Keynote. A Cooperative AI?
Updated: Oct 7
On November 12, 2025, I delivered a keynote talk entitled "Can AI Become a Labour Commons? Preventing Dystopia by Cooperativizing the Code" at the Platform Cooperativism Consortium's "Cooperative AI" conference in Istanbul, Turkiye. To read my talk and view the accompanying PowerPoint presentation, see below. The conference was a diverse mix of cooperative practitioners, high-tech workers, AI specialists, and university and independent researchers, and wonderfully educational and collaborative. It was brilliantly co-hosted by the PCC's Trebor Scholz and Stefano Tortorici, the Turkiye-based NeedsMap, and held at the Istanbul Planning Agency. Many thanks to Trebor and Stefano for the invite, and to the wonderful hosting by members of NeedsMap and the staff at the Istanbul Planning Agency. It was a truly inspiring and eye-opening experience!
Can AI Become a Labour Commons?
Preventing Dystopia by Cooperativizing the Code
Keynote written and delivered by Marcelo Vieta (www.vieta.ca)
Delivered November 12, 2025 in Istanbul, Turkiye; revised for publication September 2026, with the live-audience framing kept.
For the Cooperative AI Conference of the Platform Cooperativism Consortium (https://platform.coop/events/cooperativeai/)
A brief video summarizing my keynote and my views on cooperative AI development and governance
Part 1: The Questions and the Stakes
Good afternoon.
Today, I want to address three questions that I think are crucial for ensuring Cooperative AI:
Can we democratize AI?
Can artificial intelligence become a commons?
Can we prevent the dystopian futures that seem to be looming by cooperativizing the code?
The answers to these questions, I want to argue, are to be found at the intersection of technological transformation, democratic imagination, and the future of work.
My talk today is not technical but ethico-political, philosophical, and action-oriented. It is a call to pause, to reflect, and to reimagine the human–technology relationship before AI slips further from our grasp.
I want to offer a series of observations drawn from my research interests and work with labour and cooperative movements, the social and solidarity economy, critical theory, communication studies, and the philosophy of technology.
In my talk, I’ll move from a philosophical framing to tangible practices already emerging in cooperative movements, data and platform co-ops, and local governance.
The questions and observations I share with you today, I propose, are crucial for anyone thinking about the ethics and the future of AI.
Whether you are a technologist, a policymaker, a community organizer, a digital worker, or a tech user, I invite us to consider how we might transform the code and infrastructure of AI into something more democratic, more ethical, and more humane – into what I, with others, have come to call a labour commons.
In contrast to the “break things and apologize later” method of the for-profit tech sector, I propose that, with AI, we should take a breath and re-think it. There are still possibilities for hope, though these possibilities may be closing quickly.
Why do I say that time is short?
For one, because lost in the AI-hype of speed, scale, and the drive for “recursively self-improving “ AI models and, ultimately, the superintelligence proffered by the so-called “Big-Seven” tech giants and their evangelists are the massive impacts and unintended consequences of the AI push. Let me name just four:
The environmental impacts. Training and deploying large models require vast “compute” infrastructure, water, and energy. Data centres are today’s new factories … except that the exhaust is global and the human labour that is still needed to build them and run the large language models are dispersed (much of it to a deeply exploited army of workers in mining and data annotation in the global South) and thus invisibilized. Annual energy footprints of hyperscalers now rival those of entire cities and countries. Google’s alone in 2024 paralleled the energy use of nearly 3 million households. By the end of 2025, Google used 37% more electricity than the year before, while its data centers used about 42 million megawatt-hours, roughly New Zealand’s entire annual electricity consumption, or close to the energy use of 4 million US households.These burdens concentrate in specific regions and grids, externalizing costs to vulnerable communities and ecosystems.
Infringements on creative and intellectual labour. Much of today’s AI was trained on scraped cultural labour – writing, art, scholarship, journalism – almost all without consent, compensation, or attribution. This business model optimizes shareholder value rather than collective cultural memory. The Writers Guild of America’s 2023 strike was exactly about this.
