Wittgenstein's Octopus
Or, how to make a cephalopod AI

by Dr Stephanie Cussans Moran

I. First Contact

They’d pretty much given up on the project by then, written it off like an overripe pomegranate, to mix a metaphor. Or, not exactly written off; it had achieved some level of entity status on the plane of immanence at least. Alasdair called it a hypotype. Enough of a known conceptual entity that its publicly shared aspect had been scavenged for parts by marine biologists and artists, to live on in an afterlife of half-digested, watered down, more funder-friendly forms.

The email came from nowhere apparently, although there were appropriately labyrinthine connections, as they would later realise.


Dear Magie de la Mer and Dr Gibson,

I apologise for contacting you out of the blue like this. My name is Aspasia Pepromene and I represent a foundation called Ariadne that is working to bring a new socio-technical paradigm into being. We fund technological research that has liquid and wave-based, living, analogue ontologies from direct environmental inputs, to counter the prevailing dematerialised, mechanistic digital atomism that has caused so much planetary damage. When I came across your prototype designs for an analogue AI architecture that engages directly with the fluid dynamics of an underwater environment, it seemed to me that this project and your team’s particular position offered a perfect fit with our objectives and, moreover, that it might be eminently viable in our local environs and its research resources.

I therefore took the liberty of making some enquiries as to its feasibility with our local research institutions. I have as a result determined that the necessary research infrastructure does indeed exist here on the island of Crete. Our foundation could make available the funds to set up your institution and develop the AI, subject to agreement of satisfactory terms. This may be rather presumptuous but I believe that together we could make your prototype, and the institutional environment it requires, a reality.

May we schedule a call to discuss further?

Με εκτίμηση,
Dr Aspasia Pepromene | Δρ Ἀσπασία Πεπρωμένη
Director, The Ariadne Foundation | Διευθυντής, Ίδρυμα Ἀριάδνη
Heraklion, Crete | Ηράκλειο, Κρήτη
Greece | Ελλάδα


Of course, Mer, JMZ and the rest of project ISCRI were thrown for six by this missive. Mer, visual artist of collective 0rphan Drift, which represented one half of the collaborative project ISCRI, was almost inevitably first to pick the message up. Mer specialised in machine and octopus vision, a telepathic animal communicator, and swimmer with octopuses. The visionary speculative force behind the project, whose 0rphan Drift work If AI were Cephalopod had kickstarted everything. She was also always impeccably on top of all her comms channels, even the archaic email. Project Manager and Biosemiotechnical researcher JMZ of Etic Lab, the AI consultancy making up the other half of the collaboration, was between projects and on one of her regular immersive cultural artefact dives (submerged under several tons of actual physical books and papers on seventeenth century proto-feminist, proto-science fictional literature) when it slid unnoticed into her inbox. Forwarding the email to the whole group on a priority Bivalve channel that had lain dormant for months, it was therefore Mer who asked,

"Is this a hoax? What is this Ariadne Foundation?"

Picking up the ping from the depths, JMZ (pronounced ‘Jims’) refocused. AKA Dr J.M.Z. Gibson, she had long gone by JMZ. To the extent that no one except Bexley knew what the initials stood for, and no one dared ask either the fiercely private taekwondo black belt JMZ or the terrifyingly cyber-omniscient Bexley. JMZ’ extensive trawl of socials and edit feeds turned up little about either the mysterious Foundation or Dr Pepromene, however; even a crawl of the old net hardly added to the intel, except to establish the Foundation’s existence as a legal entity and its extreme longevity. It was registered in Greece, and seemed to go back pretty much as far as records went.

JMZ reached out to Bexley to deploy their specialist skillset. Bexley, even more hyperconnected than Mer, was already on the case and had routinely compiled a comprehensive dossier within nanoseconds of Mer's message hitting the back of their retina. Bexley Neon was Etic Lab’s Cybersentient, a synaesthesiac of sensory experience and online sentiment, cyber-tracker/-locator/-stalker (depending on your perspective) extraordinaire, a net whisperer and procurer who made things happen. They now shared their unusually thin results with the group.

Ariadne Foundation Profile

Founding date: unknown; mentions of an Ariadne organisation go back at least to the 1670s, in the aftermath of the Ottoman-Venetian Cretan war. There are rumours, on some of the deeper levels of the net, that their shadowy history can be traced back to the ancient matriarchal civilisation on Crete called ‘Minoan’ by archaeologists. It is suggested that Minoan priestesses and council members went underground when their society was colonised by the Mycenaeans. And that they have continued in secret ever since.

Organisation: research institute and investor with interests in marine and extraterrestrial technologies. Org structure appears to be a flattened hierarchy, with a female and female-identifying leadership.

Current Director: Dr Aspasia Pepromene. Background as a reproductive genomics specialist, known as a scientific lateral thinker, pioneer and connector. Her public profile dropped off a cliff when she joined the Foundation a decade ago; no recent images found.

Mission: unknown.

Etic Lab Managing Partner Alvin Hubertus, Bexley’s twin and complementary synaesthesiac of mathematical experience, observed that for a long-running organisation to leave such slight digital tracks, they must be at least either incredibly wealthy, powerful, or well-connected. Hube, as the team generally called him, when not engaged in hard-headed business negotiations, spent large portions of his working nights inhabiting the outer reaches of abstract non-Euclidian space time (like many inventive coders, he seemed to produce his best work between 1 and 4 in the morning; perhaps something to do with the thinning of the veil between dimensions, as dream and waking consciousness begin to blur). Where the others might think through problems in language or image, Hube thought in mathematical dimensions, statistical probabilities and virtual vectors. As he considered the problem, he drifted onto his smart device and into a data coma.

