AI and Biotech Convergence Makes Governance the Central Health Challenge

? jamie-metzl compresses his outlook into a single claim: after nearly 3.8 billion years of evolution, one species has gained the increasing ability to “read, write, and hack the source code of life,” engineer novel intelligence, and reengineer life itself. That capability, he argues, can become the greatest story for humanity and perhaps for life, or the worst. The determining variable is not technical performance. It is whether humans learn to use “god-like powers wisely.”

That frame matters because artificial intelligence, genetics, biotechnology, quantum computing, nanotechnology, and related fields are not developing as separate domains. Metzl’s term for the larger pattern is “superconvergence”: every technology is increasingly embedded in every other technology, with advances in one domain accelerating advances in the rest. AI helps interrogate biology; biology inspires neural networks and new chip designs; computing enables machine learning; ancient writing and number systems become the basis for code. The result is not just a set of breakthroughs but a flywheel.

The stakes become concrete in healthcare and beyond. A rare cancer mutation can turn sequencing into a treatment decision. A community oncologist can be overwhelmed by fast-changing personalized cancer guidelines. A vaccine platform can move from genome sequence to effective shots in less than a year, while research systems can also create unresolved biosafety and transparency questions. Food, materials, and data storage can be redesigned through biology, but the same tools can deepen inequity, invade privacy, amplify bias, dehumanize care, or damage ecosystems if institutions treat capability as permission.

The tension is that the flywheel does not come with a value system. Metzl repeatedly returns to a distinction: the goal is not technology, and the goal is not even health in the abstract. The point of health is to let people live their lives, love the people around them, build community, and realize human potential.

Values, for Metzl, must be embodied in governance. He uses “governance” broadly: government regulation is only one piece. Governance includes what individuals, families, institutions, companies, professional communities, national governments, and global bodies do collectively. This was also, he says, the main recommendation of the World Health Organization expert advisory committee on human genome editing on which he served: powerful technologies cannot be managed only from the top down, even though government action and shared best practices are essential.

The practical warning is directed especially at leaders and institutions asking how they should use new technologies. If the first question is “How can I best use this technology?” the process has already failed. The first question should be: Who am I? Who are we? What do we stand for? What are we trying to achieve? Only after that should an organization ask what relationship between human functions and machine functions will best advance those goals.

? jamie-metzl grounds the promise of precision healthcare in the story of his father, Kurt Metzl: a Kansas City pediatrician for 55 years, a Kansas City Chiefs devotee, a Holocaust survivor and refugee, and, in his son’s telling, an “eternal optimist” whose refrain was “Isn’t this great?” Kurt arrived in the United States at 13 with his family from a village in Austria, was resettled in Kansas City, and eventually became a community pediatrician. The family history matters because it keeps the technological argument attached to lived stakes: time, family, work, ordinary pleasures, and the chance to keep doing what one loves.

Kurt Metzl was diagnosed with stage four metastatic neuroendocrine cancer across multiple organs. At diagnosis, the family was told to hope for about one year of life. Jamie Metzl was then writing the healthcare chapter of Superconvergence, focused on precision healthcare and precision oncology. He told his father that one option was not to be treated; another was to fight for time. His father wanted to fight. Metzl’s goal was to ensure he would be treated “like it’s 10 years into the future of cancer care” rather than, as he put it, the way many people are treated, “kind of like 10 years in the past.”

Before an oncologist had been assigned, Metzl asked the internist to order two things: sequencing of the cancer cells and creation of cancer organoids from cells extracted during the diagnostic biopsy. He describes cancer organoids as balls of cancer cells against which therapies can be tested in a laboratory. The premise is not certainty but probability: if a treatment kills a patient’s cancer cells in the lab, perhaps it has a better chance of working in the person. The internist had never ordered either prescription, but after discussion, did so.

The first treatment was a well-tolerated oral chemotherapy, which Metzl says he double-checked with contacts at Mass General. His father tolerated it well, but scans showed the cancer was still growing. The next standard step proposed was aggressive IV chemotherapy. Metzl says his research suggested about a 30% chance of helping, but a certainty of the negative consequences associated with aggressive chemotherapy: damage to other cells, hair loss, nausea, sores, and other effects. For an 87-year-old whose wife depended on him and whose daily joys were still intact, that did not feel right to the family.

