Diffusion: A New Venture for Antikythera
“Utopia or oblivion. It will be touch and go until the last moment” — Buckminster Fuller
Antikythera is…
a research and development institute focused on understanding the evolution of planetary intelligence and designing its scaling through computational systems.
a school of thought that recognizes the symbiotic adaptation of living and technical systems as the milieu from which human societies emerge and grow.
a means to realize a precise rigorous thesis of how the next 50 years should unfold, advanced through a research pipeline, a global consortium, and a publishing platform.
a conduit for projects that turn early weak signals into narratives that inform investment horizons. We are a center of gravity for bespoke access to ideas, people, and projects that are frustrated by existing opportunity spaces.
a pre-first mover. Before “zero to one,” we focus on “-1” and what will become tomorrow's “zero.” We synthesize insights and developments from across disciplines and domains to provide a focused map of major veins to mine for long-term leverage.
a new connective tissue between legacy research institutions. Unlike philanthropic platforms and many other non-profits, our means and goals are aligned with transformation through direct innovation. Antikythera is nimble and directed in ways universities are not, it is truly research focused in ways private platforms often cannot afford to be, it is focused on long-term transformation in ways private equity may not be, and it is structured and focused in ways seed investing seldom is.
directed by Benjamin Bratton and based on his groundbreaking work defining the contours through which computation became a global system and an indelible constituent of planetary intelligence.
How does Antikythera see the world?
Our technology is ahead of our concepts
There are moments in history when our ideas and dreams are ahead of our technical capability. There are other moments, such as the present, when our technologies have outpaced our concepts. At such times the composition of new conceptual frameworks is a necessary and practical mission. We seek and seed ideas with which we can name and understand what is right at hand.
We are in a pre-paradigmatic moment
The terms by which the world is perceived–life, technology, energy, intelligence, value, matter– are all being fundamentally transformed by computation. Antikythera seeks to clarify the parameters of the emerging paradigm by catalyzing projects that become -even indirectly- cornerstones of the societies of this century.
The Science, Philosophy, Technology Loop
New sciences are born when philosophy learns to ask the right questions. New philosophies are born when technologies force the inception of new languages. We seed and seek initiatives that close the loop, informing all three parts at once, accelerating and clarifying a new coherence between them.
From Skeuomorphism to Emergence
We are in a time of AI Skeuomorphism. Whole sectors are scrambling to apply basic affordances of frontier models to legacy processes, structures and norms. Capital pours into transitional and temporary bridges. Instead of seeking incremental optimization, we seed and seek initiatives that represent AI-first systems and structures as foundations for the next major step in societal transformation.
The Pre-Category
Fitting into a niche is how something adapts and thrives--a living creature or a company. It optimizes itself so as to optimize that fit. But evolution also tells the story of niche construction, how a species transforms its niche and in turn transforms its own development. It remakes the world and thereby remakes itself. Our focus is not on fitting the niches and categories we have but on anticipating those that are emerging and on modeling what may fill them. These are the “pre-category” and our goal is to identify them, understand them and to enworld them.
Path dependency and contingency
History is a struggle between path dependencies that become evolutionary scaffolds upon which increasingly complex systems can grow and the irreducible contingencies of choice, chance and circumstance. With the emergence of new general purpose technologies, what initially appears deterministic turns out to be arbitrary and what seems impulsive turns out to be inevitable. There is no single path out of this bind, but all of them are based in active discernment and discrimination.
Investing in Copernican Revolutions
A Copernican revolution changes not only our understanding of the world but also of ourselves. That latter shift is often more difficult but also more transformative than the former. We seek and seed initiatives that challenge our conceptions of what it means to be an intelligent living population with planet-altering agency. Ensuring a long runway for the continued evolution of intelligence requires investing in this difficult recalibration.
The School of Thought
Orientation and steerage
Antikythera seeks to orient planetary computation as a technological and philosophical force and does so by paying close attention to what planetary computation tells us in turn about who, what, when, where, how and why we are. Computation is, as Stanislaw Lem put it, an "existential technology”
The leverage of path dependencies
We are in the very early years of the ongoing evolution of planetary computation and the human societies that emerge with it. In the present moment we are setting initial conditions that will become self-reinforcing path dependencies. This is why AI-driven optimization of obsolete systems, processes and norms is so dangerous.