Mer picked up on the Minoan reference, recollecting the 60s and 70s goddess movement expressed in the paintings of Monica Sjoo and Judy Chicago, or the performances of Betye Saar, Mary Beth Edelson, Carolee Schneemann and Ana Mendieta. JMZ had read historic (and by then unfashionable, replaced by feminist cyborgs) 70s feminist archaeomythological discourse, having devoured books with such promising titles as When God was a Woman, The Chalice and the Blade, and The Living Goddess, in the 90s while a young undergraduate. It turned out the once controversial work of Marija Gimbutas, Riane Eisler, Peggy Reeves Sanday, Heide Goettner-Abendroth, Shanshan Du, and Gerda Lerner had been more recently revisited in scholarly publications too. The ancient matriarchal social structure of Crete, an example of what Goettner-Abendroth called ‘egalitarian societies of consensus’, was reviewed over a subsection of a chapter in a work by anarchist anthropologist David Graeber and archaeologist David Wengrow. It had also been the topic of a monograph by retired Librarian Joan Marie Cichon. The Prof, Etic Lab’s eminent Cognitive Architect, was the first to voice the suspicion that this Foundation was a crazy but well-funded and powerful cult, perhaps some kind of Mediterranean matriarchal version of pop culture’s Hebridean Wicker Man.

In typical ISCRI style, the ensuing philosophical discussion meandered around the statistical probabilities of this scenario, the Crete evidence base, and pros and cons of engaging with such an organisation should that be the case. The Prof was all Yorkshire doom and gloom, while Symmer was in enthusiastic favour of engaging further to find out more. Symmer, 0rphan Drift’s specialist in developing Reinforcement Learning environments with haptic interfaces for immersive gaming, was always keen for new experiences. JMZ, scanning quickly, explained that in the most recent, comprehensive monograph, Cichon had investigated a number of possible Minoan social models based on the archaeological evidence, ranging from full matriarchy - a term disputed by Gimbutas - to more gender balanced modes. Two of these draw on anthropologist Shanshan Du’s egalitarian frameworks, based on her fieldwork in Southwest China, which offer alternative models. One being what she calls ‘gender triviality’, a predominantly ‘gender-blind’ social structure that favours individual autonomy and collective cooperation; she calls the other a ‘dyadic or gender-similarity’ framework, which minimise socio-cultural sex differences. Therefore, JMZ concluded, it is too early to entertain a stereotyped hypothesis of rampant matriarchalism as an inverted patriarchy; moreover, the Minoan evidence suggests a peaceful, contented population of women and men; none of the many artworks discovered depict sacrificial rites, or violence of any kind in fact. Hube at some point emerged from his data revery to pronounce, in typical sweeping and gnomic fashion, that the Foundation was hiding something. Mer was quick to offer the octopus perspective, explaining that octopuses [the correct term as octopus comes from the Greek, despite the more popular Latinised plural ‘octopi’] were prominent decorative and symbolic motifs in Minoan and Greek art and mythology, from which legendary figures such as the multiheaded Hydra and the snake-haired Medusa were derived. The Minoans also worshipped a snake goddesses. Snakes were thought to represent the life-force; it is possible the octopus shared that symbolism, often conflated with snakes as in the Hydra - a kind of multi-snake entity. The octopus was known as ‘polymetis’, or multi-wisdom and cunning, a tag shared with Ancient Greek hero Odysseus. Although Ancient Greece was later than Minoan, and had probably appropriated some Minoan culture. She added, the octopus could have been sacred to the Minoans, given the large number of octopus representations found, often appearing as spiral motifs on vases.

Six hours later, they’d argued around to the conclusion they were always going to reach, that they should meet with the enigmatic Ariadne Foundation. Then as a courtesy they messaged project stakeholders Evelyn and Alasdair at CreAIteur Incubator, a spinout from Ophidian art gallery that holds a repository of AI projects by artists, whose purpose is to represent their interests in negotiating with industry investors from gaming and other immersive AI growth sectors in need of creative input. This pioneering model helped artists to profit from their innovations, halting an emergent industry of digital IP (‘intellectual property’) theft and the practice of employing artist devs on a fraction of the wage of a tech sector coder. Evelyn was a supercool, unflappable young curator, perfectly verbally and visually eloquent. She worked with Alasdair, an art-tech philosophy geek who had become ISCRI’s unofficial biographer. CreAIteur had been working with ISCRI before the spinout, supporting the project’s aesthetic and conceptual development from a sophisticated contemporary art viewpoint, and with great warmth and generosity, until they had reached a mutual wall: the project was conceptually complete as far as it could be without the development work, but lacked the significant levels of funding required to carry that work out.

Having received Evelyn and Alasdair’s symbolic blessing, Mer, JMZ and Aspasia worked out the details of meeting time and agenda via email. Aspasia requested a presentation giving a technical overview of proposed work, for an initial five-year plan; in return she offered to reveal more of the Foundation’s aims and where ISCRI might fit with them. Aspasia also requested that the call take place on Etic Lab’s own end-to-end encryption video communication software, with extra privacy protocols in place.

“We are, as you have probably realised by now, a very private organisation. We do not wish to risk being overheard. We usually only meet in person without our smart devices present.”

The software was usually only used internally by the Etic team, so Aspasia - or someone at the Foundation - had clearly already carried out a high level of due diligence checks on the Lab. There was then the usual issuing and signing of respective legally binding NDAs (non-disclosure agreements), to protect the IP and confidential business information of all parties.


II. Negotiations

They assembled on the dot of 2.30 on the agreed day, with the punctuality and immediacy afforded by the meeting’s virtuality that was still unnerving even years after it had become commonplace. Etic Lab’s Ghost had set up the requested privacy protocols, with the ISCRI team meeting in person at the Wales office, around an app-less device unconnected to any personal user and away from all other devices, that was only switched on when required and accessed via a VPN currently set to Switzerland. Their own personal devices were left behind or placed in Faraday pouches before entering proximal space. This was for most of the team paranoia level H (high), although by the Ghost’s level it was pretty low. In general, the Ghost was never seen online, and only heard (like the Angels’ Charlie); even this was through vocal filters. Likewise, the Foundation were all together in their own conference room in Crete, following the same protocols.