Lost jobs. Automation reshapes work across sectors. “Routine cognitive” work in the global North is particularly vulnerable. At the same time, as we saw in Antonio Casilli’s brilliant film, the job of poorly paid annotators – so necessary still for LLMs to function – are relegated to the global South. Of little consideration to the big AI tech companies is securing good jobs and ensuring the meaning work gives us.
Fragmented political discourse. Synthetic media at scale – deepfakes, hyper-targeted manipulation – erodes already fragile trust in democracy and public life.
I could go on. Most of us in this room know these ever present dark sides of AI.
AI has thus brought us to a planetary inflection point. Why do I say this? Because this is the first time in humanity’s existence that technology is running ahead of human oversight and planetary limits at such mass scale and speed.
But, in today’s talk, I don’t want to paint a dystopia. I want to advocate for ways of preventing it by creatively and democratically appropriating AI. What I want to argue is that there is still time; there is still space for hope ... if, as all of you are so inspiringly doing, we reorient our imagination and our organizing.
What I want to highlight in my talk today is what is lost in the mainstream AI-hype: That there already are a growing number of us around the world thinking AI differently. Most of us here in this room … all of you … are at the cutting edge of re-envisioning a more just AI future.
First, a caveat: While the final phrase in the title of my talk is “Cooperativizing the Code,” I want to acknowledge that AI is much more than software and code. It is a complex socio-technical system that fuses coders and code, poorly paid workers and highly paid tech specialists, individual data and oligopolies, algorithmic infrastructure and AI users, cloud platforms and shareholders, bits and bytes and the world’s precious natural resources, learning models and human governance ecosystems.
The last point is ultimately about the possible human governance and oversight of AI. This, I want to argue today, is vital for what many of us at this conference are advocating for. People – workers and communities, not artificial agents and algorithms – are the real DNA of the AI code. And so, because this is mostly forgotten in the AI debates, the time to rethink AI is now … before we, people and communities, are programmed out further.
Part 2: The Techné of the Ancient Greeks, the Technology of Modernity, and the AI of Tomorrow
To begin to think about a Cooperative AI, I want to first step back to the philosophical and etymological roots of the word technology itself. If you will indulge me, I’d like to begin with revisiting the ancient Greeks, who still offer us insights for understanding the cultural and ecological limits being breached by an unfettered AI. They also offer us continuing echoes as to what we can recuperate by approaching AI from a different ethical disposition.
In Socrates, Plato, and Aristotle, for instance, we encounter five key concepts still useful for understanding what is at stake in the human–technology relation: physis, poiesis, techné, logos, and phronesis.
For the Greeks, these concepts were central to the intentionality of human making and the use of tools. The ancient Greeks help us still see that human making occurs with and within nature, involving a holistic worldview.
First for them there was physis, or nature – “that which creates itself…which emerges from out of itself.” Physis is self-generating, driven by what it is meant to be. For example, an acorn carries within it the potential to become an oak. The oak is not yet there, but its becoming is.
Poiesis, by contrast, refers to “the practical activity of making,” engaged with by humans when producing something. In turn, techné is the art and craft of making, embracing the skill, knowledge, and discipline of doing poiesis well. And logos was, for the Greeks, what gives order and gathers.
For the Greeks, then, techné was not just about utility; it was craft, the right way to create, to help bring things into existence closer to their essential and correct forms. But Aristotle adds something important here. Knowing how to make something well is not the same as knowing whether it should be made, or what it is for. For that, he says, we need another kind of knowledge: phronesis, or practical wisdom – the ability to judge what is good for human life. Techné gives us the skill; phronesis gives us the judgment. At its best, Greek making held the two together. Thus, craft answered to wisdom. It was a way of participating in the unfolding of the world, not dominating it.
These ideas profoundly influenced 20th century critical thinkers like Herbert Marcuse, sage of the New Left in the 1960s and a key member of the Frankfurt School. Marcuse saw in techné the possibility of a post-technological rationality that could reconcile art and tools, ethics and innovation. Marcuse contrasted this with the modern technological mindset – technological rationality – which tends to treat nature and human labour as mere instruments … as resources to be extracted, optimized, and controlled.
