Beating Cancer, Heart Disease, and Neurodegeneration

The Three Horsemen

They kill more people than everything else combined.

Heart disease claims roughly 18 million lives per year globally—one in three deaths.¹ Cancer takes 10 million.² Neurodegeneration—Alzheimer's, Parkinson's, ALS, and related conditions—kills millions more and robs countless others of their minds before their bodies fail.³

These are not the diseases of poverty. They are the diseases of success. Humanity conquered the infections that killed ancestors young, extended life expectancy by decades, and now faces the conditions that emerge when bodies survive long enough to accumulate damage. The reward for not dying of cholera or smallpox is the opportunity to develop cancer or dementia.

For a century, medicine has fought these conditions with increasing sophistication and decreasing progress. Cancer survival rates have improved, but slowly. Heart disease mortality has declined in developed countries, but the disease remains dominant. Alzheimer's has resisted every drug designed to treat it—a graveyard of failed clinical trials stretching back decades.⁴

The question this chapter asks: Can AI-accelerated research change the trajectory? Not merely incremental improvement, but transformative progress—the kind that would make these diseases as manageable as the infections humanity has already conquered?

The answer is conditional. There are paths to dramatic progress. They are not guaranteed. But the tools are becoming powerful enough that the question is no longer whether such progress is possible, but whether humanity can navigate the obstacles between here and there.

2026 Snapshot — Where Things Stand

Cancer

Cancer is not one disease but hundreds, sharing a common feature: cells that grow without constraint. Each cancer type has distinct biology, distinct causes, and distinct responses to treatment. "Curing cancer" is not one problem but a family of related problems, each difficult in its own way.

Current treatment involves some combination of:

Surgery: physically removing tumors (works when cancer is localized)
Radiation: destroying cancer cells with targeted energy (also works for localized disease)
Chemotherapy: drugs that kill rapidly dividing cells (effective but toxic)
Targeted therapy: drugs designed for specific molecular abnormalities (less toxic, only works for matching mutations)
Immunotherapy: reprogramming the immune system to attack cancer (transformative for some cancers, ineffective for others)
Cell therapy: engineering a patient's own cells to fight cancer (CAR-T, remarkable for blood cancers, still experimental for solid tumors)

Five-year survival rates vary enormously: over 90% for some cancers caught early (prostate, thyroid, melanoma), under 10% for others (pancreatic, glioblastoma, late-stage lung).⁵

AI is already contributing to cancer care:

Imaging analysis that detects tumors earlier and more accurately
Pathology that identifies cancer subtypes from tissue samples
Genomic interpretation that matches patients to targeted therapies
Drug discovery that identifies new compounds and targets
Clinical trial matching that finds appropriate studies for patients

But these are augmentations to existing approaches. They make current medicine more efficient. The question is whether AI can enable fundamentally different approaches.

Heart Disease

Cardiovascular disease encompasses a range of conditions: coronary artery disease (blocked heart arteries), heart failure (weakened pumping), arrhythmias (irregular rhythms), stroke (blocked or bleeding brain vessels), and others. The common thread is damage to the circulatory system—the infrastructure that delivers oxygen and nutrients to every cell.

Current treatment focuses on:

Risk factor management: controlling blood pressure, cholesterol, blood sugar, and weight through lifestyle and medication
Medications: statins, blood pressure drugs, anticoagulants, and others that reduce risk and slow progression
Interventions: stents that open blocked arteries, ablation that corrects arrhythmias, pacemakers and defibrillators that regulate rhythm
Surgery: bypass grafts that route blood around blockages, valve repairs or replacements
Heart failure management: medications and devices that support failing hearts, with transplant as last resort

Heart disease mortality in developed countries has dropped substantially since its peak in the 1960s—a public health success story driven by smoking reduction, blood pressure control, cholesterol management, and better acute care.⁶ But heart disease remains the leading killer, and progress has plateaued in recent years.

AI applications include:

Risk prediction models that identify high-risk patients earlier
ECG and imaging analysis that detects abnormalities
Monitoring through wearables that catch arrhythmias
Drug discovery for new cardiovascular targets

The fundamental challenge with heart disease is that prevention works—the knowledge of how to reduce risk exists—but implementation fails. People don't take their medications, don't change their diets, don't exercise. The disease that kills most people is largely preventable with existing knowledge.

