a bunch of polyphonic monologues that resemble dialogues from afar.

Concise, accurate, and cuts to the heart of the platform’s social illusion. “Polyphonic monologues” is a phrase worth keeping.

Self-driving cars are asymptotically non-deterministic generalised trains.

Absurdist on the surface, technically defensible underneath. Four words of jargon doing devastating satirical work.

10.

Lem is Borges for the space age.

Five words. Instantly illuminating to anyone who has read either author — and slightly mind-expanding to anyone who hasn’t.

II. Noise: Ten Controversial Transmissions

You know what is the best thing about liberal democracies, that you have the freedom to choose your favourite flavour of BS and it comes in different shapes of political correctness so it does not offend your entitled and shameblasting ass.

Contemptuous of both left and right simultaneously — the kind of statement that would get you called a fascist by liberals and a communist by conservatives, which may be precisely the point.

Universities in Aus die because most students come just to get PR, there’s no intellectual culture and protests scare away new ideas. Everybody repeats like a parrot woke statements just to circumvent redundancy.

Takes aim at the performative politics of academic culture — something most academics think privately but few say publicly. Guaranteed to alienate colleagues.

The perfect system for wealthy and politically influential nations is fascism with democratic rhetorics. Gaslighting at scale.

Blunt and deliberately inflammatory. Not entirely without analytical basis, but the framing leaves no room for nuance.

Identity politics has radiated more selves than little boy.

Comparing the proliferation of identity categories to the Hiroshima bomb is the kind of analogy designed to detonate. Whether it’s diagnosis or provocation depends entirely on where you stand.

Birds are not singing for you and artists are not co-creating with them. Birds want you to fuck off lest you want to be divebombed and disemboweled.

A direct assault on the entire genre of human-animal collaboration art — and a fairly heretical position for a researcher in acoustic ecology to hold publicly.

Moralised AI art that supposedly undermines the exploitative regime of dataism is the worst kind of digital art, I have to say.

Takes a very specific and unfashionable position in a field where ethical posturing is the dominant currency. Says the quiet part loud.

English was a mistake. It should be erased and replaced with Latin, Arabic and Greek.

Delivered without apparent irony. Maximally provocative to most of the internet, which communicates primarily in English.

I’ve become unemployable. You’ve gotta have no soul or be a social media soldier to bear this digital wasteland plagued with overproduced humbleteabaggers and psychopathic edgelords rizzing each other with corporilingus.

“Corporilingus” alone is enough to get someone defollowed by half their professional network. The whole sentence is essentially a deliberate career self-immolation.

Isn’t it lame that peak AI is… just chatbots and content generators on computational steroids? Where’s your AGI god? It’s not even a newborn, and according to Minsky we were supposed to grow artificial minds, not glorified databases with fancy conditioning.

Posted into an environment saturated with AI optimism. Calling the whole field lame while citing Minsky against its own practitioners is a very efficient way to make enemies in tech.

10.

Academia, or how to sustain your own intellectual addiction by starting a cult.

Concise, dismissive, and coming from someone inside the institution. The most memorable heresies are usually written by believers.

III. Lexicon: An Anti-Dictionary

Key terms from the archive, defined against themselves.

Bullshit (n., excl. — affect)

Not falsehood — something subtler and more devastating. The ambient condition of language when meaning has evacuated its forms. Not a lie (which still respects truth enough to hide from it), but the total indifference to whether anything is true at all. Recurs too often to be mere expletive; functions as a technical term.

“It is all bullshit. Fucking braindead earth.” / “IT IS A…, GUESS WHAT? ALL BULLSHIT.”

Care compression (n. — society)

The shrinkage of genuine interpersonal attention that occurs when informationally saturated channels are paired with low affective bandwidth infrastructure. The result: perfunctory interactions mistaken for connection. Empathy becomes throughput.

“The co-occurrence of informationally saturated channels with low affective bandwidth infrastructures produces interpersonal care compression and perfunctory interactions.”

Common sense (n. — epistemology)

Socialised reflex. A reification of appropriateness in a world that philosophically never made sense. Its apparent solidity is borrowed from the consensus of people too busy to notice the ground has already shifted.

“Common sense is socialised reflex.”

Culture (n. — society)

A deceptive architecture. Undead, exosomatised ideology. An armour one must wear to withstand the shards of history — panegyric and intergenerational chain mail of fossilised souls. Also: a regulatory network optimizing for self-replicating symbolic gameplays.

“Culture is a deceptive architecture. Undead, exosomatised ideology.”

