Interspecies Communication Research Initiative—ISCRI—exists as a proposal: to bring an octopus into a temporary and reciprocal learning relationship with a machine learning agent. The goal, to attempt a learning scenario in which the human is withdrawn from the loop, and to learn something about what other, higher-dimensional forms of communication could emerge in its absence. It is a collaboration between artist collective 0rphan Drift (0D) and machine learning consultancy Etic Lab, incubated for an extended period by the Creative AI Lab at Serpentine and King’s College London.
ISCRI emerged at a time of converging urgencies that seemed to compound. The deployment of ubiquitous semi-autonomous systems, in supermarkets, hospitals and battlefields; of the potential value of nonhuman ability being diminished through ecosystemic degradation; of complex systems that could collapse unpredictably, vastly more difficult to restore than prevent. These urgencies exasperated each other: accelerated energy expenditure impacting environments, autonomous systems making erroneous decisions with escalatory existential consequences. But some speculated that a different kind of AI system could be developed through reciprocal learning—if we were open enough to being stewarded, ourselves, by modes of understanding that disrupted conventional wisdom. Though never viewed as a solution to these multiple crises, such divergent strands of research might act as a proof of concept for counterhegemonic AI paradigms that evaded the logic of the market and instead sought different grounds for optimisation.
The contribution of ISCRI lies not only in how data was to be gathered, handled, and conceived of. It proposes early on what Stephanie Moran of Etic Lab emphasises as a ‘whole-systems approach’ in which the entire system was the artwork, the manifestation of a holistic, collectively constituted research apparatus. This forwarded a new collaborative configuration in which a first-order collaboration—cephalopod-agent—was scaffolded by a second-order collaboration—a system of humans attempting to facilitate the core relationship with minimal intervention. In this, something is reflected of ‘collaboration’ for art practice in particular but for collective work more generally. The delegation of labour is never simple; it is complex, multiagent, inequitable across multiple dimensions not just of quantity of work done, but of kind of work. This insight has transformed my own research profoundly, and holds for many other edge cases in which a collaboration is complex.
Tapping into this potentially transformative possibility, whilst wary of the assumption that it would certainly work, ISCRI is recognised as a high-risk, high-reward artistic research project. Though it attracted widespread attention in the artworld, this potential for failure made it challenging for traditional institutions and funding bodies to invest in. How a project could both test a new hypothesis and guarantee the deliverables required by organisations became a challenge to the arts ecosystem that in the end was never adopted by a specific organisation in full. The project has brought to the fore a set of memetic concepts that have become increasingly important in the artworld in recent years: interspecies communication; multimodal data; high-dimensional pattern recognition; new collaborative configurations; embedded intelligence. Interspecies communication, as a theme, became of widespread interest to both artists and scientists working concurrently or later as a unique affordance of machine learning’s specific capabilities. Sometimes this has been characterised as ‘translation’, often as ‘interpretation’; ISCRI defined and still holds a unique position in the field in its agnosticism, and desire to simply establish contact, to find a signal that could tempt a response, not to name this relationship before it emerged. High-dimensional pattern recognition was relatively new when the ISCRI system architecture was designed: within its undulating, ineffable enfoldment might lie a recurrent signal, otherwise invisible to the human, which could be decrypted only by an agent trained through algebraic calculation rather than an analogue sensorium. This form of ‘multimodal’ data analysis levels a challenge at the conventional scientific method, in which a single variable is permitted and its change interpreted and theorised. By observing multiple variables simultaneously, new patterns could be observed and new kinds of inference made.
This was grounded by Maggie Roberts and Ranu Mukherjee, co-channelers of 0D, specifically in the shared affordance of machine learning and artistic research: to conduct very different practices which could both be called pattern recognition, conducive to new forms of knowledge. Like the unpredictable patterning of the cephalopod’s skin; it’s morphological form; its chemotactile signals extending into its immediate environs, a machine learning agent could triangulate multiple modes of data, finding patterns across these in what Etic Lab’s Alex Hogan called ‘the rate of change of rate of change’, imperceptible to the human, but the core domain of the appropriate neural architecture. The artist, likewise, works with patterns which operate beyond the confines of natural language to convey, interpret, and formulate knowledge claims outside of linguistic reduction. Finally, ISCRI established a concrete methodology for investigating embedded cognition. They were willing to rethink the octopus as a fixed 'subject', instead conceiving it as an embedded intelligence, a sensorium closely integrated with its environment but which nonetheless retained its own bounded moral status. This conception in itself can teach us something about machine learning systems, always embedded in their deployment contexts, both in its input and output, as well as their recursive logic which complicates any linear understanding of their function.
Despite the absence of a build phase, the ideas went through many iterations of feedback and have become profoundly important as a hypothetical thought experiment (or ‘hypotype’). They were rigorously and imaginatively tested against disciplinary standards across marine biology and machine learning, philosophy and ecological psychology. The proposal was iterated upon countless times; scrutinised and revised, fortified by many minds who helped to shape this strange and expanding scheme, an idea pulled from the void of speculative possibility; to make it conform, in some minimally viable way, with our current reality and expectations for rigorous inquiry. By taking soundings from experts in these diverse fields, criticisms of the proposal were absorbed and changed the schemata. They were, in the end, not tested against the constraints of the world itself.
There are specific challenges with pushing through a project like this in ~2020-23. That this type of interdisciplinary research design is not yet fully recognised might be the reason it lacked a funding stream. But it is because of this that we must rethink how we bring speculative projects, on the edge of viability, into being. The project underlines the need for both technical and philosophical expertise in curatorial contexts, capable of identifying the point where artistic speculation aligns with contemporary technological possibility, without compromise to ideas which might expand contemporaneous assumptions. We need institutions which can bear greater risk exposure, with the competency to assess highly speculative and resource intensive technological projects.
What comes next for these ideas? For 0D, they continue to imagine an encounter between an octopus and an AI in a multidimensional artwork in development called 9Brains. For Etic Lab, it continues to inform the possibility space for ML research and value creation today. For me, the ‘whole-systems approach’ and the schematics for this hypothetical collaboration has deeply informed my doctoral research. For now, though, the ISCRI hypotype becomes an early research phase for other future projects which will come to completion, in which the ‘whole-systems approach will be further realised’. Perhaps someday it will be discovered on the Wayback Machine and pulled out of the cryogenic chamber. But its ideas have been seeded, proliferating through the cultural undercarriage that is our dimension of the Artworld.
Alasdair Milne
November 2023