Neurodegeneration

The brain is the hardest organ. Neurodegeneration is the hardest problem.

Alzheimer's disease destroys memory and cognition over years, affecting over 50 million people globally.⁷ Decades of research and dozens of clinical trials have produced almost nothing that changes disease progression. The few approved drugs provide modest symptomatic benefit or, in the case of recent antibody therapies (aducanumab, lecanemab), unclear clinical impact despite measurable biomarker changes.⁸

Parkinson's disease causes movement disorders through the death of dopamine-producing neurons. Existing treatments (levodopa, deep brain stimulation) manage symptoms but don't slow degeneration.

ALS (amyotrophic lateral sclerosis) progressively paralyzes patients, typically killing within a few years of diagnosis. Treatments extend survival marginally.

Huntington's disease, frontotemporal dementia, and other conditions follow similar patterns: progressive neuronal death, limited therapeutic options, poor outcomes.

Why is neurodegeneration so hard?

The brain is protected: The blood-brain barrier keeps most drugs out, limiting treatment options
Mechanisms remain poorly understood: Despite decades of research, the fundamental causes of most neurodegenerative diseases remain unclear or contested
The brain doesn't regenerate: Unlike other organs, the brain has limited capacity to replace lost neurons
Clinical trials are slow: Neurodegeneration progresses over years; trials must run for years to detect effects
Good biomarkers have been lacking: Until recently, disease progression could not be measured except through cognitive tests and imaging of advanced damage

AI is beginning to help:

Analysis of brain imaging to detect earlier signs
Interpretation of biomarkers (CSF, blood) to track progression
Drug discovery targeting novel mechanisms
Patient stratification to identify who might respond to treatment

But the honest assessment is that AI has not yet produced breakthrough therapies for neurodegeneration. The tools are being applied, but the fundamental scientific problems remain unsolved.

Notable Players

Cancer: Research and Treatment

Major academic cancer centers drive research and set treatment standards: MD Anderson, Memorial Sloan Kettering, Dana-Farber, UCSF, Johns Hopkins, Mayo, and international equivalents.

Immunotherapy leaders have transformed treatment: Bristol-Myers Squibb (Opdivo, Yervoy), Merck (Keytruda), Roche (Tecentriq). These checkpoint inhibitors have become foundational for many cancers.

CAR-T developers: Novartis (Kymriah), Gilead/Kite (Yescarta), Bristol-Myers (Breyanzi, Abecma). Second-generation approaches aim to make cell therapy more accessible and effective for solid tumors.

mRNA cancer vaccine developers: BioNTech and Moderna are conducting trials of personalized cancer vaccines that train the immune system against patient-specific tumor mutations.

AI-driven oncology: Tempus uses AI to analyze clinical and molecular data for treatment matching. Foundation Medicine provides comprehensive genomic profiling. Multiple startups apply AI to various steps of cancer research and treatment.

Cardiovascular

Device makers: Medtronic, Abbott, Boston Scientific, and Edwards Lifesciences provide pacemakers, stents, valves, and monitoring devices.

Pharma: Major cardiovascular drug developers include Novartis, Pfizer, Bayer, and AstraZeneca. Recent advances include PCSK9 inhibitors for cholesterol and SGLT2 inhibitors that help both diabetes and heart failure.

Continuous monitoring: Apple, Fitbit, AliveCor, and others are building consumer devices that detect cardiac abnormalities. Implantable monitors from Abbott and Medtronic provide longer-term surveillance.

Digital therapeutics: Omada Health, Livongo (now part of Teladoc), and others attempt to change behavior through digital coaching—addressing the implementation gap that limits cardiovascular prevention.

Neurodegeneration

Pharma: Biogen and Eisai developed the controversial amyloid antibodies (aducanumab, lecanemab). Roche, Lilly, and others have their own programs targeting various mechanisms.

Diagnostics: Companies like C2N Diagnostics and ALZpath are developing blood tests that can detect Alzheimer's pathology years before symptoms, potentially enabling earlier intervention.