This shift from techné to technological rationality, from pre-modern to modern ways of thinking, marked a profound transformation in how humans understood and engaged with the world. This shift was central to the birth of the scientific method and capitalism. Technological rationality is not just a way of thinking … it’s a worldview. It’s a lens through which everything is seen as discernable, dissectible, utilitarian, and exchangeable. Everything is now a tool or a resource, a means to an end.
As Andrew Feenberg, Marcuse’s PhD student and one of my MA supervisors, puts it: things are now “simply there, unresistingly available for human use.” Technological rationality thus shapes our modern economies. It tends to organize our entire experience of reality. It tends to tell us what is worth knowing, what is worth doing, and what is worth valuing.
Marcuse and Feenberg show how this rationality narrows what counts as knowledge. Postcolonial thinker Gayatri Spivak has a name for what happens when one way of knowing is imposed as the only legitimate one: epistemic violence. Ethical, aesthetic, spiritual, and non-Western ways of knowing are simply dismissed as irrational or irrelevant. Today, in our Western-centric and technologically rationalist ways of thinking, it’s not just that we use technology to manipulate the world; it's that we stop asking whether we should.
But here Feenberg adds a crucial twist, and it’s where he parts ways with his teacher, Marcuse. Technological rationality is powerful, but it is not destiny. Technology is neither neutral nor an unstoppable force. Every technology carries values built into its design. These values are choices about what counts, whom it serves, and what it leaves out. And what was designed one way can be designed otherly. History shows this again and again: in the 1980s, AIDS activists forced medical research to open experimental treatments to patients. Workers and users have repeatedly turned technologies toward purposes their designers never imagined. As we’ll see, workers at the UK’s Lucas Aerospace and Italy’s GKN have done exactly this. Technology, in other words, is a site of struggle … and of choice.
Indeed, now perhaps more than ever, with AI we are at the dawn of a new historical rupture. With AI, we may be reaching technological rationality’s escape velocity – tools designed to augment us begin to operate at scales and speeds that disembed them from human judgment, obscure accountability, and intensify externalities. Can we still recover the ethical core of techné – craft, proportion, and care – in the age of AI? Feenberg’s answer, I think, is yes. But not by going back to the Greeks. We can’t return to their world. What we can do is carry their questions into our designs: into how AI is built, how it is governed, and who gets a say.
With unfettered AI, Mary Shelley’s prophetic novel, Frankenstein, or the Modern Prometheus, comes to mind. It’s often read as a warning against creating monsters. But look closer: Victor Frankenstein’s real failure is not that he creates. It’s that he abandons what he has made. He refuses responsibility for it. Frankenstein’s creature is not born a monster; it becomes one through neglect. That is the question AI puts to us: not whether we can create, but whether we will remain responsible for what we create.
What kind of rationality is guiding our AI systems? What values are embedded in their design? Who benefits? Who bears the cost? Can we bring AI back to human scale and reorient our algorithmic imagination toward care and co-responsibility with each other and the planet? Or will we continue down the path of severing our technologies from the judgment and care that should guide them, building ever more powerful tools while forgetting to ask what, and whom, they are for?
Part 3: Relationality, Indigenous Ways of Knowing, and Ethical Technology
As the philosopher Yuk Hui reminds us, there has never been just one kind of technology. Different cultures have joined technology, ethics, and the cosmos in different ways. Technological rationality is one path among many, not the only one.
To imagine a different path for AI – one that resists technological rationality – we must look beyond the dominant paradigms of Eurocentric philosophy. Indigenous ways of knowing offer us invaluable guidance.
Indigenous economics and ways of knowing are not about markets or transactions. They are holistic worldviews, integrating at all times relationships with land, community, ancestors, and future generations. Past, present, and future are considered together in human action. Wealth, for many Indigenous peoples, is not measured by accumulation, but by the health of the pluriverse of these relationships.