First, introductions. On the ISCRI side, spread around the couches of the chilly and somewhat untidy Etic head office deep in rural mid Wales was Mer, JMZ, Bexley, Bexley’s pet AI chatbot Geraint, Hube, The Prof, Symmer, The Ghost, 0D computational artist Corvin who specialised in coding animal skin patterning, and Etic Lab newcomer Mariella Ayodele. Mariella’s area of research was comparative epistemology and the cognitive neuroscience of religious experience; she was also a skilled divination practitioner and researcher of African traditional religions. She had approached Etic Lab with a view to becoming an Associate a couple of months before the mysterious Ariadne invitation. ISCRI’s research phase had concluded that the first philosophical-cognitive step towards interspecies communication is developing the capacity, or at least willingness, to shed epistemological assumptions and to speculatively enter into different onto-epistemological paradigms. They had reached their conclusions through a cross-disciplinary approach, spanning New Materialist and posthuman hydrofeminist philosophy, Mer’s telepathic animal communication swimming with octopuses, as well as recent zoological and cognitive scientific research. As such, Mariella’s presence here - although not previously part of the team - represented a new stage in their thinking. Mariella had an immediate rapport with Aspasia, who introduced the members of the Foundation present.

Aspasia presented the members in attendance as the Foundation Council. There were nine of them including Aspasia, seated around a circular conference table in an airy office brightly lit by the Cretan sunlight filtering in through what looked like a forest of vegetation covering the outside of the large windows. They were introduced by first name, pronouns and area of responsibility only, with no bios or personal information offered: Artemis (they/them/their), zoology and livestock; Demetria (xie/hir), botany and agriculture; Hera (she/her), economics and ecology; Melissa (co/cos), tech and comms; Athina (she/her), archives and information; Selene (she/her), cosmology and astrophysics; Kallisto (they/them/their), buildings and community; and Posidaeia (she/her), marine ecology. The responsibilities seemed a weirdly eclectic mix, but not much stranger than the range of research represented by the ISCRI team. There were notably no ‘he/his’ pronouns given, rather a mix of ‘she/her’, ‘they/them/their’, ‘co/cos’ and ‘xie/hir’. Although a familiar organisational makeup for 0D accustomed to the world of contemporary art, it marked a welcome change for Etic Lab from the usual tech industry ‘bros’ and machos that dominated the VC investor landscape.

Mer and JMZ were ISCRI’s opening presenters, beginning by sketching out the project’s philosophical basis and aims.ISCRI presentation title slide

J: Here, we apply our understanding of Wittgenstein’s private language ‘argument’ to the problem of interspecies communication, in the context of more recent whole-systems thinking. ‘Private language’ refers to the idea of a language that is only comprehensible to its [sole] originator, because its vocabulary is based on things that are inaccessible to others - whether phenomenological or conceptual, or most likely a combination of both. Wittgenstein contends that the existence of a private language is impossible, that there is nothing in language that is inherently inaccessible or hidden. Language can only discuss what is open to the world.

slide 2, there is no such thing as 'private language: nothing is unknowable in language because we inhabit a shared environment.

This makes perfect sense following Maturana, Varela, Gregory Bateson and J.J. Gibson. Ecological systems thinking and ecological psychology make clear that no organism can exist separately from or communicate outside of its environment, and therefore any communication must be knowable within that environment. We define communication here as an exchange that effects a change in both (or all) communicating parties. Meaning is always created relationally, co-evolved within shared environments and cultures, and so there can be no inherently unknowable or inaccessible 'internal' language. Which underlies our approach to interspecies communication - an understanding that because 'language' (or, better, communication) is inseparable from its environment, it can be phenomenologically modelled, from the outside, through behavioural and environmental observation (in this case via an AI).

slide 3, Why an Octopus? They display cross-species curiosity and play and collaborate with other species.

Most current interspecies communication AI projects are based on social species that communicate verbally between themselves, as well, of course as gesturally and visually. The projects tend to focus on deciphering verbal communication, sometimes using the visual cues for context labelling. Machine Learning (ML) algorithms are capable of parsing humanly unimaginable volumes of data at speed once set up, and have been able to identify patterns in the audio data at this scale. We have in this way learned that many other species, from bats to birds to whales, have verbal patterns of communication similar to human language: they have syntax, grammar and proper nouns (individual names for each other).

Octopuses represent a particular challenge and opportunity. A challenge in that they are solitary animals or, more accurately, largely asocial among their own kind, with the exception of some rare sites where they have been forced into closer cohabitation; either way, communication between themselves is usually limited to signals that seem to convey “stay away”, “fight” or “fancy a shag? / want to breed with me?” and “sure thing / no thanks”. The opportunity is that they are highly intelligent, curious, and are known to form relationships with other species. They display interest in other species including humans, apparently out of curiosity, or for purposes of obtaining food or engaging in play. O. cyanea and other reef-dwelling octopuses are known to collaborate with fish in hunting, for example, and to engage in cross-species signalling to achieve this. Octopuses have been observed to play games with both fish and humans, and have been known to form bonds with humans. Where they communicate, so far as we understand it (and we cannot make any claim to understanding), it is often across species and visually rather than verbally.

So we have focused on learning to communicate, using the mediation of an AI (artificial intelligence) in their dynamically fluid environment, rather than trying to learn a ‘language’. That is, we wish to construct a system that invites the octopus to communicate with us, should it wish to. This is where our AI gaming model comes in, deploying deep reinforcement learning (DRL). (D)RL is a form of ML used (among other things) for gaming and gamified environments for AI such as ChatGPT. Our use of it is intended to take us beyond the mechanistic models of Skinner (in psychology) or the Churchlands (in neurophilosophy). Our model is based on observed ecological-cultural behaviours within and inseparable from their environments.

Slide 4, Communication through Play, an AI gaming model, a system that invites a response.

M: 0rphan Drift have created visual computational prototypes, developed from experiments with forms, textures, patterns and speed of video streaming, imagining into what we understand - from ISCRI’s research - may spark an octopus’s recognition and curiosity. Art for octopuses, drawing on classic looming stimuli simulating a possible approaching predator; flitting stimuli of varying shapes and sizes; patterning simulating octopus and other cephalopod and local marine species’ skin pattern changes. We have been experimenting in 3-D software such as Blender, incorporating Corvin’s octopus skin patterning coding and Symmer’s work on visual-haptic objects in virtual environments, and transmuting the resulting 3D-visual vocabulary into polarised displays. Polarised light is made up of lightwave oscillations on a plane, a phenomena undetectable by human visual apparatus but that octopuses can detect.