Research institutions: Broad Institute, UK Dementia Research Institute, Allen Institute for Brain Science, and major academic centers drive fundamental research.

AI in neuro: Verge Genomics, Recursion, and others apply AI to identify drug targets and compounds for neurodegenerative diseases. The field remains early-stage.

Cancer: Paths to Transformation

How might AI-accelerated research transform cancer from a death sentence into a manageable condition? Several paths are emerging.

Personalized Cancer Vaccines

The concept: every tumor is genetically unique. Mutations that drive cancer growth create proteins (neoantigens) that the immune system can potentially recognize as foreign. If you can identify these mutations and train the immune system to attack them, you might clear the cancer.

The process:

Sequence the tumor to identify mutations
Predict which mutations create neoantigens likely to provoke immune response
Synthesize mRNA encoding those neoantigens
Inject the mRNA; cells produce the neoantigens; the immune system learns to recognize them
Immune cells seek and destroy cancer cells displaying those neoantigens

This is precision medicine at its most precise: a treatment manufactured for a single patient based on their unique tumor biology.

Current status: BioNTech and Moderna are conducting Phase 2 trials of personalized cancer vaccines, often in combination with checkpoint inhibitors. Early results in melanoma and other cancers have been promising.⁹ Adjuvant trials (treating patients after surgery to prevent recurrence) are underway.

AI's role: Predicting which mutations will generate effective immune responses is computationally intensive. AI models analyze the patient's HLA type (immune system genetics), the characteristics of potential neoantigens, and vast training data from other patients to predict which targets are most promising. Without AI, this prediction would be impossibly slow and inaccurate.

Trajectory: Near-term likely for some cancers. The technology works in principle; the questions are efficacy across tumor types, manufacturing scale, and cost. If successful, cancer vaccines could become standard treatment within a decade for cancers with high mutation burdens.

AI-Designed Cell Therapies

CAR-T cell therapy has produced remarkable results in blood cancers: patients with terminal leukemia or lymphoma have achieved complete, durable remissions.¹⁰ The approach involves extracting a patient's T cells, genetically modifying them to recognize a cancer marker, expanding them in the lab, and reinfusing them.

Current limitations:

Works mainly for blood cancers: Solid tumors present a hostile environment that defeats current CAR-T approaches
Expensive: Each treatment is custom-manufactured, costing $400,000-500,000 or more
Toxic: Cytokine release syndrome and other side effects can be severe
Manufacturing delays: The weeks required to produce cells can be fatal for fast-progressing cancers

AI is attacking these limitations:

Improved target selection: AI analyzes tumor biology to identify targets expressed on cancer but not healthy tissue, reducing toxicity.

Off-the-shelf cells: Rather than modifying each patient's cells, allogeneic approaches use donor cells that can be manufactured in advance. AI helps design modifications that prevent rejection.

Armored cells: AI-designed cells include additional modifications that help them survive the hostile tumor microenvironment, secrete beneficial factors, or resist exhaustion.

Manufacturing optimization: AI optimizes production processes to reduce time and cost.

Trajectory: Plausible for substantial expansion of cell therapy. Solid tumor CAR-T remains challenging but is the focus of intense research. Off-the-shelf approaches could dramatically reduce cost and improve access if they prove equivalent to autologous cells.

Tumor Digital Twins

The concept: create a computational model of an individual patient's tumor—its genetics, its microenvironment, its interactions with the immune system—and use that model to predict which treatments will work.

Currently, oncologists rely on population-level evidence: drugs that worked for patients with similar cancers. But cancers are heterogeneous. A drug that works for 30% of patients with a given cancer type does nothing for the other 70%. Knowing which group a patient falls into before treatment would save time, toxicity, and lives.

Digital twin approach:

Gather comprehensive data on the tumor: genomics, transcriptomics, proteomics, imaging, pathology
Build a computational model that captures the tumor's biology
Simulate the effects of different treatments on that specific tumor
Recommend the treatment predicted to work best

Current status: Early-stage research. Companies like Champions Oncology create mouse avatars (patient-derived xenografts) for drug testing; computational approaches aim to do this in silico.

AI's role: The models are too complex for traditional analysis. AI integrates multi-omics data, learns from outcomes across patients, and generates predictions that would be impossible through manual analysis.