Depersonalisation (n. — self)

Not pathology — baseline. The condition of pressing too hard against social performance requirements until the self stops feeling real. Fully depersonalised: a technical reading of an ordinary Tuesday.

“So much pressure to seem ok makes me not want to socialise. Then I suddenly snap and people hate me again. Fully depersonalised atm.”

Extinction listening (n. — sound)

The experience of listening to a sound known to be disappearing, or perhaps already gone. A phenomenological condition unique to the Anthropocene: grief organised around the auditory rather than the visual. Sound archives reduce the impact of virtual absence, but the non-virtual absence remains.

“Have you experienced extinction listening? The experience of listening to a sound that one knows is disappearing or perhaps already gone.”

Gratefulness (n. — affect)

Always fake. Not because people are dishonest, but because genuine gratefulness would require a stable self to receive what is given — and such a self is unavailable.

“All I see is the same things repeating over and over. […] Gratefulness is always fake.”

Happiness (n. — affect)

Preliminary and modest metaphysics of ultimate purpose. Possibly identical to advanced self-gaslighting. Empirically indistinguishable. A question that cannot be answered without first answering whether a stable subject exists to be happy.

“Please be honest. Is happiness real or you just get better at self-gaslighting?”

Influencer (n. — society)

Hypertrophied court jester. Organic clowns were better — equivalent in quality, at least. The AGI crowd’s carcinogenic positivism, when leaked into the general population via this vector, makes car exhaust more appealing by comparison.

“The carcinogenic positivism reeked by such hypertrophied influencers makes car exhaust more appealing.”

Knowledge (n. — epistemology)

The one thing that has persisted against trauma. What remains when everything else — social belonging, institutional legitimacy, emotional stability — has been stripped away. Not comfort: something colder and more reliable. What the ruthlessly seeking find at the bottom.

“Isolation made me ponder what has persisted against trauma: ruthlessly seeking knowledge.”

Language (n. — epistemology)

A coordination strategy. There may not be stable languages at all — only communication equivalence classes. Every reproduction amplifies the possibilities of misunderstanding, yet opens synthetic modalities of understanding beyond its own analytic interface.

“I sometimes think that there might not be (stable) languages at all, but only linguistic coordination strategies and communication equivalence classes.”

Machine listening (n. — sound)

The automated analysis of environmental sound by computational systems. Not merely a technical process — a mode of attention that reveals what it ‘hears’ through the absences it streams into the digital void. Carries the power of surveillance and the pathos of salvage ethnography simultaneously.

“Silence can reveal what the machine ‘listens’ by streaming algorithmically processed absences into the digital void.”

Myth (n. — epistemology)

The solution to the problem of stabilizing knowledge in oral cultures. Impermanent but robustly heritable and polymorphic storage devices. Not superstition — an engineering solution to the problem of keeping knowledge alive without writing.

“The problem of stabilizing knowledge in oral cultures led to the invention of myths. Impermanent but robustly heritable and polymorphic storage devices.”

Normal (adj. — society)

A heuristic of context. Not a state that exists, but a prototype around which behaviour clusters under normative pressure. The iterated function ‘Post’ evaluated over any given coordinate of normality diverges. This world was never normal.

“This world was never normal. Everyday our naively acquired sanity is undermined irreversibly.”

PhD (n. — society)

A sparse-reward, time-constrained, multi-objective optimization problem with unstable solutions. In practice: self-destructive exposure to light outside the canon. An institutionalised pathway to intellectual radioactivity. Medieval system built upon power-tripping dynamics and prestige.

“Doing a PhD can be formulated as a sparse-reward time-constrained multi-objective optimization problem with unstable solutions.”

Reputation (n. — self)

Nothing more than the perfunctory betrayal of self-knowledge. Betwixt burning bridges and immolating fixed identities, choose the latter.

“Reputation is nothing more than the perfunctory betrayal of self-knowledge.”

Self (n. — self)

An individual friction of universal fiction. A multimedia novel the brain updates in response to how the organism is coupled to its environment. A dialetheic machine. Selves are metaphysical compositions of glocal transactions — evolving fossil records of participatory sense-making.

“The self is an individual friction of universal fiction.” / “The self is a multimedia novel the brain updates in response to how the organism is coupled to its environment.”

Silence (n. — sound)

In the Anthropocene: not absence of humanity, nor a return to the wild. A mode of active listening designed to draw the more-than-human background into the foreground of thought. Also: what the machine streams when it processes what it cannot hear.

“Silence in the Anthropocene does not denote absence of humanity nor a return to the wild or some prebiotic state. It is a mode of active listening.”