Indeed, “technology” as a separate, abstract category is itself a recent invention of industrial modernity; most Indigenous knowledge systems have not needed such a concept because tools have been understood within relationships. Indigenous traditions tend to hold that tools and artifacts are never separate from community and the other-than-human. They are interconnected. From traditional birchbark canoes and long-held ways of navigation to digital storytelling and knowledge-keeping today, tools, artifacts, and innovations are expressions of cultural and ecological continuity and care
Today, many Indigenous communities are engaging with digital tools to preserve language, resist data colonialism, assert data sovereignty, and tell their own stories … in their own terms.
My colleague at the University of Toronto, Jennifer Wemigwans, has been working with Indigenous communities and Elders to collect digital bundles that hold cultural memory, sacred knowledge, and teachings via digital platforms.
University of Victoria’s Vanessa Andreotti’s AI bot Aiden is an AI-driven conversational bot created as part of their research in the Gesturing Towards Decolonial Futures (GTDF) collective. The tech combines natural language processing with curated content from GTDF’s research corpus in collaboration with Indigenous communities. Aiden avoids extractive or prescriptive interactions and rather encourages users to question assumptions about progress and development, emphasizing relational ethics and plurality of perspectives. Andreotti is currently looking to house the server with a local community involved in further developing Aiden.
Resonating with Indigenous ways of knowing is the work of Ursula Franklin, the Canadian and University of Toronto physicist and philosopher whose 1989 Massey Lectures, published as The Real World of Technology, remain a touchstone for ethical tech thinking.
Franklin argued for “technology as practice.” Technology is not just a collection of tools. It is a way of doing things. Every technological system embodies values, assumptions, and power structures.
Franklin further distinguished between holistic and prescriptive technologies. Holistic technologies allow people to control the entire work process, like artisanal or craft work. Prescriptive technologies fragment the work process; they are control-oriented. They break tasks into discrete steps, controlled by systems or authorities, like assembly lines or bureaucracies. They diminish human agency and foster conformity. Franklin’s holistic technologies are, in a sense, the modern heirs of techné; her prescriptive technologies are technological rationality at work on the shop floor, in the office, and in the marketplace.
One compelling example illustrating Franklin’s vision in practice was the Lucas Plan of 1976. Faced with the threat of thousands of layoffs at Lucas Aerospace (a major British manufacturer of military and aerospace components), the company’s shop stewards, organized across its many plants in the Lucas Aerospace Combine Shop Stewards Committee, asked their fellow workers a simple question: “Given your skills, and the machines you already work with, what could we make that society actually needs?” The result was an Alternative Corporate Plan proposing some 150 socially useful products, among them heat pumps, wind turbines, kidney dialysis machines, and a road–rail bus. The workers managed to build several of these ideas as working prototypes. Management, to no surprise, rejected the Plan outright. But the Lucas workers had proven something profound: that the people who do the work can redesign technology around human and social needs. What stood in their way was not technical at all. It was the question of who gets to decide. In Feenberg’s terms, the Lucas workers showed that what was designed one way could be designed in another way.
Where the Lucas workers were stopped, others are going further. Ex-GKN in Campi Bisenzio, Florence, is a former car-parts factory whose 422 workers were dismissed by email in July 2021. The workers responded by occupying the plant in what has become the longest factory occupation in Italian history. Universities, engineers, and students have since partnered with the ex-GKN workers and redeployed their expertise to begin transitioning the plant into a solar panel and cargo bike manufacturer. In doing so, they joined the fight to establish an ecological transition for Italian industry and inspired the Insorgiamo (“Let's rise up!”) movement, a name that recalls the slogan of the partisans who liberated Florence in 1944.
Other European examples of workers and communities recuperating former capitalist firms into worker and community co-ops, as well as transitioning their production into more just and ecological social production, have been: Milan’s RiMaflow (another former car parts manufacturer turned worker cooperative and social economy hub), Rome’s Officine Zero (a former rail maintenance facility that became a multi-stakeholder community cooperative), and, near Marseille, Scop-Ti (a worker cooperative founded in the former Fralib plant, where Unilever produced Lipton and Éléphant teas) and L’Après M (a former McDonald’s franchise in Marseille’s northern districts that its workers and neighbours turned into a community-run restaurant and solidarity food hub).
