超知能は病気を根絶できるか：AIが掲げる最大の約束

In December 2020, DeepMind’s AlphaFold solved a problem that had confounded biologists for half a century: predicting how a protein folds from its amino acid sequence. Within two years, it had mapped the structures of virtually every known protein — over 200 million — a task that would have taken experimental crystallographers centuries.

That was narrow AI. The question now consuming researchers, investors, and policymakers is what happens when artificial intelligence becomes comprehensively smarter than all of us combined — and pointed at the single greatest source of human suffering: disease.

The promise is extraordinary. The reality is more complicated. But the trajectory of AI in healthcare has already crossed thresholds considered decades away, and the implications demand attention from every leader whose decisions shape human welfare.

The Present: AI Is Already Transforming Medicine

Before reaching for superintelligence, it is worth recognizing how profoundly today’s AI is reshaping healthcare — not in theory, but in clinical practice.

Drug discovery has been compressed from decades to months. Traditional development takes 10 to 15 years at $2.6 billion average cost, with a 90% failure rate. AI is dismantling those numbers. Insilico Medicine advanced a pulmonary fibrosis candidate into Phase 2 trials in under 30 months. Recursion Pharmaceuticals runs millions of experiments weekly using machine learning to find therapeutic relationships invisible to researchers. Isomorphic Labs, the DeepMind spinout, applies AlphaFold’s insights to design molecules with engineered precision.

AI diagnostics are matching and exceeding specialist accuracy. In radiology, deep learning systems detect lung nodules and brain hemorrhages with sensitivity comparable to board-certified radiologists — and they never fatigue. In pathology, AI achieves diagnostic accuracy exceeding 95% for certain cancers. In dermatology, smartphone-based tools classify skin lesions with specialist-level precision, bringing expert screening to populations that have never had access to a specialist.

Personalized medicine is becoming operational. Genomic sequencing, health records, wearable data, and AI analysis converge to enable treatments tailored to individual biology. Tempus matches cancer patients with therapies most likely to work for their specific tumor profile. Foundation Medicine identifies actionable mutations across hundreds of cancer genes. One-size-fits-all medicine is ending — because AI can process the complexity that personalization requires.

These are not pilot programs or research curiosities. They are deployed systems affecting patient outcomes today.

The Superintelligence Hypothesis: Medicine’s Moonshot

Now consider the leap.

Superintelligence — artificial intelligence that surpasses the cognitive capabilities of all humans combined — remains vigorously debated. Some, like Demis Hassabis of DeepMind, believe it could arrive within a decade. Others argue it remains centuries away. The timeline is uncertain. The implications, if it arrives, are not.

A superintelligent system applied to medicine could, in principle, model the entirety of human biology as a single integrated system — every gene, protein interaction, metabolic pathway, and environmental variable — and identify the precise interventions needed to prevent or cure any disease. Cancer, which is not one disease but thousands, could be mapped and countered at a pace that renders multi-year clinical trials obsolete. Alzheimer’s, whose mechanisms have resisted four decades of concentrated effort, might yield to a system that simulates neurological processes at fidelity no research team can approach.

The most provocative claim — that superintelligence could solve aging itself — rests on a defensible premise. Aging is a set of identifiable molecular processes: telomere shortening, cellular senescence, mitochondrial dysfunction, epigenetic drift, accumulated DNA damage. A mind vast enough to model all of them simultaneously could, theoretically, transform aging from an inevitability into a treatable condition.

This is the boldest promise in the history of medicine. It is also the one most vulnerable to hubris.

The Reality Check: Why the Path Is Not Straight

The enthusiasm surrounding AI in healthcare must be tempered by an honest accounting of present limitations.

Biology is not a solved engineering problem. Predicting protein structure is not the same as predicting protein function, and predicting function is not the same as predicting how a molecule behaves in a living body. Drug candidates that perform brilliantly in silico fail in human trials with dispiriting regularity. AI accelerates hypothesis generation. It does not eliminate the fundamental unpredictability of biology.

Data quality constrains AI potential. Medical datasets are plagued by gaps, biases, and inconsistencies. Clinical trial populations have historically underrepresented women, racial minorities, and elderly patients. An AI system trained on biased data does not transcend bias. It scales it.

Regulatory frameworks are struggling to keep pace. The FDA has approved over 800 AI-enabled medical devices, but the regulatory model was designed for static technologies that do not change after approval. Modern AI systems improve through continuous learning, raising fundamental questions about certification. The PMDA in Japan faces the same challenge, compounded by the need to harmonize with international standards.