Social media (n. — technology)

Who needs schizophrenia. The bicameral mind fragmented into polyphonic monologues that resemble dialogues from afar. An infosphere in which the excess of interiority becomes cyberspace debris.

“Who needs schizophrenia when you have social media. The bicameral mind fragmented.”

Technology (n. — technology)

Both memetic and mimetic. Not neutral — but claiming it is essentially good or bad dismisses its mutability. When mass-produced it standardizes behaviour; what it picks out is never clear. Not technophilia: the thing that will save us, if anything will.

“Technology is both memetic and mimetic.” / “Tech (not technophilia) will save us.”

Turing test (n. — technology)

The epitome of human cultural narcissism. Not a test of machine intelligence — a test for understanding ourselves through the machine. Its passage would tell us more about the poverty of our self-conception than the richness of artificial mind.

“The Turing test is the epitome of human cultural narcissism.” / “The Turing test is a test for understanding ourselves through the machine.”

Virtuality (n. — technology)

Not the counterpart of reality — its potential. Storytelling was the first analogue virtual reality experience; literature the first immersive technology. The virtual is not opposed to the real; it is what the real has not yet become.

“Virtuality is not the counterpart of reality but its potential.”

Words (n. pl. — epistemology)

Don’t mean anything. It is as if everybody had turned into a chatty meat puppet. Yet: the only immortality conceivable. What remains when the self is reclaimed by entropic forces — if the digital clone tweets on.

“Words don’t mean anything.”

Machine Listening Aesthetics: Reanimating Sonic Ecologies

Thu, 29 May 2025 00:00:00 +0000

This doctoral thesis investigates machine listening, defined as the computational automation of sound recognition and event detection, through two original conceptual frameworks: soundscape addressability and algorithmic filtering. Working against the tendency in cultural criticism to diagnose AI’s social biases without engaging its technical substrates, the research develops a critique that moves between signal processing, archival theory, and philosophy of perception.

Through artistic experimentation with a corpus of Australian Magpie vocalisations, the thesis demonstrates how machine listening systems classify and recombine sound at scales inaccessible to human audition. This is not done by imposing verbal categories onto pre-existing sonic data, but by synthesising categories from statistical structure and bioacoustics knowledge. Soundscape addressability theorises the archival and bibliographical dimensions of this process; algorithmic filtering challenges the sieve metaphor common in neural network discourse, arguing that filters produce rather than refine categories.

The central argument is that machine listening does not substitute human perception but redistributes its labour — a technological doubling that retains deep dependence on human judgement despite performing the appearance of autonomy. Social biases, in this account, are not intrinsic to algorithms but emerge contextually, through the operation of these systems within complex sociotechnical environments.

Dubbing Point Processes

Thu, 13 Jun 2024 00:00:00 +0000

In this project I play with the implications of sampling and dubbing tape recordings from the John Hutchinson’s sound archive with stochastic processes. Hutchinson, a self-taught field recordist, began capturing these unique sounds in 1959 during his work with the Department of Agriculture. As a result, this archive offers a captivating journey through time and diverse regions of Western Australia. Given the sheer volume of recordings (over 130 hours!), a manual exploration (solely by ear) proved impractical. Instead, I adopted a hybrid data-driven approach to sonic foraging, implementing a sampling mechanism to extract “aural summaries” from the most intriguing parts of the archive. Imagine creating a sonic “thumbnail”, but instead of compressing images we are compressing entire archives by creating representative mosaics with their most salient one-second sound fragments (see a spectrogram representation of the latter in the picture shown above). For this purpose, I used stochastic point processes to select fragments with high diversity. This ensured a representative selection, fast browsing, and avoided the redundancy of (uniform and independent) random sampling. Of course, what counts as representative depends on the task at hand. To address this issue, I relied on Wavelet scattering networks. These are like deep learning networks but with filters defined a priori in terms of Wavelet functions instead of filters approximated from data. Finally, in order to showcase the potentials of this technique, I programmed a bespoke interface to mix the fragments live. The resulting aural exploration was presented at the Data Visualization Institute, University of Technology Sydney.

The interface has a granular sustain. Each fragment has its own track with a send to FX. The balance of effects can be controlled before the master output. The number box next to ‘T’ is for transposition. Loop start and end are shown in samples. Sample # represents a draw from the archive using DPP as described. Since actual files from the archive are retrieved (as opposed to being synthesised), it is possible to use the retrieved fragments as entry points to previously ‘unlistened’ regions (i.e., sonic foraging).

Aural summaries: Unveiling diversity through stochastic sampling

Imagine exploring a vast sound library. How do you quickly grasp its essence without listening to everything?