J: We’ve drawn on recent research, in particular Shelby Temple’s work at the University of Bristol’s Ecology of Vision Laboratory, exploring octopuses’ polarised vision. The research suggests that polarised light perception enhances cephalopod vision similarly to the way that perceiving variations in lightwaves as colour does for humans. It is thought that at least some species of octopus may also produce polarised light signals on their skin, as a form of communication invisible to animals such as humans that do not possess this sense. A kind of secret code. Polarised light forms are hard for humans to detect even with technological prostheses, but Temple et al. point to a number of effective methods. Polarised signals could be intended for their crustacean prey, fish collaborators or shark predators.

M: Elements from the 3-D ‘visual vocabulary’ we have developed will be streamed intermittently for octopuses via modified liquid crystal and holographic displays in the sea (represented in the system diagram, fig. 1, as ‘emitters’) to deliver dynamic polarisation stimuli during the project lifespan. Individual Common Octopus participant candidates could readily be found in the waters off the coast of Crete, where they are numerous. They inhabit the shallower waters, and tend to live their lives within a few metres of their dens. This should make it relatively easy to trial communication with a number of individuals. The AI artefact will mediate the image content as it learns from the octopus responses via a multi-sensor array capable of detecting a wide range of sensory cues (‘detectors’ in the system diagram). Thus each octopus would have control over what is streamed; or, to put it another way, would be able to communicate through their interaction with it. Whether or not the octopuses choose to engage, the model will learn to become part of its environment, as the sensor array will be observing each octopus and its environment. Changes in the rate of change will be reflected in changes in, or edits to, the visual interface. The streamed content will change over time in ways we cannot predict, as the AI artefact learns.

slide 5, Visual Vocabulary, communicated via modified LCD and holographic displays

Bexley: The sensor array aggregates as many underwater environmental sensors as are available, including audiovisual recording, temperature, light level, colour spectrum, water pressure, movement, audio and a range of chemical sensors.
Symmer: This facilitates the procedural environment through which the model can learn, and an octopus can respond, if it wishes.

Slide 6, system diagram showing the reinforcement learning system, learning from the octopus in and as part of its environment.

Hube: In this way, we anticipate that the AI will learn visual language model protocols and guidelines that are immanently structured through the epistemic ends or values of its environment [unlike ChatGPT, which has a fudged epistemological teleology due to its use of subjective labellers; an effect of Silicon Valley platform capitalism].

J: People find the idea of communicating through the whole environment hard to grasp, although it’s not a new idea; it is related to Maturana and Varela’s insight that nothing can exist outside of its environmental context. It’s just that most people in the west think of themselves as individuals who are separable from their environment, who dominate their environment. And we extend this idea of individuality, to a lesser extent, to other animals too.

slide 7, Learning from an octopus in its environment. Changes in the environment, including the octopus, are reflected in the visual output.

The other bit that’s hard for people to wrap their heads around is how the visual learning works. It’s a rather more sophisticated version of Geraint, Bexley’s pet AI instance.

B: Geraint was built for me by The Ghost, a curated language model trained primarily on a diet of octopus and fluid dynamics research. When asked a question, whether octopus-related or not, he responds within an ‘octopoid’ epistemology. From a human-centric, scientific perspective; but still, a small step towards cephalopod comms. And Geraint is our working model, a very simple prototype for what we hoped to achieve with the first ISCRI AI. Imagine an AI equipped with a visual interface and a range of sensing technologies that responds to a variety of stimuli including movement, gesture and pheromones, with liquid-dynamic, octopus-influenced imagery.

slide 8, A Visual Language Model Trained by an Octopus, a system that offers an interface for an octopus to edit visual artworks.

Ghost: We have seen how sensing technologies have become central for environmental monitoring since our research began. There is a growing range of marine sensing technologies for the ‘blue economy’. Our work can draw on this new research in developing a wider range of sensory inputs for the AI to learn from.

Prof: We know and communicate as embodied entities.

Hube: The RL model is not the only idea we have been working on. We believe that developing complementary top-down and bottom-up combinations of behavioural and mechanistic models will further our understanding. We therefore propose a triangulated approach, by building complementary experimental prototypes:

• Communicating with an octopus in environment, as already described.
• Based on our research into octopus visual cognition, and especially 0D’s computational visual explorations of how an octopus sees, a series of ML models that simulate ways octopuses use vision, including tactile hapticity, the polarisation of light, distributed luminous and/or chromatic inputs from octopus arms and suckers, and the fluid dynamics of their environment. These may eventually be aggregated and feed into and from number 3.
• ML models based on octopus highly distributed neural architecture and visual cognition.

Together, these AIs would offer us a better understanding of octopus visual cognition that could be applied in virtual simulations of underwater environments and for octopoid robot vision.

Here, Aspasia leaned forward, her face animated. She said, “this is the kind of project that we are well equipped to support, here on Crete. And it could be a good fit with some of our own objectives, if you are interested in collaboration. I believe you are acquainted with Dr Dimitris Tsakiris from his time in Wales? He is now Director of Research in our ancient city of Iraklion, or Heraklion, part of a major Greek scientific research institution (FORTH-ICS, the Foundation for Research and Technology – Hellas, Institute of Computer Science). He has an ongoing robot octopus project that may benefit from octopus-like vision for navigating underwater or in outer space. His lab, the Computational Vision and Robotics Laboratory, would be able to help you with the underwater visual work. Furthermore, this could be supported by the Institute of Oceanography, part of the Hellenic Centre for Marine Research (HCMR), housed at the Thalassokosmos complex, also in Heraklion.”