The Lucas Plan, ex-GKN, and many worker-recuperated firms from around the world exemplify what Franklin called redemptive technology: technology as a practice rooted in justice, care, and democratic participation.
These techno-ethical insights and practices offer the Cooperative AI transition a compass.
Part 4: AI as a Moment of Inflection in Humanity’s Technological History
With AI we are at a turning point in the history of the human–technology relation. We are now building technologies that not only extend human capacities exponentially, they begin to operate independently of us. We are training AI systems to learn and make decisions without direct human oversight or guidance. What future is this shaping?
Philosophers and historians of technology like Lewis Mumford, David Noble, Ursula Franklin, and Andrew Feenberg have reminded us that technology is never neutral. It, rather, always reflects the values and power structures of its creators. What values and structures of power, then, undergird AI if the goal is to artificially replicate, enhance, and ultimately replace human capacities and human thinking?
If, on the other hand, AI development and deployment were to be taken up within critical and ethical frameworks – with relationality, with care, with democratic design, with the inclusion of community voices … perhaps with a narrow AI (AI designed for specific, bounded tasks, rather than the pursuit of general “superintelligence”) – could we avoid or at least mitigate the dilemmas that loom?
I want to reach out to one more critical thinker of technology: Marshall McLuhan, the Canadian and University of Toronto media theorist who gave us the concept of the Global Village, the Medium is the Message, and who, in many ways, predicted the Internet.
Towards the end of his life, McLuhan developed a framework called the Laws of Media, published as a book with the same title in 1988 after his death and co-written with his son Eric McLuhan. Also known as the tetrad, it’s a diagnostic tool for understanding how any new technology affects society.
The tetrad asks four questions:
What does a technology enhance or amplify?
What does it make obsolete?
What does it retrieve? That is, what older practices, technologies, or values does it bring back?
What does it reverse into when pushed to its limits?
In October 2025, I asked the version of Microsoft’s Copilot licensed by the University of Toronto to run a tetradic analysis of AI. Here is what it came up with:
Enhances: It amplifies decision-making speed, pattern recognition, predictive analytics, and personalization. It automates cognitive and creative tasks.
Obsolesces: It obsolesces routine human labour, replaces manual data analysis, disrupts traditional learning and judgment.
Retrieves: It brings back predictions and insights, and even oracular knowledge (think ancient seers, prophets, and sages). It revives apprenticeship models through personalized learning and retribalizes media consumption via recommendation engines.
Reverses: When pushed to its limits, AI flips into surveillance, bias, misinformation, and the dehumanization of decision-making. It risks creating dependence on machines for thinking, creativity, and governance.
A pretty good tetrad! But notice one thing in particular: Copilot says AI retrieves oracular knowledge. That’s worth reflecting on further. Remember that for the Greeks, what made techné genuine knowledge was logos – the ability to give an account of why something works. An oracle gives no such account. It pronounces; we obey, or, at best, we interpret. Many of today’s AI systems work the same way: they give us answers that even their makers cannot fully explain. In that sense, AI risks retrieving not techné, but the oracle.
But we can go further; allow me to add a few more tetradic insights of my own:
AI enhances or amplifies: simulation and simulacra (pace Baudrillard!).
It obsolesces: jobs and professions, human ways of learning, human intuition, human creativity, “sweat equity,” learning-by-doing, and human wisdom – what Aristotle called phronesis, the practical judgment about what is good for human life, which can’t be automated.
It retrieves: religion and myth, and, more darkly, forms of serfdom and even slavery in the hidden, poorly paid labour of data annotators, the forced and child labour found in some of the mines that supply AI’s hardware, and a growing dependence on the platform owners who control the tools, the data, and the terms of work. This all echoes the old dependence of serfs on the lords who owned the land.