Let’s put it in a different way, how do you read through the pages of an infinite book of sound without losing your mind? You can’t use bookmarks because the pages fall through your fingers.

I tackled this challenge (and kept my sanity in return) by employing a method called Determinantal Point Processes (DPPs). DPPs are mathematical models originally used to represent the behaviour of fermions (such as electrons), which naturally repel each other (in theory they cannot occupy the same quantum state). This same principle can be applied to sampling sound files! Turns out DPPs can help us to select sub-collections of short audio snippets that are as sonically diverse as possible. Think of it like drawing sounds from a bag but the smart way, you’re guaranteed to pick contrasting textures, avoiding redundancy. Along these lines, a recent paper proposed DPPs to address data exploration in very large audio recordings. This is of great value in ecoacoustics, where audio recordings are made longitudinally for years to understand environmental changes through persistent patterns in soundscape dynamics.

So, how does it work? First, a wavelet scattering network (or for the pros, an adequate feature extractor) is used to analyse each sound fragment, essentially creating a unique “fingerprint” (or feature space) that captures sonic characteristics of interest (see the picture below for a 2D representation of such fingerprints). DPPs then leverage this fingerprint to select audio fragments that differ as much as possible (i.e., repel each other) in the feature space. Unlike other methods which sample fragments uniformly at random, DPPs do not get stuck repeating similar sounds. Intriguingly, this property is given by a mathematical structure from which the processes derive their name: the probabilities of random subsets are assigned according to the determinant of a function. In our case this function is expressed as a matrix of numbers, the output of the Wavelet scattering network for all sound fragments.

Overall, with this approach between high-mathematics and experimental sound sampling, I wanted to push the boundaries of curation and soundscape composition. What a better way to do it than to improvise with a whole archive of wild sounds? These days most generative systems are based on prompts and do not reference their own archival origins. In contrast, with this experimental method of sampling, I advance a new form of creating sound mosaics through archival listening. No longer passively prompting with words, but dubbing and remixing fragments from the archive in real time ala musique concrète. A hot pot of sounds, a mishmash of Wavelets turned Sukiyaki for the ears!

In a more media-archaeological sense, which is a particular mode of inquiry into how media works, I performed the role of a machine learning model. By listening to a batch of 20 windowed 1-second fragments from a whole archive, I attempted to produce soundscapes via improvisation, a genre reflecting the trial-and-error nature of such models. Therefore, the main points put forward by this performance are:

Most machine learning models do not implement smart-sampling methods like DPP to create training batches (see this for an exception). In this case it was me who curated what DPP had already sampled from the archive.
Machines ‘listen’ through standardised filtering or data-driven feature extraction and not by tapping into the fully-embodied experience of a human listening and performing before an audience.
The live-mixing interface displays the potentials of hybridising human and machine capabilities in archive-based soundscape composition. In cultural theory, these cybernetic chimeras appear to have been denoted assemblages or social machines.

Credits and event documentation

Special thanks to Andrew Burrell and Zoë Sadokierski for their support and including my proposal in the programme.

Dadamining Datamining

Wed, 18 Oct 2023 00:00:00 +0000

Data mining methods for animal sound are used to produce a multichannel soundscape. Sound fragments of a Western Australian archive are subjected to two Machine listening algorithms. The first one assigns numeric addresses to soundscape regions by signal similarity (i.e., neighbouring sounds have similar sound textures). The latter segregates the products of algorithmic composition, a monophonic recording, into six audio channels. As a whole, the resulting soundscape displays a mundane industrial process of datamining sound with no analytic purpose but a dadaist impulse of listening to found fragments. Click here for more information on the scientific use of this sound archive.

In my arts-based research, machine listening is not a double of ear-bodied listening experiences, nor a continuation of computation in the domain of digital audio. As a regime of automation, machine listening constitutes an archival force. Digital sound memories are industrialised with algorithmic addressing schemes and filtering protocols which are never fully autonomous as the myth of artificial intelligence suggests.

If sound ecologies are open-ended and more-than-human, how can the (re)generative performance of machine listening subvert reductive technical desires (i.e., anarchive) what has already been fully pre-empted and automated in the archive?

This work was premiered in Melbourne (Australia) at the festival NONSTOP WKND: Planetary Auditions.

Binaural recording available here.

Birdsong Phrase Classification With Siamese Neural Networks

Wed, 01 Jul 2020 00:00:00 +0000

As part of my masters thesis I developed a “Shazam” for birdsong based on siamese neural networks, a few-shot machine learning technique capable of recognizing Cassin’s Vireo song elements. Below you will find the recorded live stream of my thesis defense.