The sequencing result changed the decision. Kurt Metzl’s cancer had a very rare BRAF V600E mutation, almost unheard of in neuroendocrine cancer but more common in lung cancer. That opened the possibility of using a targeted cancer therapy that had been used in many lung cancer patients and, according to the medical records Metzl consulted, in four people with his father’s cancer. He told the oncologist he did not know the odds with such a small case base, but he did know that wiping out his 87-year-old father with aggressive IV chemotherapy would be a major blow, and that someone knocked down at that age might not get back up.

The family chose the targeted therapy. Metzl says he was nervous because if the scans worsened after declining the standard aggressive chemotherapy, his brothers, all physicians, would have blamed him for exercising too much influence as “this jackass with a PhD.” Instead, the scans showed not only slowed spread and not only stopped spread, but a remission that he says had not previously been seen in a patient with his father’s type of cancer and that mutation.

More than a year after diagnosis, when his father had already outlived the initial expectation, Metzl and his brother took him to the Super Bowl, where the Chiefs won in overtime. ESPN, after hearing that Kurt Metzl was a Holocaust survivor, cancer survivor, and Chiefs superfan, arranged Chiefs gear for him, including a helmet signed by Patrick Mahomes. The moment carries the larger point: the goal of technology is not technology; the goal of health is not even health. The goal is to enable people to do the things they love.

The case later moved from one family’s treatment decision into the medical literature. The oncologist, initially hesitant about the unusual approach, published a case report in the Journal of Gastrointestinal Cancer, according to Metzl. The article shown on screen was titled “Prolonged Response to Dabrafenib/Trametinib in Grade 3 Metastatic Pancreatic Neuroendocrine Tumor (NET G3) with BRAF V600E Mutation.” The displayed abstract described a patient with grade 3 pancreatic neuroendocrine tumor metastatic to the liver, lung, lymph node, and scalp, treated with dabrafenib/trametinib after detection of a BRAF V600E mutation. Its visible conclusion stated that tumor testing for actionable mutations should be undertaken at diagnosis and/or progression and that dabrafenib/trametinib should be considered a first-line option for BRAF-mutated metastatic pancreatic neuroendocrine tumors, especially in the grade 3 setting.

The lesson is not that every patient can or should replicate that path. It is that cancer care is changing fast enough that patients and families cannot treat medicine as a passive transaction. Everyone needs to be part of the care team. A cancer diagnosis, in Metzl’s formulation, is not a moment to step back and wait for the doctor to decide everything. The patient and loved ones are part of the care team.

Many people still think about AI too narrowly. If asked whether they have engaged AI today, ? jamie-metzl says, most people will think of opening ChatGPT and asking a question. That is the wrong mental model. If asked how they experienced electricity today, people would not say they visited a power plant. They would recognize that electricity was embedded in the alarm clock, the building, the clothing, the carpet, and nearly everything around them.

AI, in this view, should be understood the same way. It is not something people “go to” when they visit a chatbot. It is already in phones, traffic lights, and many other systems, and it will become more pervasive. The point is not that AI is magical technology. The real story, Metzl says, is “magical humans.”

The civilizational context matters. A century ago, the world had about 2 billion people and a literacy rate of roughly 15%, meaning around 300 million people could participate in knowledge shared beyond their immediate communities. Today, with more than 8 billion people and an 85% literacy rate, about 7 billion people can participate in that shared knowledge world. Because people are networked together, no one has to reinvent what has already been invented. Metzl contrasts that with the Copper Age, Bronze Age, and Iron Age, when thousands of years could separate the emergence of a technological capability in different societies. Without the recipe for copper, bronze, or iron, a society might spend centuries unable to contribute new ideas about what those materials could enable.

The acceleration of technological change comes from this combination: more minds, more educated minds, global networks, and converging tools. Metzl traces a long arc from fire to agriculture to writing to computing. Control of fire allowed humans to cook food and process animal proteins outside the body, which he says eventually allowed energy to be reallocated toward the brain. Agriculture produced surpluses, surpluses enabled larger settlements, and larger settlements required recordkeeping systems. Writing, alphabets, and numbers became the distant basis for computer code: Roman and Phoenician letters, Hindu-Arabic numerals, ones and zeros, A’s and B’s and C’s.

That long arc is meant to make the present less mysterious. Machine learning and AI are not alien artifacts; they sit atop accumulated human systems. In turn, AI is now being used to interrogate living systems, such as how a plant diffuses nutrients through a leaf, and to develop better, faster computer chips. “Technology begets technology,” Metzl says. “Innovation begets innovation.”