Costs and benefits
The costs are potentially infinite. The alternative is to focus on first principles research. It is the only way to discover the more fruitful paths before they are closed off by myopic default to the familiar. The costs of getting it wrong are astronomical, perhaps existential. The benefits of getting it right may secure futures we can hardly imagine, as life thrives on this planet and eventually others nearby.
Focus on the horizon
Antikythera is a filter for channeling capital, ideas and talent toward that goal. Our work directly and indirectly refocuses investment theses by extending the horizon of roadmaps. Our focus is on what is truly fundamental about future societies and thus immune to short-term hype and bust cycles.
R&D Units
It starts with a thesis
An R&D unit starts with a synthesis of macrotrends from seemingly disparate domains that focuses them into a clear set of interlocking provocations. The thesis is refined in research and development studios, salons, workshops, our book series and peer-reviewed journal with MIT Press: from inkling to exploration to modeling to message.
Making is Thinking is Making
We believe that ideas come from making just as much as making comes from ideas. The cycle is to externalize ideas into prototypes and in turn draw new ideas from those artifacts. Across the board, to artificialize something is to rediscover what it is. If we are invested in anything, it is in the continuing rapid rediscovery of the familiar as a newly alien form full of wonder.
Pop-Up R&D labs of underutilized creative polymaths
Over the past ten years Antikythera’s studio leadership has drawn creative overachievers from diverse disciplines and backgrounds who yearn to build but are frustrated by existing research contexts. We curate pop-up labs of the best and brightest doing what they wish they could do elsewhere. Given the chance, they come to us from and go on to places like MIT, Oxford, Google Research and Google DeepMind, Santa Fe Institute, Center for Humane Technology, Planet Labs, Thiel Fellows, Cambridge, Cal Tech, Berkeley, and many more.
A studio-based approach to software
Our research and development studios enable us to discover what others overlook. Society now asks of software what it used to ask of architecture, namely the programmatic organization of societies. We learn from architecture’s studio model in which the discipline breaks itself down to first principles and builds them back again, following ideas and clues to new places. A software society deserves a similar structured platform for discovery.
First research and then development
We divide the R&D unit into distinct phases. A Research Studio that turns pre-category starting points into new category projects. It is open-ended and willing to follow a lead to see where it goes. The second turns those insights into blueprints. Development Fellows bring projects that fit the thesis and prepare them for incubation or acceleration. We go deep on “explore” before we focus on “exploit.”
Scaffolding not solutionism
To merely optimize and overfit to the present is ultimately inefficient because this sets suboptimal initial conditions upon which further development relies. We prioritize not immediate solutions but building blocks upon which better questions are asked and answered.
Leverage before impact
Every major step in the evolution of planetary computation was inspired by a book that first mapped a new space of possibility or warned against an unforeseen dead end. This form of influence is based on leverage before impact, on inspiring and establishing the new conditions of exploration, not counting clicks. Our book series and peer-reviewed journal continues this, publishing bleeding-edge philosophy, theoretical computer science, science-fiction and more, inspiring a unique self-selecting global audience to compose what is most needed for the worlds we share.
What we promise
We do not promise to make companies, but rather to generate the core tenets on which they are founded. There is no shortage of platforms for formal incubation. There is a shortage however of truly new ideas. We promise insights with real consequences for investment horizons.
Synthetic Intelligence:
The Alignment Gap
General Purpose Intelligence and Bidirectional Alignment
Just as human societies transformed in relation to earlier general purpose technologies, they will again with the ongoing evolution of machine intelligence. Alignment of AI to human society as it exists is not enough. Alignment will be bidirectional. We seed and seek projects that will aid in the adaptation to new opportunities afforded by utility-scale AI. These include focused purpose-driven tools as well as inventive conceptual frameworks with which to orient new realities.
The Alignment Gap
The pace at which the capability ceiling of frontier models expands is accelerating beyond the pace at which society incorporates and culture internalizes those capabilities. This gap grows every release cycle and as it does, bidirectional alignment becomes more difficult.
Bending the Curve
Unsuccessful bidirectional alignment between society and general purpose intelligence will introduce and extend unnecessary harms, will suppress broadbased benefits and will destabilize the economic scaling of intelligence providers of all sizes. Antikythera works to bend the curve so that societies can close the gap by accelerating their own creative and symbiotic uses of synthetic intelligence.