Trajectory: Plausible for improving treatment selection. True "digital twins" that reliably predict individual responses remain challenging, but intermediate approaches—better biomarkers, improved patient stratification—are advancing.

Earlier Detection at Scale

Cancer is most curable when caught early. The five-year survival rate for stage 1 breast cancer exceeds 99%; for stage 4, it's around 30%.¹¹ Similar patterns hold across cancer types.

The liquid biopsy technologies described in Chapter 3—detecting cancer from blood draws—could shift the stage distribution dramatically. If most cancers are caught at stage 1 rather than stage 3 or 4, outcomes transform even without new treatments.

The math: Catching cancers two stages earlier might double or triple survival rates for many cancers. Applied at population scale, this could save millions of lives annually.

AI's role: Analyzing the complex signals in blood (ctDNA, methylation patterns, protein markers) requires sophisticated machine learning. As detection improves, AI will also help interpret findings—distinguishing true positives from false positives, predicting the source tissue, and recommending follow-up.

Trajectory: Near-term likely for high-risk populations; plausible for population-wide screening within a decade. The technology exists; the questions are sensitivity, specificity, cost, and integration with clinical workflows.

Heart Disease: Paths to Transformation

Heart disease kills through accumulated damage—decades of arterial plaque buildup, gradual weakening of heart muscle, progressive dysfunction. Unlike cancer, where you can point to a tumor and say "remove this," cardiovascular disease is diffuse, systemic, and deeply entwined with lifestyle.

The transformative approaches fall into different categories than cancer: prevention optimization, damage reversal, and monitoring that catches problems earlier.

Precision Prevention

The risk factors are known: blood pressure, cholesterol, blood sugar, smoking, weight, activity, diet. Interventions demonstrably work: statins reduce heart attacks, blood pressure control prevents strokes, lifestyle changes reduce risk.

The problem is implementation. Prescriptions go unfilled. Medications are taken inconsistently. Lifestyle changes fail. The knowledge exists; behavior doesn't follow.

AI-enhanced prevention:

Personalized risk prediction: Models that incorporate genetics, biomarkers, and lifestyle data to generate individualized risk scores more accurate than traditional calculators. Someone told they have an 8% ten-year risk might not act; someone told they have a 25% risk might.¹²

Precision medication: Genetic factors influence drug response. Some people metabolize statins differently; some respond better to certain blood pressure medications. AI can help match patients to optimal drug regimens.

Behavioral intervention: AI-powered coaching (chatbots, apps, connected devices) that provides personalized, timely nudges toward healthier behavior. Early evidence suggests these can improve medication adherence and lifestyle factors, though effects are modest.¹³

Continuous monitoring: Wearables that track activity, diet, sleep, and physiological markers, feeding data to AI systems that identify risks and recommend changes. The shift from annual checkups to continuous optimization.

Trajectory: Near-term likely for incremental improvements. The question is magnitude—whether AI-enhanced prevention can close the implementation gap or whether the fundamental behavioral challenges remain.

Plaque Regression and Vascular Repair

The holy grail of cardiovascular medicine: not just stopping plaque growth but reversing it. Removing the buildup that blocks arteries and causes heart attacks.

Current approaches:

Aggressive lipid lowering: Very low LDL cholesterol (achieved with high-dose statins and PCSK9 inhibitors) can modestly shrink plaques in some patients
Inflammation reduction: Drugs targeting inflammatory pathways (like colchicine) may stabilize plaques
Emerging targets: ANGPTL3, Lp(a), and other targets offer new approaches to cardiovascular risk

AI-accelerated possibilities:

Novel target discovery: AI analysis of genetic data, omics profiles, and clinical outcomes to identify new drug targets for plaque regression

Drug design: AI-designed molecules that can more effectively clear arterial plaque or stabilize vulnerable lesions

Regenerative approaches: Longer-term possibilities include treatments that repair damaged blood vessels, regenerate heart muscle, or rejuvenate aged cardiovascular tissue

Trajectory: Plausible for improved treatments within a decade. True plaque reversal sufficient to restore arteries to youthful condition remains challenging but is the focus of active research.