? jamie-metzl uses Google DeepMind as a compact illustration of the speed and compounding nature of technological progress. In his account, DeepMind’s AlphaGo defeated Lee Sedol in four games out of five in 2016 after being trained on existing Go games played by human masters. The following year, AlphaZero was given no access to human Go games. It received the rules, played against itself, and learned through a reward mechanism that reinforced moves that improved its odds of winning. After two days, Metzl says, it went from the worst Go player in the world to the best by defeating AlphaGo.

The next move was from games to protein folding. Proteins are central because understanding how a protein functions requires more than the sequence of amino acids represented as letters. Researchers also need the shape, because protein shape largely determines function. Knowing both sequence and shape can affect health, industry, and many other areas because proteins are the building blocks of life.

Metzl says DeepMind’s AlphaFold program initially came in 20th in a 2018 competition to predict protein shapes from amino acid sequences. After the algorithm was reworked, he says, it returned in 2020 and won so decisively that Nature characterized the challenge of predicting protein shapes from amino acid strings as largely solved. In 2021, according to his account, DeepMind released predicted structures for 350,000 proteins, including all proteins in the human body. In 2022, it released predicted shapes for 215 million proteins known to science.

The scale of that shift is central to the argument. Since the 1950s, Metzl says, X-ray crystallography had been used to characterize proteins: crystallize a protein, bombard it with X-rays, take large numbers of images, and infer the shape. On average, characterizing one protein took about three years and required highly trained researchers. If 215 million proteins were known to science and each took about three years to characterize, the work represented roughly 645 million years of human time.

The point is not merely that AlphaFold saved labor. It is that once such information is made available, the frontier shifts for everyone at once. Metzl compares it again to the recipe for metals: if everyone had received the recipe for copper, bronze, or iron on the same day, many societies’ histories would have been radically different. In the protein case, predicted structures that would have required immense cumulative labor were “thrown back into the pot.”

AlphaEvolve, in his description, is a further step. The system asks what constellation of AI algorithms might help solve a problem, tries different approaches, discards what hurts, retains what helps, and learns to write better code to do things differently. DeepMind is only one company in this telling. Thousands of companies are working on comparable transformations in other domains.

The healthcare transformation ? jamie-metzl describes has several linked shifts: generalized to personalized, episodic to continuous, sick care to healthcare, fee-for-service to incentivized outcomes, physical to physical-plus-virtual, opaque to user-friendly, atomized data to interoperable data, and human to human-plus-AI.

At the center is precision healthcare. Generalized medicine treats people as humans, based on population averages. Metzl calls that better than anything humanity has previously had, but not as good, in appropriate cases, as treating a person based on being that particular person. To do that, clinicians and systems need more than genetic information. They need systems-biological information integrated with electronic health records in a useful, privacy-protected dataset over the course of a person’s life.

The broad formula applies beyond healthcare: high-quality datasets, more computing power, and stronger algorithms generate deeper insight into complex systems, and those insights enable more precise manipulation of those systems. In healthcare, Metzl points to early gene and cell therapies, including Victoria Gray, whom he describes as one of the first people treated and largely cured of sickle cell disease through gene therapy. He also points to pharmacogenomics, where prescriptions will increasingly be matched to individual biology to improve the odds that a drug will work.

Today’s wearables are not treated as already sufficient. Metzl notes that many people use Oura rings, Whoop bands, and similar technologies, and agrees with Zeke Emanuel’s view that people are overdoing it relative to the actionable information currently available. But he does not take that as an argument against sensors. Rather, he argues for a more systemic and systematic view of how information can be useful. Sensors will be on bodies, inside bodies, and in surrounding environments. That will expand and decentralize the time and place of healthcare.

The current system, in his phrase, is largely “episodic sick care”: a person has a symptom, goes to a care facility, and is treated based on that symptom. He imagines a distributed model with hospitals at home, referencing existing models in St. Louis and elsewhere. Patients would not simply be discharged with “good luck”; they would have sensors and support that allow a hospital-like experience at home, connected to control centers.

Digital twins are another part of this transformation. At the individual level, Metzl describes a digital twin as a digitized representation of a person outside the body, analogous to a car dashboard where warning lights activate before failure because sensors detect emerging problems. He mentions stories of people receiving signals from an Oura ring or Apple Watch before a COVID diagnosis, indicating that something was wrong. The future version, he says, will be more sophisticated.

But individual data is useful only in a collective context. Normal and abnormal can only be understood against population patterns. As more people have digital representations and as massive systems-biological datasets expand, Metzl expects digital twinning at a societal level. That could reveal trends and support prediction and prevention. He points to Pitkin County wastewater sensors checking for measles and other signals as a simple example of environmental surveillance. With more dynamic connected technologies, public health could have much more actionable information.