The superintelligence timeline is deeply uncertain. Current AI systems are remarkably capable, but they operate through pattern recognition — not the deep causal reasoning that solving cancer or Alzheimer’s likely requires. The gap between “very good AI” and “AI that understands biology more completely than every scientist who has ever lived” remains an open question that honest researchers acknowledge rather than paper over.

Japan: Where AI Healthcare Meets Urgent Necessity

No major economy faces the healthcare challenge that AI promises to address more acutely than Japan.

With 29% of its population over 65 and life expectancy of 84.6 years, Japan is a triumph of healthcare and a case study in the pressures of success. Healthcare expenditure surpassed JPY 46 trillion ($310 billion) in 2025, with age-related conditions driving relentless cost growth.

AI offers a structural response. The Fugaku supercomputer at RIKEN’s Center for Computational Science in Kobe has been deployed for drug simulation, modeling how candidate therapeutics interact with target proteins at atomic resolution. Takeda and Daiichi Sankyo have established AI drug discovery partnerships. The University of Tokyo is pioneering clinical decision support systems designed for Japan’s specific disease burden.

The PMDA has been developing frameworks for AI-based medical devices, drawing on the precedent set by the 2014 Act on the Safety of Regenerative Medicine. Japan’s challenge is not willingness to adopt AI in healthcare — it is ensuring that adoption keeps pace with the demographic urgency that demands it.

The Ethical Imperative: Intelligence in Service of Equity

The most consequential question about superintelligent medicine is not whether it will work. It is who it will work for.

Access inequality could become the defining injustice of the century. If AI-driven therapies serve only wealthy nations or premium patients, the result betrays medicine’s commitment to universal welfare. Healthcare is a fundamental human right. Technology that makes the best care dramatically better while leaving the world behind undermines that right.

Bias scales with power. Dermatology AI trained predominantly on lighter skin tones has demonstrated lower accuracy for darker-skinned patients. Superintelligent systems trained on biased datasets will not transcend disparities — they will automate them.

Life-or-death decisions demand accountability. When an AI recommends a treatment or flags a scan as benign, who bears responsibility if it is wrong? The liability frameworks governing human physicians do not map cleanly onto algorithmic decision-making. The black-box problem in healthcare is not an inconvenience. It is potentially fatal.

The concentration of capability raises governance questions. A handful of companies control the most advanced AI research on earth. If superintelligent medicine emerges from these laboratories, global health governance could rest with entities that have no democratic mandate.

These are not problems to solve after the technology arrives. They determine whether it serves humanity broadly or merely its most powerful members.

The Convergence: Why This Moment Matters

The trajectory of AI in healthcare is not a straight line toward superintelligence. It is a series of compounding advances — better models, richer datasets, faster compute, smarter molecular design — each delivering incremental benefit while moving the frontier closer to transformative capability.

That compounding is accelerating. BenevolentAI identified baricitinib as a potential COVID-19 treatment early in the pandemic — a connection invisible to human researchers, subsequently validated in clinical trials. AlphaFold’s database now serves over two million researchers worldwide. AI-designed antibodies are entering clinical trials.

Whether superintelligence arrives in ten years or fifty, the decisions being made today — about data governance, regulatory frameworks, access models, and ethical guardrails — will determine whether this revolution serves global equity or deepens division. These decisions cannot wait for the technology. They must precede it.

Join the Conversation

On April 26, 2026, the Tech for Impact Summit convenes senior executives, policymakers, and technologists at Tokyo Garden Terrace Kioi Conference to confront the questions defining our trajectory toward 2050. The theme — “Beyond Boundaries: Building 2050 Together” — speaks directly to the boundary between human intelligence and artificial superintelligence.

Confirmed speakers include Taro Kono (former Minister of Digital Affairs), Charles Hoskinson (Cardano), Yoshito Hori (GLOBIS), Kathy Matsui (MPower Partners), Ken Suzuki (SmartNews), Jesper Koll (Monex Group), Sota Watanabe (Astar/Startale), and Hiroshi Aoi (Marui Group).

Whether you lead a healthcare enterprise, a pharmaceutical company, or a policy institution grappling with AI governance in medicine, the superintelligence question is no longer speculative — and your voice in shaping its direction has never been more needed.

Explore partnership and membership opportunities →

Watch highlights from previous summits: youtu.be/ujy7ZXflrt4

---

The Tech for Impact Summit is an invitation-only executive gathering taking place April 26, 2026, in Tokyo as a partner event of SusHi Tech Tokyo. Learn more at tech4impactsummit.com.