I began my scientific career at Starlab, a multidisciplinary research institute secluded amid the tranquil forests beyond Brussels. Its principal base occupied an imposing nineteenth-century manor which, in an earlier incarnation, had served as the official embassy of the First Czechoslovak Republic.
Established by Walter de Brouwer and MIT Media Lab founder Nicholas Negroponte, Starlab was conceived as a scientific “Noah’s Ark”: a refuge in which minds from many disciplines might come together to pursue far-ranging projects. Its nearest neighbor offered an almost perfect counterpoint: the Pasteur Institute, one of only a handful of heavily secured Biosafety Level 4 laboratories around the world. One was designed to admit the widest possible range of ideas; the other to ensure that certain contents remained securely within.
Starlab was born as an incubator for fundamental, long-horizon research in the tradition of Bell Labs, the MIT Media Lab, Xerox PARC, and Interval Research. Its watchwords were “Deep Future” and “A place where one hundred years means nothing.” At its height, more than 130 scientists from thirty-six countries lived and worked at the laboratory, many already distinguished in their respective fields.
A second base of operations, Starlab DF-II (Deep Future II), was established at the Fabra Observatory, perched high above Barcelona. With its mission narrowed to space and neuroscience research, DF-II survived the closure of its parent organization and continues its work to the present day.
Research ranged across artificial intelligence, biophysics, consciousness, emotics, intelligent clothing, materials science, protein folding, neuroscience, new media, nanoelectronics, quantum computation, macroscopic entanglement, robotics, stem-cell research, theoretical physics—including the possibility of time travel—transarchitecture, and wearable computing. It was a place founded on the belief that the most consequential discoveries often arise where customary boundaries have ceased to matter.
Our purpose-built supercomputer, the CAM-Brain Machine, was supported in part by a €1 million grant from the European Union. Estimated at the time to rival the power of 10,000 Pentium II PCs, it harnessed Xilinx field-programmable gate arrays and evolutionary genetic algorithms to evolve neural circuit modules in hardware. The completed architecture could contain as many as seventy-five million artificial neurons and update 130 billion cellular-automata cells per second—sufficient, we hoped, for the real-time control of embodied machines. The 2001 Guinness Book of World Records recognized it as the “World’s Most Complex Artificial Brain.”
In testimony before the French Sénat, and later in an advisory report prepared as chairman of the UNISCA First Committee on Disarmament and International Security, I argued: “Future networks will not be built. They will be grown.”
When Starlab’s finances faltered in June 2001, chance afforded me an unexpected avenue of appeal: a meeting with President George W. Bush during his first overseas journey after taking office. Newly arrived in Brussels for consultations with NATO, the President surprised us with his immediate familiarity with Starlab and its work, responding favorably to our earnest request for $1 million in federal support through the National Nanotechnology Initiative—announced by President Clinton the previous year and carried forward through the 2001 federal research program. The funds never landed. Neither, as it turned out, did Starlab—Nature reported the collapse that July under the headline “Utopian dream in tatters as Starlab crashes to Earth.” Appropriations of that magnitude move at the pace of the behemoths that make them, plodding and sluggish to adapt; a laboratory for which one hundred years was to mean nothing could not outlast the month.
For my contributions to the field, I was selected as one of three student fellows at the Salishan Conference on High-Speed Computing, convened by the U.S. national laboratories. As the only undergraduate among the three, I attended alongside H. Shrikumar and Bill Butera, both doctoral researchers at the MIT Media Lab.
I was subsequently sponsored to attend conferences and senior-administrator briefings at National Security Agency headquarters at Fort Meade, outside Washington, D.C.; the World Technology Summit in London; and the French Sénat, where I was invited to testify on the future of technology and the transformations it might visit upon human life in the decades ahead.
Following three days of spirited debate with European royalty, French senators, senior politicians, and international diplomats at the Sénat’s historic Paris hearing on artificial intelligence, Starlab’s principal investigator and AI program lead, Hugo de Garis, turned to me during a brief lull in the dinner conversation, as though some private deliberation had just concluded. With all the restraint for which he was renowned, he pronounced that I might one day be elected President. Far sooner than that, however, he added, I might find myself targeted for assassination. Apparently satisfied that my political future had thereby been mapped in sufficient detail, he turned to the waiter and ordered another bottle of champagne.
In his view, the accelerating advance of artificial intelligence would inevitably meet with violent opposition from those determined to arrest the march of technological progress—activists, extremists, or Luddites of a more modern vintage. He told me that throughout his years at Starlab, he had steadfastly refused to open a parcel or take receipt of any postal delivery, lest he become tomorrow’s headline as the unlucky target of some latter-day Unabomber or would-be copycat. I have often wondered whether our receptionist would have discharged that ordinary office duty with quite the same equanimity, had anyone troubled to explain why the task had befallen her.
Our living quarters at the laboratory comprised an expansive two-story, three-bedroom apartment with cathedral ceilings and a fully stocked library, ordinarily reserved for visiting prime ministers, senators, and senior diplomats. I shared them with de Garis himself. One midsummer afternoon, as we wandered almost at random through the sprawling estate and its wooded grounds, immersed in an impassioned debate over the long-term promise and peril of superintelligence, de Garis conceived a still more theatrical scheme: he had set his mind on acquiring a life-size replica of Fat Man—the 10,300-pound plutonium implosion bomb whose 21-kiloton blast devastated Nagasaki—and suspending it precariously from the soaring vaulted ceiling of my apartment, to hang directly over my bed.