Alignment Discovery
Alignment cannot arrive simply through legislation and guidelines, because alignment discovery is an active complex system. Instead of top-down policy, our focus is on how meaningful bidirectional alignment occurs one user, one task, one insight, one breakthrough at a time. The alignment interface is everywhere. It is anywhere that individuals and whole societies touch and are touched by machine sensing and cognition. Alignment is not a set of rules imposed by command; it is how a society learns to reshape itself one touchpoint at a time. Bottom-up experimentation will better steer societal adaptation than top-down mandates.
Diffusionism as Ethos
Green Lines
The discourse is saturated with “AI red lines” over which we must never cross. Some are sensible and some are coping mechanisms. It’s not always easy to know which is which. AI red lines may be necessary if well drawn but they can also suppress the realization of important benefits. Alignment overfitting is a serious risk. Instead we are focused on “AI green lines,” the necessary preconditions for assuring that the greatest range of possibilities is available to as many people as possible..
Symbiogenesis not Skynet
Evolution is driven by symbiosis and symbiogenesis as much as competition. We take as a core principle that the co-evolution of human-machine intelligence is as well. We see the future of synthetic intelligence as the mutual augmentation of different forms of mind across diverse substrates, some biological and others mineral.
Ecological Theory of Automation
Automation is misunderstood. We define it as contingency fixed into scaffolds on which other systems can depend. Automation is exemplified by an agent reliably providing significant outputs for another, like a flower’s pollen for the bee, or the city’s signals for an autonomous vehicle for the city. Automation is less a newly imposed layer of control onto natural systems than a quality of all stable systems, evolved or designed. No wonder then that AI is increasingly diffused into physical things interacting with the world and one another. Diffusion also means the radiation of AI from data centers to everyday environments.
The complexity of culture
Diffusion into human cultures implies a far more complex challenge than the engineering of AGI-level models. It will be an intricate integration of their new possibilities unlocked with diverse and deeply-embedded perspectives. Presuming that alignment must always be bi-directional between society and new technologies, Antikythera probes the phases and strategies by which a diffusion revolution will unfold.
The means of production as a monthly subscription
Widespread general purpose machine intelligence has the potential to disturb concentrated cathedrals of power and prestige by quickly making artificially scarce forms of agency more ubiquitous. Each new improvement in model capability raises the stakes but also makes the changes more difficult. We seek and seed initiatives that accelerate diffusion against incumbent interests.
“Diffusionism”
The augmentation and amplification of collective and individual intelligence will upend many cherished institutions. It will therefore be extremely contentious, especially for incumbents whose hold on cultural power is challenged. “Diffusionism” posits that the long term impact of making functional general intelligence widespread, inexpensive, accessible and useful will be, on net, extremely positive, but also bumpy as hell.
A Different Direction
Our agenda is driven by planetary scales, not by degrowth ideology. We are not like other philosophically-minded programs. We are not interested in deconstructing our way to a better world. We see the official academic-activist commentariat is a PR wing of a pre-AI cultural and economic establishment doing everything it can to discredit the tides as they wash away their sand castles. Our role is to draw what happens once they crumble. We maintain a clear vision for what is emerging but which is being sidelined by defensive groupthink.
The Model is the Message
Superintelligence is Social
All complex intelligence is social, and so will be complex hybrid superintelligence. The fast takeoff will scale better if it integrates multiple modes of cognition, capability and agency. For some, a human is the pilot. For others, a human is a collaborator. For others still, a human is a wide-eyed observer of a world alive with unfamiliar forms of mind.
Simulation as cognitive technology
Simulations are fundamental to how our minds work and to how our science works. We construct models of systems through which we interpret new experiences and new data, updating the model for the next cycle. Simulations allow us to investigate reality-in-a-box, synthesizing patterns into a conceptual whole in ways otherwise impossible. They are not just formal mathematical abstractions but conceptual models with which we heuristically navigate reality. What we today call “frontier models” and “world models” allow us to communicate directly with those synthesized patterns as a new component of that reality. The simulation is made real.
AI is a scaffold for what?
Selection brings forth the stable systems we call “life.” Life learns to artificialize its environment to scale its populations. Artificialization requires strategic intelligence. Strategic intelligence requires abstraction and symbolic language. Large language models feed on the artifacts of symbolic abstraction to model the world. What will build on AI as we know it? Will AI teach us more about what thinking is than we will teach it? What is the next step? We seek and seed projects that may help us understand those questions and address them.