He also expects more research on humans and complex biological systems to move “in silico.” This is not a claim that computation will replace reality. Biology is extraordinarily complex, and Metzl says humanity knows “next to nothing” about its full functioning, estimating current understanding at roughly 3% of the full complexity. Real-world work will remain necessary. But in silico research and medicine will still be transformative. He mentions the virtual cell initiative as an aspiration: virtually replicating and understanding a single cell. Even that, he says, is years or decades away.

Research itself will change. Self-driving labs automate functions in wet labs and allow faster experimentation. More consequential, for Metzl, are “co-scientists”: AI systems strung together across the scientific process. He breaks that process into understanding the world, scanning the horizon, developing hypotheses, designing tests and experiments, interpreting results, and publishing papers. The aim is not to replace humans, he says, but to empower humans to be “great humans.”

? jamie-metzl is sharply critical of the idea that institutions can simply “unleash AI” and expect better healthcare. After showing a Wall Street Journal headline — “Cancer Care Is Getting Personal. Local Doctors Can’t Keep Up” — he says it does not matter how many innovations exist if people at scale do not reach them. He also invokes former NIH director Monica Bertagnolli’s warning that most people in America receive care in community facilities for which these advances are overwhelming.

His objection is not to AI in medicine. It is to a simplistic model of AI as the answer to everything. No concept, he says, will get more people killed than the idea that “AI can do that” — that AI will provide the best healthcare, write novels, love children, and solve human obligations if simply set loose. When AI succeeds, it does not look like a robotic hand pressing a glowing artificial intelligence button. It looks like doctors, patients, and families able to do the essential human things better because machines remove burdens or perform specific functions better than humans can.

The human-plus-AI model requires task-level negotiation. Some jobs will be destroyed and others created, Metzl says, but the main transition will happen within jobs. Every job contains multiple tasks. Some will prove essentially human; others will prove essentially machine-suited. The challenge is to identify what humans do best and what machines do best, rather than forcing humans to compete as “second-rate machines.”

He rejects the claim that the future belongs to AI and humans are “the new Neanderthals.” He also rejects, in unusually blunt terms, the popular framing of artificial general intelligence as a moment when AI systems can perform all cognitive tasks better than all humans. He summarizes his view as “AGI is BS” and tells the audience not to let people sell them on it. Humans, he says, are a “magical species” with unlimited potential.

But he immediately names the problem with this optimistic humanism: humans. Humans contribute to solutions and to dangers. In Darwinian terms, he frames human systems through random mutation and natural selection; in contemporary terms, diversity and competition. Competing models drive progress, but they also drive risk.

The concrete danger list he gives for healthcare and related technologies includes integration, cost and return on investment, efficacy, privacy, bias, and dehumanization. A system can imagine extraordinary new capabilities, but if it bankrupts itself pursuing novelty and cannot preserve the best of what already works, it has failed. If tools do not integrate into hard-won systems, if they deepen bias, compromise privacy, or make care less human, the technical achievement is not enough.

? jamie-metzl uses human genome editing and COVID-19 to illustrate the gap between capability and governance. He says he was closely involved in the issue of CRISPR babies. When his previous book, Hacking Darwin, was in production in November 2018, he called his publisher after news emerged that the first two CRISPR babies had been born in China. The book was about the future of human genetic engineering, and he did not want it published without reference to the event.

The bad news, in Metzl’s telling, was that the event had already happened. The good news, from the standpoint of the book’s argument, was that he had predicted the broad pattern: that it would happen, that it would happen in China, and that CCR5 was among the genes he had identified as likely to be targeted first. Afterward, he says, Dr. Tedros invited him to join the World Health Organization Expert Advisory Committee on Human Genome Editing. The committee, co-chaired by Peggy Hamburg, spent three years producing recommendations on guardrails for what Metzl calls one of the most powerful technologies in human history.

COVID-19, in Metzl’s account, shows the same duality. He calls the vaccines among the greatest achievements in human history: moving from a sequenced genome to a fully functional, highly effective vaccine in about 10 to 11 months. Operation Warp Speed, he says, was a tremendous success, and Donald Trump should, in his words, be running around “pumping their fists in the air” for it, though he says changed perceptions of vaccines have prevented that and have become dangerous.