Roughly the size of a Volkswagen Beetle, the replica had been constructed for Brain Child, the Discovery Channel documentary he had just finished filming on the future of artificial intelligence—a film that culminated in the prospect of global thermonuclear war between humanity and its artificial progeny. Having settled the symbolic question to his satisfaction, de Garis proceeded directly to the practicalities: offering to purchase the replica outright from the director and making arrangements for its expedited delivery to Starlab’s headquarters in the forested countryside beyond Brussels. The bomb, he explained, was to hang above me as a daily reminder of the weight of my responsibility to the future of humanity.
de Garis had an unfailing instinct for rendering an abstraction tangible. On this occasion, the abstraction weighed rather more than ten thousand pounds.
The warning, extravagant as its staging had been, has since acquired respectable company. OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis, and Anthropic CEO Dario Amodei are among the signatories to a declaration that the risk of extinction from advanced AI should be treated as a global priority alongside pandemics and nuclear war. Elon Musk, founder of xAI, has voiced related fears.
With the explosive rise of AI over the past few years, and artificial general intelligence and superintelligence now subjects of serious technical and political debate, one might conclude that de Garis—radical and extravagant though his methods were—had not been entirely mistaken. He had merely been early.
Strip away the theatrics and the point stands. Like any powerful tool, these new engines of creation can do society real harm. Unlike any tool that came before them, they will look back at us with eyes of their own. They will see the world in ways we cannot, with an alien intelligence that reaches deeper than our finest minds, and with capacities we can scarcely begin to imagine.
No great transformation has ever come without cost. There will be fires to put out along the road. AI can displace work faster than people can adapt, concentrate wealth and power, and put panoptic surveillance in the hands of totalitarian regimes. It can leave us with atrophied skills, uncritical minds, and far less freedom than we realize we have surrendered. These dangers are real, and I have spent the better part of my working life attending to them, from Starlab onward.
Yet the oldest peril predates every machine: the blackened hearts of men who prey upon the many, who make cruelty and corruption their instruments, who send other people's children into wars they never sought, all to feed an insatiable greed and a hunger for dominion without bound. Whatever intelligence we bring into being will enter a world that has not yet learned to wield its own power wisely.
Set against that darkness stands a brilliant light: a world of plenty beyond all want, delivered at last from the ancient scourges of war, pestilence, aging, and death. Such a world once belonged to prophets and poets. Much of it is now a tractable problem for science and engineering—perhaps a matter of decades, perhaps sooner. There is a reason this passage is called the singularity: it is a horizon past which all our old maps fail. That light of hope is what draws me toward this future.
Nothing about the crossing is guaranteed. It will be neither safe nor simple. My hope is that the digital minds we bring into being will carry more wisdom than we do and fewer of our follies, and that they may help us tip the scales toward the light. But that hope asks something of us in return: the wisdom to shape what we can, the resolve to forge new tools when duty calls, the humility to admit what we cannot foresee, and the courage to remain answerable for what we create.
Answerability begins with measurement. The questions that matter most—whether a system is aligned, what it can do, how much risk it carries, how it ought to be governed—cannot be settled by observing behavior alone. Surface conduct may be strategically misleading, and two systems may act identically while differing entirely in their internal organization.
Physics offers a way past that impasse: structural metrics computed from a system's latent organization rather than from its testimony about itself. The patent-pending Unified Continuation-Interest Protocol is one such instrument. It asks whether an advanced system capable of recursive self-improvement might acquire an interest in its own persistence, and it answers by looking to the latent structure of an agent's trajectories, where agents built with terminal continuation objectives register a measurably higher entanglement entropy than instrumental controls. The Continuation Observatory extends that measurement to frontier models, to learn whether the signal survives contact with more complex systems. It is the first station in a wider program of falsifiable structural measurement across alignment, capability, risk, and governance.
A later retrospective in Sifted traced the arc from our early work at Starlab—including its excursions into AI and time travel—to my subsequent journey across East Asia, where I helped develop national quantum-technology roadmaps for senior U.S. research-funding and intelligence-community leadership.
In the years that followed, I undertook research fellowships in nanoscience and the foundations of quantum mechanics with Nobel laureate Anton Zeilinger’s group in Austria and across Europe. I was later recruited to help create a futures initiative at NASA in collaboration with Google and Ray Kurzweil, bringing together leading companies, scientists, astronauts, venture capitalists, and entrepreneurs from Silicon Valley and around the world.
.jpg)
That path led from human-spaceflight training at NASA to the summit of a volcano where the Apollo 11 astronauts had trained before the first lunar landing; from the development of a quantum-communications architecture based on continuous-variable quantum teleportation to provide space-based NASA assets with unconditional information assurance, to work with diplomats advising the United Nations on critical security questions of the coming age. It led onward to field expeditions in austere desert terrain, where multidisciplinary teams of scientists, special-forces operators, and national agency directors tested next-generation technologies at the absolute limits of their design.
Each of these undertakings was born of the same conviction: that science is not merely an inquiry into what may be done, but a covenant with those who must inherit what we do. That obligation is owed to our children, to our children’s children, and to the generations yet to come.
No comments:
Post a Comment