Another Foundation member who had been introduced as Posidaeia, with responsibility for marine ecology, added, “We’re thinking specifically of the Ocean Physics (fluid dynamics) and Marine Chemistry (underwater chemoreception) Departments, and the Marine bio-optics group whose work on underwater colour, custom underwater platform combining standard CTD measurements and hyperspectral optical measurements, and previous research observing inherent and apparent properties of seawater, could be crucial for your AI project. They could help you to custom develop underwater sensing technologies, including chemofactory and visual sensors. HCMR also has cephalopod expertise at the CretAquarium and in Dr Teuthis and other marine researchers at the Hydrobiological Station of Rhodes, not very far away…”

The ISCRI crew suddenly and collectively had the sense that there was a rather more developed plan on the Foundation’s side than they had even begun to suspect. They had, indeed, met Dr Tsakiris several years ago, in the unlikely environs of Aberystwyth University, when Etic Lab had been invited to work on a collaborative project with the Computer Science department. It hadn’t gone anywhere, but they had mentioned ISCRI in conversation as Tsakiris had been developing his robot octopuses even then. Aspasia prompted, “tell me more about 2 and 3”.

Hube, leaning back, looking down and, deceptively, appearing faintly bored, replied, “first tell us what’s in it for you. What is the Ariadne Foundation?” At the edge of the holocamera’s field of view, The Prof shifted back almost imperceptibly too. A retired Cognitive Psychologist, The Prof was adept at performing reads of others’ intentions but didn’t want to be too obvious about scanning Aspasia and her colleagues’ expressions and gestures.

Aspasia was prepared for this question; of course, she must have been expecting it, she had promised some explanation.

“Those two enquiries are deeply interconnected. Our archivist, Athina, will tell you more.”

“The Ariadne Foundation’s philosophy, purpose, organisational structure and way of being all have their origins in our history. We have, as I am sure you are already aware, a very long history; and these elements are now entirely entwined and inseparable one from another. It is therefore hard to know where to start; but I shall begin with the org structure, and how it emerges from our history.

“You might have noticed there are no male-identifying Council members here. For historical, cultural and symbolic reasons, male-identifying members cannot be appointed to the Council. The Foundation is only the modern form of an ancient society. We are descendant of the culture that lived here on Crete before ancient Greek civilisation. Archaeologists have mislabelled this Minoan culture; a mislabelling that is wrong in both its anachronism and its assumption of a male ruler (they based it on the later myth of King Minos; if he did exist, he lived thousands of years after our first temple-palaces were first built, and hundreds of years after the Mycenaean invasion that destroyed our nation). They made the common mistake of reading the remains through their own cultural bias. Our name for ourselves is Panagian.

“Historically, our culture is what you might call matriarchal: administered by a Council of women, with heritage traced through the female line from mothers to their children. This is the only way when you do not have a nuclear family structure, with a clear father. Our society was traditionally spiritually organised around Goddesses endowed with the symbolic power of various animals; most famously the snake, but also including other animals such as bears, birds and hedgehogs. We have retained, as you may have observed, our matriarchal structure. We use the term ‘matriarchy’ not in the sense of an opposition to or inversion of patriarchy (in fact, it preceded patriarchy), but as a completely different kind of social order. The Greek word ‘arche’, the suffix shared by both, has two meanings. It can mean ‘domination’, or it can mean ‘beginning’. While patriarchy takes the dominance aspect (modelled after the 'dominant father’ or male leader), matriarchy for us is about our shared beginnings, the sacred source of life: our interconnection with and dependence on the earth as well as each other. In ancient Panagian culture this was symbolised by the Goddess, but today might be better understood as the living system Gaia - the Earth herself, the original mother Goddess. This acknowledgement of interconnection is something we recognised in your AI philosophy.

“We have a very long history of maintaining a nonhierarchical, structured society. We are the last remnant of an indigenous social form that once extended across Old Europe, before the invading patriarchal Kurgan culture swept down from the Russian Steppes. We persisted here for longer than the rest of the old world, isolated as we were on our island, too small for the early conquerors to bother with. Cooperation and egalitarianism are our core values, embedded in every aspect of our structure: economic, social, political, and cultural-spiritual. Economically, we are self-sufficient and autonomous as an organisation, we own land and buildings collectively, and we operate an economics of reciprocity between ourselves. Socially, we are clan-based and live collectively in large, extended, intergenerational kinship and affinity groups. Politically, we operate a consensus decision-making model. Culturally, our mode is cooperative, based on connection and the flow of mutual empowerment. We have a Director but she is not a leader in the way you understand, at the top of a hierarchy. Likewise, we have a Council rather than the more conventional Board; this is a structure inherited from our ancient past. Aside from Aspasia, all of the Council members’ names given today are - as you may have guessed - symbolic names, or titles. I cannot tell you the symbolic functions of each Council role, as that information is only for initiates. Suffice it to say that in our contemporary form as the Foundation, we continue to orient ourselves around the sacred living entity and ecosystem Gaia, and that each role has a symbolic function that is necessary to our society’s economic, political and cultural-spiritual life.

“Before the Mycenaean invasion, we did not have the concept of private property or land ownership, only usage rights. When the Mycenaean patriarchy invaded, we did not oppose or fight them. Our way is to evade and persist. We would not change our way of being in order to preserve our society; for by becoming militant, we would have changed our values and patriarchal culture would have won. Once started on that path, it can only escalate into an arms race. We have become like the octopus, an animal that shape shifts, hides and throws blinds to evade predators. We maintained a thread of our culture underground, passed from mothers to daughters in secret across the centuries and millennia. We did not, for many many centuries, commit any of this to writing; it has been passed down orally embedded and embodied in a series of secret rites and rituals, for initiated members. Our culture was never reliant on written text to pass ideas on; we have always had a vibrant immanent philosophy, embodied in our rituals and famous arts - from dancing and music to ceramics and painting. We have over the last millennium begun to record our philosophy, obscured by symbolic language and reference to ritual observations, in a series of esoteric texts. After the Mycenaean invasion we placed our skills at the disposal of the colonisers in order to be able to continue our work. We dispersed, and retained our knowledge through our art forms, injecting at least aesthetic beauty into the new patriarchal societies; something to mitigate the misery of war and oppression.

“Modern men are the inheritors of that warring, patriarchal culture. They are trained to dominate rather than cooperate; to impose rather than to listen and share. They cannot help it, it is deeply culturally engrained and reinforced.” Here she cast a glance at Hube and the Prof. Hube continued to emanate nonchalance, while the Prof looked somewhat sheepish.

“We operate in the real world, in a white-male-dominated sphere,” he offered. “And we have to make a living.”