It reverses into: extreme exploitation of labour, hallucination, cheating, human ignorance, human illiteracy, zombiism, and, perhaps, when the superintelligence ostensibly takes over, the singularity – that moment when AI surpasses human intelligence and becomes an autonomous force, perhaps even our overlord.
McLuhan’s tetrad helps us see the full arc of technological impact. Not just what a technology does, but what it undoes, what it resurrects, and what it risks turning into. The tetrad helps us understand that every innovation carries unintended consequences. And, it challenges us to think holistically and ethically about this new technology – AI. Read alongside the Greeks, the tetrad shows us exactly what is at stake: AI amplifies technique while threatening the judgment that should guide it. Without this judgement, AI is a techné without phronesis (human wisdom).
Part 5: Cooperativizing AI and the Labour Commons
If McLuhan’s tetrad helps us diagnose the impact of AI, then cooperatives and the commons offer us remedies for its pitfalls; a way to reimagine and reclaim it.
Cooperatives all have at their core a deeper concept that has guided much of my own research work and activism: autogestión, or workers’ and community self-management. This is not just about running a business differently. It’s about reclaiming control over the means of production and the technologies and conditions of work, by working people themselves.
In Argentina, Italy, Canada, and elsewhere, I’ve studied and worked with worker-recuperated enterprises. These are companies that were abandoned or shuttered usually during economic crises and then taken over and run democratically by their workers.
These workers don’t just save their jobs. They transform their workplaces into spaces of solidarity, care, and community engagement. In Argentina, the workers themselves call the collective and cooperative practices they take on in the recuperated workplace autogestión.
It’s a word that comes from the Greek autos (meaning “self”) and the Latin gestio (meaning “to manage” or “to bear”). More evocatively, it can be understood as self-gestation – the act of collectively creating, controlling, and sustaining a collective’s own productive and creative life.
Autogestión is not just a management model. It’s a political and ethical stance. In this sense, autogestión recuperates something of what the Greeks meant by physis and poiesis: a making that emerges from within a community, rather than being imposed on it from outside.
This brings us to the idea of the labour commons, a concept I’ve developed with Dario Azzellini in our just-published book Commoning Labour and Democracy at Work: When Workers Take Over.
The labour commons reframes work, not as a commodity to be bought and sold, but as a shared community resource – a commons – collectively governed by working people themselves, and socially embedded.
In a labour commons, the workplace becomes an organizational commons. The labour performed becomes a commoning practice. And the surplus generated becomes a commonwealth, where the created value and surplus are reinvested in the community, not extracted for profit. This model challenges the capitalist norms that commodify labour and alienate workers. The logic shifts from extraction to reproduction of life.
Can AI be woven into a labour commons? Yes! When data, algorithm models, and platforms are treated as common pool resources, collectively governed and locally accountable. And when people are not exploited in the process.
The Drivers Cooperative, which is one of the US’s largest worker cooperatives, is an example of a digital labour commons. It uses algorithmic dispatch and route optimization while aligning control and value with drivers, who are co-owners and members of the co-op. The tech is similar to Lyft and Uber; the ownership and governance are not.
Consider also data cooperatives, which offer a democratic alternative to what Shoshana Zuboff has termed our era of “surveillance capitalism.” A data cooperative is a member-owned and governed organization where individuals voluntarily pool their personal data. The goal is to collectively manage, protect, and benefit from that data, and retain data control, rather than leaving it all in the hands of large corporations.
One example is Megha Mandli, a cooperative of some 1,000 Adivasi (Indigenous) women farmers in Gujarat’s Tapi district and part of India’s SEWA Cooperative Federation. It coordinates agricultural tools, inputs, and services, and, with the research organization IT for Change, is piloting a data cooperative in which the women themselves own and govern their data. Data becomes collective, and a means of livelihood and self-determination, not extraction and patriarchal exploitation.
We reviewed similar types of digital co-ops in our cooperatively written 2023 book, Cooperatives at Work.
These platform and data co-op models align with the values and practices of autogestión and the labour commons, offering suggestive ways that a narrow AI could be governed within democratic, community-driven structures. These examples show us that it is possible to build complex AI systems that are ethical and community based.