Planetary Computation
The lesson of the Antikythera mechanism
The first computer was more than a calculator. It was also an astronomical device, capable of modeling the past, present and future of the sky. Computation begins here. Computation is fundamentally about orienting intelligence in relation to its planetary condition. It remains so today. It is more than an instrument; it is a technology that changes how we understand the world. We seek and seed projects that if successful will use planetary computation as the means to tell us something fundamental about our realities. The value of success is then multiplied many times over.
Artificialization is natural
Niche construction is the name for how all species artificialize their environments and thereby alter themselves. Evolution selects for those that are good at artificialization. It is adaptive. We explore artificialization both for what new capabilities it can provide but also for what new insights and truths it has to show us.
Technology literally evolves
All technologies are built from previous technologies. Each not only fulfills a function now may become part of a future technology that fulfills a function we can’t imagine. We seek and seed projects that promise to become scaffolds for future projects and the adaptive societies that incorporate them and breathe life into them.
The Discovery of Computation
Computation was discovered as much as it was invented. It is essential to how life works and replicates itself. Computation is now the dominant mode of planetary infrastructure, linking pipelines of light beneath the ocean to networks of satellites beaming signals across the surface of a planet transforming itself at unprecedented pace. We frame all questions with this totality in mind. Computation is not just about apps, chats, or clicks. It enables a major step in the evolution of planetary intelligence.
The Stack
The emergence of planetary computation produced a new kind of accidental megastructure, which we call The Stack. It both twists modern Westphalian political geography in its own image but also produces new territories over which the geopolitics of the future is contested. These are hemispherical stacks. As mapped in 2015 in the book The Stack: On Software and Sovereignty, it is composed of functionally defined layers: Earth layer, Cloud layer, City layer, Address layer, Interface layer, and User layer. However, AI upends this architecture and is fast reconstituting a new stack configuration in its image.
The timeline is 100x Planetary Compute
By our calculations the total amount of available computation at any given moment on Earth is approximately 1021 FLOPS. Call that 1 Planetary Compute unit. The total is doubling quickly. A new scaling law is observable by zooming out and looking at the whole. If current trends hold the total will be 10x Planetary Compute within a decade and 100X within two. The rate will probably be faster. Building for a world of 2x or 3x is not our goal. Instead we seek and seed projects that are based on three critical questions: What does a world with 100x more compute capacity than the present look like? What does it need? What are the initial conditions that can be set today that will orient it favorably?
Planetary Intelligence
We have always lived with superhuman computation
The prospect of living in the midst of computational intelligence that far exceeds our own does not frighten us because we realize that the world already is full of forms of computation that go far beyond what our brains are capable of doing. Intelligence, like life itself, is planetary. It is the result of accumulated interactions across vast distances and depths of time. All human species emerged in worlds surrounded by “superhuman” computational intelligence. It is called the Biosphere. This is our home.
Intelligence is something that a planet does
Evolution works through many interlocking loops, among them a macroloop by which energy is transformed into intelligence and intelligence produces technologies that capture more energy and thus drive more intelligence. Just as the imperative of life is to produce more life and to construct its niches to support this, so too is the imperative of intelligence to create more intelligence. This is not just something that happens on a planet, it is something a planet does. At least one of them.
The solar economy is the ceiling
Ultimately this whole cycle exists within a solar economy. It is driven by the nuclear fusion reactor in the sky we call the Sun. The practical limit of this cycle, if there is one, is unlikely to be contained by a single planet. There are limits to how much entropy one astronomic body can sustain and still transform into systems that beget more complex systems. So, outward we go.
Extraplanetary computation
As we look over the horizon, where all the roadmaps end, we envision planetary computation evolving into extraplanetary computation. We foresee the Earth itself building more technologies that extend its involvement with its native solar economy in new ways. Working back from that, the contours of the incipient AI Stack for planetary computation comes into view. It is less an extrapolation of the stack we have than the next stage of the future state of the energy/ intelligence/technology loop that has been running for billions of years.
Join us
We seed and seek projects that already understand themselves in this way, that have a critical role to play in getting us there, that require a space to develop they do not already have, and which are driven by those whose sense of urgency compels them as much as ours compels us.