“However,” JMZ added “it’s been important to ISCRI’s ethical, epistemological and technical integrity that it has evolved outside of the tech innovation economy or the Silicon Valley-style VC [venture capitalist] extropian, transhumanist, accelerationist culture. It has come into being within an artistic paradigm, a space where ideas do not (yet) have to become platforms for acquiring huge volumes of human user data for marketing or socio-political engineering, or as fuel for next generation AIs.”

Bexley qualified, “even in the apparently progressive, idealistic space of contemporary art, where artists make work that challenges colonial, patriarchal hierarchies, they make and exhibit within a highly hierarchical, competitive field largely owned by white upper classes and wealthy elites. Reinforcing and reinforced through institutional gatekeeping practices and internalised hierarchies, ironically often expressed through possessing the phallus of the correct ‘language.’”

Athina nodded and continued, “We are trying to change that, slowly and immanently, through new exosomatic technological forms we coevolve with. Our ‘matriarchal' culture is not an inversion of patriarchal structure; it is an entirely different structure, based on different ways of being and knowing. The all-womxn council is not a patriarchal-style leadership, it is an administrative function. We administer and serve our society, seeking balance, facilitating consensus-based decision-making. In contemporary times, we have moved on from worshipping female Goddess forms, but have retained our prohibition on masculine council members due to the external prevailing socio-cultural conditions. Whether overt or implied, formal or informal, patriarchal organisational culture imposes dominating structures of control where feminine ways of being, or female ontoepistemologies, are implicitly viewed as weak and as such are eroded and crushed. This kind of immanent aggression is incapable of affording the socio-cultural transformations we wish to see. We aim to remove all trace of internalised hierarchies, the effects of millennia of patriarchal trauma. The spiralling form of the octopus is a potent symbol for us in many ways. Our org structure is not arranged hierarchically, in verticals and horizontals; it is, rather, circular. It resembles a curving, spiralling circuitboard, or a labyrinth form, more than a tree or a ladder. Our work has contributed to shifts that are finally now beginning to manifest in serious challenges to this paradigm. We are moving towards a non-binary culture.

“Paul B. Preciado argues that technological advances allow us to alter our bodies and procreate differently; to escape the naturalisation of normative gender identities and behaviours. We plan to go beyond human desire, to pursue an immanent non-binary flow of Gaian interconnectedness. In your project, we see possibilities for learning from other species, including the octopus; for radically changing the human through deanthropocising our technological prostheses. We see this potential in the species-fluid artworks of Keiken, whose interspecies prosthetic pregnancy bellies generate a sense of planetary interconnectedness. The prevailing western socio-technological paradigm has been based on a Platonist, idealised mechanistic binary atomism, suppressing anything that did not follow the principle of ideal forms, anything messy or noisy. It was a science of the dead, of cut-up bodies and unnatural stillness that took place in laboratories and morgues.”

“Yes,” The Prof interjected. “Science created the concept of the white cube before modern art did: the idea of an autonomous space where pure experiments take place, abstracted from the world, that claim to reveal truths about the world.”

“Precisely. An alternative form of atomism has always co-existed, however, pre-dating Platonism in Panagian culture. The spiralling mathematics immanent to the whirlpool, the wheel, the snake and the octopus are sacred to us. The octopus is associated with the moon, summer solstice and the depths. They embody the mystical centre and the creation of the Universe, and we frequently depict their spiral geometries in our decorative artworks. Since ancient times, the octopus has been connected to the Great Mother, the all-goddess Panagia or Gaia. Philosopher Michel Serres shows how these ideas remained in Lucretian atomism, a text that was long banned as heretical. He connects this atomism based on life and liveliness to the mathematics of Archimedes, theorist of the spiral and the contingent, showing how it is a philosophy and a science of turbulence, of fluidity.

“With the fall of matriarchal culture, the octopus became the prototype for the dangerous female figures of the multiheaded Hydra and the paralysing Medusa, part of a misogynistic semiotic and mythological disinformation campaign. Sailors in patriarchal culture viewed the octopus as one of the most hideous and scary sea terrors, strong enough to wreck a large ship, despite a lack of any evidence. Homer’s hero Odysseus is given the octopoid epithet “polymetis,” symbolic of patriarchal vanquishing of the feminine order. The rehabilitation of the octopus in the west, centuries after Lucretius’ poetic recuperation, coincides with the return of a Panagian physics in the form of nonlinear dynamics, fluid mechanics and chaos theory. The lines of flight out of binary, linear patriarchal culture have always been held within its excluded middle. The Cantor set shows how non-binary branching paths carved out of the infinite states between zero and one. This ternary set is the immanent nonlinear, chaotic fluidity of life, of Gaia. It leaks and then bursts out joyfully. Simulacral worlds and socio-technical relationships based on a substrate of binaries, of zeroes and ones, can equally enact the non binary occult evasive vectors of their own emancipation.

“We invest in projects we believe will help push this agenda forward, to work with Gaia rather than pursuing the impossible task of trying to contain her; and to develop technologies for cohabiting with all Gaian entities, through pleasure rather than conflict.”

Athina stopped with finality, and Aspasia resumed, “if you are satisfied with our account of ourselves, I would very much like to hear the rest of your plan now.”
Mer, JMZ and Bexley were beaming with enthusiasm for the Panagian vision, while Hube was now visibly enraptured. The Prof had an expression somewhere between bemusement and incredulous fascination. Without needing to confer, JMZ and the Ghost took up the presentation where Hube had left off.