Digital network and movement building are also vital for the solidarity stack that Trebor Scholz talked about yesterday. In Argentina, FACTTIC – the federation of tech and knowledge worker co-ops – builds scale through shared projects, mutual support, and sectoral advocacy. It is innovating capacity building and social protections for tech workers that are collectively negotiated.
Some municipalities and countries around the world are also experimenting with democratic AI:
Barcelona’s Decidim offers open source platforms that extend participatory decision‑making to citizens.
Helsinki’s AI Register and the Dutch government’s Algorithm Register, actively used in cities such as Amsterdam, innovate transparent listings of municipal AI uses and public oversight of algorithmic governance.
The City of Montréal and MILA are creating responsible AI frameworks with local communities.
And Switerland’s Apertus LLM AI initiative (an open LLM released in 2025 by EPFL, ETH Zurich, and CSCS), as well as the cooperatively run Public AI Switzerland, all embed transparency and citizen participation into digital governance.
Moreover, the Digital Public Goods Alliance is a promising global initiative that contributes to the United Nations’ Sustainable Development Goals.
These are not utopian pipe dreams; they are living prototypes that show how to embed into AI participatory governance and public‑interest.
Andrew Feenberg has a name for this: democratic rationalization. It’s what happens when the people who build, use, and live with a technology – workers, users, citizens – intervene in its design and governance, and rewrite the values built into it. The drivers who own their platform, the SEWA women who control their data, and the cities that open their algorithms to public scrutiny are doing exactly this. They show that what I have been arguing today is not just philosophical, it is about concretely redesigning technology in other, more purposefully humane, more ecological sound, and more democratic ways.
Part 6: Is it too late to Recuperate Techné and Cooperativize the Code?
But are these experiments in cooperative AI too late? Some, like AI safety expert Roman Yampolskiy and the Nobel laureate “godfather” of AI, Geoffrey Hinton, warn that advanced AI may soon outrun human control altogether. “Deep learning” (neural networks with many layers trained on data); the drive toward “recursively self-improving systems”; hyperscaled AI’s inherent platform power, planetary spread, opacity, and sunk costs … all underscore that AI may have already outrun human and democratic intervention.
But I am not 100% convinced that we are there yet. Perhaps I am still too hopeful, but hope and collective action might be what we have left to leverage for change … for a new AI.
I am still convinced that, while time is running out, we can still intervene. We can still rein in and rethink AI. We built these systems. And, unlike Victor Frankenstein, we do not have to abandon what we have made. We can redirect it. Or slow it down and re-design it. Perhaps the goal is not to seize every existing tech stack, but to start anew where necessary, as some of us have been suggesting at this conference: recoding architectures, recuperating human-centred data practices, and reconceptualizing incentive structures around human scale and ecological limits.
I turn to you, the experts, for some of the answers and possibilities. Can we still do this? When there are clear, transparent and participatory processes, can AI not be harnessed for deepening democratic participation and equity? Can’t we recuperate a digital commons? AI as a labour commons? Can AI be a redemptive technology – one that codes ethics into its very infrastructures, that redistributes power equitably?
The brief examples I’ve shared today, and the many initiatives that you are all sharing with us, are not isolated cases: they are happening all over the world. They are the budding seeds and flowers of a new and collaborative digital economy. Cooperative AI!
Let me close by returning to where I began: with the Greeks. Like the acorn that carries the oak within it, these seeds carry a different future … not yet here, but already becoming. This, finally, is what recovering techné means today. Not returning to the ancient world, but joining skill to judgment – techné to phronesis – in how AI is made, and in who gets to make it. Autogestión, the labour commons, and cooperative AI are ways of doing just that: of making our technologies with and within our communities and the living world, rather than independently from them.
Our future generations deserve that we grapple with both the perils and the possibilities of AI now. You are all showing the way for how to cooperativize not only the code, but the conditions under which code is made. That, in the end, is where a truly cooperative AI begins.
Thank you. I look forward to the conversation.