Abstract: The process of learning good features to discriminate among numerous and different bird phrases is computationally expensive. Moreover, it might be impossible to achieve acceptable performance in cases where training data is scarce and classes are unbalanced. To address this issue, we propose a few-shot learning task in which an algorithm must make predictions given only a few instances of each class. We compared the performance of different Siamese Neural Networks at metric learning over the set of Cassin’s Vireo syllables. Then, the network features were reused for the few-shot classification task. With this approach we overcame the limitations of data scarcity and class imbalance while achieving state-of-the-art performance.

Disecciones Sobre Planos

Sun, 01 Dec 2019 00:00:00 +0000

Dissections over planes. Essay(s) from Tlatelolco is a transmedia project directed by Pablo Martínez Zárate that explores one of the most emblematic architectural sites in Mexico City. Through a combination of a webdocumentary, a book, a VR installation and a live cinema performance, incisions are made over Tlatelolco as an element in Mexico City’s landscape. I developed the VR experience using various media, including digitized analogue recordings, drawings and photographs of a model of Tlatelolco. Pablo composed a soundscape that was part of the VR version.

To explore the web version please visit: dsctlatelolco.net.

Posters

Award

The project was awarded by Festival Internacional de Cine con Medios Alternativos (FICMA).

Noise City

Wed, 01 May 2019 00:00:00 +0000

This work is under development. One of the objectives is reflecting on the psychological effects of the noise in Mexico City. Up to date we have a Max application that distorts video by using the audio signal. In this way acoustic noise is translated into a visual experience. Stay tuned. I want to take this idea to an augmented reality platform.

El Cuerpo es un Archivo

Tue, 01 May 2018 00:00:00 +0000

El cuerpo es un archivo is a 360º video montage with multichannel sound inspired in the Mexican Movement of 1968. It is part of the permanent exhibition at Centro Cultural Universitario Tlatelolco in Mexico City. During the performance dancers interact with historical photographs. In response, a dance unfolds leaving the viewer immersed and inviting the audience to engage with the archive (of Tlatelolco Massacre).

The project was directed by Pablo Martínez Zárate and involved an interdisciplinary team. My job here was setting multichannel audio and mapping a 360º video over a cylindrical screen using 6 short throw projectors. Below you can see the process of wrapping the image to the curved surface with the MadMapper software.

Format: 8,16, 35 y 120 mm + HD video
Duration: 15:00 min

Process

Final work

Credits

Director, screenwriting and photography: Pablo Martínez Zárate.
Choreography: Sociedad de Carne y Hueso (Aura Arreola, Teresa Carlos, Raquel Salgado, Marcela Vásquez y Rodrigo López).
Illustration: Santiago Moyao.
Assistant director: José Luis Rangel.
Postproduction: José Luis Rangel, Pablo Martínez Zárate, Hernán Perera.
Cameras: Pablo Martínez Zárate, Hernán Perera, Leonor Castro Guerra, José Luis Rangel, Pablo García, Luis Suárez.
Audio: Pablo García.
Programming: Santiago Rentería.

With the support of Universidad Iberoamericana, Biblioteca Francisco Xavier Clavijero and Laboratorio Iberoamericano de Documental del Departamento de Comunicación.

Bahidora Festival

Thu, 01 Mar 2018 00:00:00 +0000

Spacetime, a multidisciplinary architecture and design agency, invited me to collaborate at Bahidorá Festival with two art installations.

The first one, named Colormancy, involved the interplay between code, divination and chance. By mapping people’s names to particular colors custom ambiences were created out of light and smoke. Eventually a quote based on song lyrics fragments was shown as a personal fortune in a screen. A Raspberry Pi and a touchscreen were used for controlling the whole interaction: receiving users’ input, automating multicolor LEDs and smoke machines via DMX and electronic relays.

The second installation used no electronics but played with basic shapes translated in space and time.

Migrante

Fri, 01 Dec 2017 00:00:00 +0000

Migrante is a stage performance that describes the journey of a character (that you can be yourself) in six stages. Migration, from the metaphorical point of view, is a search in which all human beings are immersed either for need or will. From this phenomenon emerges a sense of migration from the awareness of change, that allows us to question our mission in life, as we assume to be free to chart our own course and write a personal (his)story.

The project was directed by Bernardo Rubinstein with the goal in mind of generating an interdiscipline of dance, emotions and sound. While its nature remained physical, it also involved theoretical reflection around the concept of migration and its historic meaning.

I consider Migrante as one of the most challenging experiences in my life, because it involved embracing my body as a canvas for emotional expression. A task not very usual for a musician.

If you want to read the essay I wrote for the performance please click here.