J: The second triangulation vector is to prototype new kinds of octopoid AI vision artefact, or octopus “seeing objects” as I like to think of them, in an analogy with Svetlana Alpers’ conceptualisation of seventeenth century Dutch paintings. These will adapt and invent new forms of Deep Reinforcement Learning (DRL) for computer vision, to imitate what we think we know about octopus vision rather than human vision. For example, we might train Convolutional Neural Networks (CNNs) for saliency and object detection, video analysis and object tracking on footage taken by underwater polarisation cameras. This could be enhanced with peripheral light and colour detection, in imitation of visual inputs from octopus arms, suckers and skin. Essentially, we could train a CNN to produce any potential feature of octopus visual perception. Unlike seventeenth century Dutch paintings, the octopoid seeing objects will mediate moving, embodied points of view, learning from visual data and enhanced by other sensory data, as vision is not static and not separable from other senses. Seeing objects based on coastal octopuses of the shallows and reefs, such as the Common Octopus (O. vulgaris) and the Day Octopus (O. cyanea), will be optimised for dynamic vision in fast-moving underwater environments. Seeing objects based on octopuses of the deeps (like the octopus-squid Vampyroteuthis infernalis) will be optimised for vision in low light conditions and bioluminescence. This research and prototype development will inform and be informed by developments from the other two strands. It will output simulations of what octopuses might see, from modified video streams; we hope it may eventually incorporate more of the distributed visual information prototyped in experiment 3.

slide 9, Octopoid Seeing Objects - a behavioural simulation of octopus vision.

G: We envisage initially training these in simulated, virtual underwater environments based on existing ocean models and the learning from our first RL model. If we were able to combine these prototypes with Dimitris’ octopus robotics, we could train and test them in real marine environments.

J: There is a growing body of research on cephalopod cognition, a fascination with how octopus brains work, but the relation between ‘brains’, or embodied cognition, and behaviour is complicated by the lack of knowledge of octopus behaviours within their environments. This is hard to study, due to the difficulty and expense of underwater research and the fact that many octopus species live in environments such as coral reefs that are legally protected.

G: Octopus-based robots (that move and see like octopuses, that possess octopus-like features) equipped with underwater video recording capacities can explore and observe octopuses in their environments over long periods.

B: Octopoid robots, optimised for navigation and movement in underwater environments and equipped with a range of sensors could be used more generally for underwater exploration and ocean monitoring. We imagine different kinds of octopoid robot and octopoid computer vision, corresponding to the diversity of octopuses cognitive apparatus afforded by their different environments. The vision and behaviour of octopuses that have evolved in the faster-moving, bright shallows and reefs is distinct from dwellers of the slow deeps, lit mostly by bioluminescence. This is likely to have applications for outer space exploration too, for operating in similarly low gravity, high pressure situations. It could even be useful for somaforming, or bodily modification technologies, for future living in watery or outer space environments - something explored in Pat Cadigan’s SF short, ‘The Girl-Thing Who Went Out For Sushi’.

slide 10, The Shallows, using DRL Convolutional Neural Networks trained for object saliency and detection, video analysis and object tracking. slide 11, The Depths, with CNNs optimised for the low-light conditions, bioluminescent illumination and slow speed of high pressure environments.

M: The third side of our triangle is Machine Learning models based on the distributed neural architecture of octopuses. Think about the way your hand reaches for your device when it buzzes, before you’ve consciously made the decision; the way your fingers type, directly interfacing sentences without having to consciously make a decision about each letter or word. Imagine having outsourced a greater level of autonomy to each of your limbs, and having evolved in a way that afforded and was afforded by eight arms and their fingers reaching out in different directions, each containing visual, tactile, gustatory light- and chemical-sensing interfaces. A brain in each of eight arms, each moving in spiralling, Gaussian motion; that are semi-autonomous, to a greater extent than the autonomy of human limbs.

slide 12, Distributed Neural Architecture, for decentralised ML models that learn from heterogeneous data streams.

H: We plan to design and build new, distributed decentralised ML models. In some ways, these might be similar to our existing Federated Learning (FL) models that learn autonomously from heterogeneous datasets held in different locations, pooling the results of their learning without sharing the data. They will be trained, in an unsupervised fashion in a simulated underwater environment, on sensory data streams from experiment 1. They will have the capacity to extend light and colour receptors out in different directions via a virtual bend propagation mechanism, in spiralling pursuit of interesting or novel information. A centralised model will pool that data with optical video data, allowing it to override any of the ‘arms’ autonomous explorations if more interesting or novel information is detected in other directions.

slide 13, Novelty-seeking ML models that simulate an octopus' distributed decision-making and central override.

These are entirely speculative, unpredictable experiments; whatever the outcome, we will learn new things from them.

There followed a long discussion about the project’s feasibility, requirements, ethics and how it could work on Crete. Foundation members were thorough, but not with the usual investor or funder focus on ROI, route to market, investment sought and costings. They had clearly done their research, asking informed questions about the team, with questions put to each of the team about their ethical and political values. Posidaea and Artemis enquired about research ethics, how the team planned to work with octopuses and in the sea in ways that would cause minimal stress to its inhabitants and leave minimal trace. Hera posed the usual questions about environmental, social and governance plans, how ethical and sustainable practices would be pursued. She also cross-examined their rough three-year and detailed first-year plan. Selene asked some specific questions regarding the outer space and interspecies communication applications. The implications seemed to be that if an AI could learn from an octopus in and through its environment, it could potentially also learn from extraterrestrial life forms. The life forms would just need to be detectable via a sensor array. In theory, this was the kind of problem the Institute of Astrophysics (IA-FORTH) on Crete might be interested in.

In return, the ISCRI team asked about the resources, support and marine research infrastructure that would be available to them on the island. It seemed that most resources and research needs, including marine researchers and additional developers, could be provided through the existing marine and computational institutions spread across Crete and Athens, and through the Foundation’s membership. They would invite some octopus visual cognition researchers, divers and experienced octopus handlers from across the sea at Stazione Zoologica Anton Dohrn (SZN), Napoli, in Italy, although no handling would be necessary for this project. They debated the project org structure, incorporating anticipated partner organisations and an advisory board. Hera began speculating on terms negotiations, IP split… Aspasia leaned in before they could get too far.

“We very much want to work with you to realise this project. The learning and the new technological structures you (and hopefully we) derive from these experiments have the potential to culturally transform cross-species relations and communication, as well as to contributing to a change in our conception of ourselves - from the colonial idea that we are individuals at the centre of our own universes and destinies, to an understanding that we are not separable from our environments or our relations to others, including other species. Physically, bio-modifications derived from this could lead to new adaptations for humans that enable our survival in rapidly changing environments and on other planets, emphasising our embodied and ecologically embedded natures, while bringing us bodily, sensorially closer to other animals, in ways that are not abstracted or detached from our physicality via screens - as in Keiken’s deeply somatic artworks. It could lead to new non-binary and extra-human adaptations.