At the same time, Metzl says his own view — while acknowledging differences of opinion — is that the origin of the COVID-19 pandemic was very likely associated with a research-related incident in Wuhan. He says he has written extensively on the topic and was the lead witness in congressional hearings. He claims the truth may never be known because, in his account, the Chinese government destroyed samples, hid records, and made it criminally punishable for Chinese scientists to speak or write about the topic without government clearance.

His hypothesis is specific, and the article should be read as reporting Metzl’s account rather than resolving the dispute. He says the most likely origin was not a bioweapon. Rather, he believes China had a transparent collaboration, with some NIH funding, pursuing one set of work, while a secret effort on the side attempted to develop a pan-coronavirus vaccine. In his telling, the motivation was a leapfrogging success similar to the impulse behind the CRISPR babies case; something went wrong; young researchers were involved; and the early illness was not diagnosed properly.

For Metzl, the vaccines represent extraordinary scientific achievement, while his concern about a possible research-related origin represents a failure of governance. The larger point is that powerful biotechnologies create upside and downside at the same time, and the technologies themselves do not decide which path dominates.

? jamie-metzl extends the AI-genetics-biotech convergence beyond healthcare. The same technologies, he argues, are already beginning to revolutionize agriculture and animal products, offering ways to obtain desired animal products without cruelty, without slaughtering 93 billion land animals per year, and without the current levels of land, energy, water, fertilizer, and antibiotic use.

He anticipates resistance, especially from people who say they do not want “science” in their food and prefer organic food from Whole Foods. His response is that nearly everything people eat is already “radical biotechnology.” Agriculture itself is radical biotechnology. If someone truly wants to avoid radical biotechnology, he says, the available list becomes tiny: cactus fruits, wild swordfish, and little else. The real question is not whether food involves biotechnology, but what kind and under what values and systems.

Industrial materials could also shift from being cut down or dug up to being grown. Metzl does not dwell on the details in this talk, but he places that shift within the broader pattern of learning from living systems.

Data storage is his most developed non-health example. The reason so many data centers must be built, he says, is that what people call the cloud is stored on media such as magnetic tapes that erode in roughly 30 to 50 years. Humanity is creating more data every two years than in all previous human history up to that point, he says, and then must repeatedly copy it forward. Without a different approach, the world risks becoming dominated by data centers.

Nature, in his view, has already solved the data-storage problem. DNA is a million times denser than silicon and, under the right conditions, can store data reliably for two to five million years. He compares the situation to Karl Benz inventing the car in the 1880s while having a Porsche in the workshop as a reference. When humans face hard problems, they can look to how nature has solved analogous ones: data storage, information processing, and low-energy computation.

Again, the excitement is conditional. These capabilities can be useful, but the difference between the good story and the bad story is whether human values are built into systems.

? jamie-metzl’s advice to individuals has three main parts. First, humans should not accept the story that they are obsolete. But they also should not seek care or professional guidance from people who refuse to use modern tools. If an oncologist says they are a naturalist, do not use electricity, do not believe in sequencing, and reject new tools, Metzl says to run the other way. The future is not humans alone or AI alone. It is humans and AI working together.

Second, people must never stop learning. He tells the Aspen audience that they already understand this at one level, or they would be lying on a beach in Cabo rather than attending the conference. But he draws a distinction between line-of-sight learning — the knowledge people already know they need — and adjacent learning. People should spend about two hours a day on adjacent learning, he says. The cue is noticing that something in one’s familiar domain looks different or strange. When that happens, ask why. Follow the answer to another why, and then another. That is how people keep up with changing contexts.

Third, values must precede implementation. This applies to companies, health systems, families, and individuals. If the first question is how to use the technology, Metzl says, the person or organization is lost. The better sequence is to define identity, commitments, and goals, then decide how human and machine functions should relate to achieve them.

The final image for responsibility is Georges Seurat’s A Sunday on La Grande Jatte. Every person is a dot in the painting. If each dot says it has no responsibility for the whole, the painting fails. The better posture is to recognize that each person is a dot with partial responsibility for the whole picture.

This is also how Metzl reframes AI itself. A common image of a human hand touching a robotic hand can make the technologies feel alien, as though something nonhuman has arrived from outside. He insists they are not aliens. AI systems represent thousands of years of human evolution in math, science, logic, writing, computer science, and related domains. The data being mined is human. These are human technologies.

If humans want those technologies to reflect human values, the work has to happen now: through governance, regulation, institutional choices, and serious conversations that become sparks for action in each person’s “own little world.” The best future is not something to wait for. In Metzl’s formulation, if people want to live in the best future they can imagine, they have to build it.