“Before we proceed, however,” she warned “there are a few risks and necessary mitigation measures I need to make clear on our side. If you do choose to work with us, you should be aware that we have some enemies and there is some danger to your work and to yourselves if they connect you with us. Far-right activists have identified our work as a threat, making us targets in Greece and across Europe. A number of our members are on their hit lists. We are troubled by the rise of ultra-patriarchal far-right parties all across Europe. In Greece, we’ve seen the election of three far-right parties after we thought the threat from the Golden Dawn had disappeared. The Spartans, the pro-Russian party Greek Solution and the ultra-Orthodox party Niki (Victory) all have seats in the Greek parliament. Far-right violence is on the increase, seemingly backed and stoked by hostile foreign state actors. We are fortunate, though, to have a very strong anti-fascist movement in Greece. Indeed, we are safe here on Crete; we threw the far-right out and they have not returned. If they connect you with us, they may begin to grasp some of the implications of your work that will be a threat to them.

“I would like to suggest - and I think this will work well for you in many ways - that you undertake the work here on Crete in secret, and that your institution’s public face and place of registration is in Athens, where you can interact with other marine biological and AI researchers, and the wider community. It is important for this work that we are outward-looking and inclusive. We can protect you while you are here on Crete, and we can arrange for very discreet regular transport between here and Athens.”

This unexpected turn put the stakes into a wider and more immediately urgent perspective. The ISCRI team’s lively conversation slowed as they digested the new information and its implications. Having sketched out a heads of terms, they agreed to reconvene after internal discussions.


III. Initiation

Mer, JMZ and Bexley flew to Crete to set up the Institute and assemble the wider project team, getting things ready for the rest of the ISCRI crew. They arrived in the ancient harbour city of Iraklio (anglicised as Heraklion) in the hot, dry autumn. Medieval, Venetian and modern architectures in pale stone and concrete and painted pastels glowed warmly in the September sun. Fortifications, churches, hotels, cafes, bars, restaurants and shops huddled densely around the shining, very blue curve of the harbour; suburbs and villages flowed back towards the beginnings of purple-tinged mountains. The Foundation had arranged for a concealed office/digital lab and accommodation in the old town, near the Venetian port and within walking distance of the archaeological museum. Accessed via a generically inconspicuous modern doorway, a whole new apartment block had been (bought? rented? …?) acquired for their sole use. There were individual apartments for the team to stay in, while the whole of the top floor had been converted into a spacious open-plan office. This was not overlooked by any other nearby buildings and fully equipped with the high spec computing power for data processing and for Mer’s creative visual modelling.

Mer and JMZ had been officially appointed ISCRI co-Directors; practically this would split into Mer overseeing work with octopuses, visual set-up and generative work with the dev team, while JMZ took responsibility for having a broad overview and understanding the work of each research department and work group involved, making connections and convening cross-disciplinary panels to meet at appropriate intervals. She would continue to have an overall Project Management role, consulting regularly with Mer who would take the Executive oversight role. Aspasia, the first member to be appointed to the Institute’s Board, was to make introductions at all levels of the research departments and all across Greece, smoothing the way for strong connections and good communication.

Bexley’s role was Creative Producer of all the project’s strategic comms, including commissioning and overseeing documentation. The plan was to keep the research dissemination relatively restricted for the first two to three years, and to shape how the project was then exposed to the world. During those first years, Bexley and Geraint were to build the narrative and documentation, including commissioning and facilitating a full documentary film tracking the Institute’s process. This would be important for disseminating the work in a way that was consistent with the philosophy behind it.

Mer, JMZ and Aspasia together began the process of recruiting the research and technical teams, facilitated by Aspasia’s introductions. They recruited two new dev team members from the Foundation, Arkoudia and Glauca, to work with Mer, JMZ and new ISCRI developer, Willow, on the seeing objects. Arkoudia was an expert in generative AI, while Glauca specialised in vision AI. The rest of the cross-disciplinary team, each experts in their field, was assembled from various departments of ICS-FORTH and the Hellenic Centre for Marine Research, including Dimitris’ octo robotics lab. The rest of the original ISCRI team arrived in instalments, gradually moving in to the empty apartments and meeting new colleagues. They were given use of a Foundation building on the less populated south coast, from where Mer and the divers could scout for octopuses, and Hube, Symmer, Corvin and the Ghost could start setting up the RL environments in the sea within individual octopus’s territories. The house was also equipped with heavy computational power, connected to the Foundation’s own hydro-cooled eco data servers under the mountains.

Once everything was set up to begin work covertly on the island, Mer, JMZ and Bexley turned their attention to the Institute’s new official HQ in Athens, where they could host external meetings and conferences and have regular project gatherings. They took apartments in the city’s “anarchist quarter” Exarcheia. JMZ’ and Bexley’s spouses, Bataille scholar and philosopher of Dick [the science fiction writer who had reactivated and revised the atomist concept of simulacra] Dr Jean Gibson and sound engineer Protocol-F, respectively, joined them. Jean took up a research post at the university, while Protocol-F rapidly became part of the Athenian music scene.

Mer, JMZ and Bexley scheduled regular open meetings for all project participants who were interested, to maintain their ongoing philosophical thread. Evelyn and Alasdair often beamed in to these, and occasionally took a trip to the island to meet in person. The Foundation’s heads of marine ecology and zoological research, ‘Posidaeia’ and ‘Artemis’ were regular contributors. Aspasia was regularly present too. Hube kept the group updated on the latest ML research papers. The Prof and Mariella, ISCRI Director of Research and Research Associate, responsible for coordinating the Institute’s written research outputs, embargoed until it went public, gathered all the latest material on cephalopod cognition and nonverbal and animal communications to present and discuss.

Over the first six months, some unexpected results from their initial AI modelling began to become apparent. At the same time, they were introduced to more Foundation members and more of its mysterious culture. As time went on it seemed more questions were being raised than answered.



To be continued…