Continuous Cardiac Monitoring and Early Intervention

Many cardiac deaths are sudden—heart attacks, arrhythmias, strokes. The patient is fine until they're not. If problems could be detected earlier, intervention could prevent catastrophe.

Current capabilities:

Apple Watch and similar devices detect atrial fibrillation and alert users
Implantable monitors track heart rhythms continuously for years
Continuous blood pressure devices are emerging

AI-enhanced monitoring:

Pattern detection: AI that identifies subtle changes in heart rate variability, activity patterns, or other signals that precede cardiac events—potentially by days or weeks

Risk stratification: Identifying which detected abnormalities require immediate attention versus which can wait

Automated response: Systems that automatically alert clinicians, schedule appointments, or in extreme cases dispatch emergency services

Integration: Combining data from multiple sources (wearables, implants, home devices, EHR) into unified risk assessment

Trajectory: Near-term likely for improved detection of arrhythmias. Plausible for prediction of heart attacks or strokes before they occur, though this remains an active research challenge.

Neurodegeneration: The Hardest Problem

If cancer is hard and heart disease is challenging, neurodegeneration is something else entirely. The field has been humbled repeatedly. Confident theories have failed. Promising drugs have collapsed in trials. The brain remains largely mysterious despite decades of study.

Can AI change this? The honest answer is: possibly, but nothing is guaranteed. Here's what might help.

Understanding Mechanisms

The fundamental problem with neurodegeneration is that science does not fully understand what causes it. Alzheimer's has been dominated by the "amyloid hypothesis"—the idea that accumulation of amyloid-beta protein causes the disease. Billions of dollars have been spent on drugs that clear amyloid. The results have been largely disappointing: some biomarker changes, minimal clinical benefit.¹⁴

Maybe amyloid is wrong, or incomplete, or only part of the story. Maybe there are subtypes of Alzheimer's with different causes. Maybe the timing of intervention matters—too late and the damage is irreversible. Maybe multiple pathways need to be targeted simultaneously.

AI's potential contribution:

Multi-modal data integration: Combining genetics, brain imaging, biomarkers, clinical data, and molecular studies across thousands of patients to find patterns human researchers miss

Target discovery: Identifying new drug targets through analysis of genetic variants associated with disease, protein interaction networks, and other biological data

Causal inference: Distinguishing correlation from causation in complex biological systems—which changes drive disease versus which are downstream effects

Patient stratification: Identifying subtypes of disease that may respond to different treatments, rather than treating "Alzheimer's" as a single entity

Current example: Google DeepMind's AlphaFold has predicted structures of proteins involved in neurodegeneration, providing clues to their function. AI analysis of UK Biobank and similar datasets has identified novel genetic associations.¹⁵

Trajectory: Plausible that AI accelerates mechanistic understanding. Whether this translates to effective treatments depends on whether the fundamental biology is tractable—and that's not guaranteed.

Earlier Detection and Intervention

Neurodegeneration begins years or decades before symptoms appear. By the time someone shows memory problems from Alzheimer's, substantial brain damage has already occurred. If the disease could be detected earlier, intervention might occur when more brain tissue is still salvageable.

Emerging biomarkers:

Blood tests: P-tau217 and other markers can now detect Alzheimer's pathology from blood draws, years before symptoms¹⁶
Retinal imaging: The retina is an accessible extension of the brain; changes there may reflect brain changes
Digital biomarkers: Subtle changes in speech patterns, typing behavior, or phone use might signal cognitive decline

AI's role: Detecting subtle patterns in these biomarkers that indicate early disease. Integrating multiple data streams (blood, imaging, behavior) into unified risk assessment. Predicting who will progress to symptomatic disease.

The challenge: Early detection only helps if treatments exist that work when given early. Currently, no convincingly effective treatments exist at any stage. Early detection without effective intervention creates distress without benefit.

Trajectory: Near-term likely for improved early detection. Whether this translates to improved outcomes depends on therapeutic development.

Disease Modification Through Novel Modalities

If traditional small molecule drugs haven't worked, perhaps different approaches will.

Gene therapy: Correcting or compensating for genetic variants that cause or contribute to neurodegeneration. Early approaches are in trials for rare genetic forms of disease.

Antisense oligonucleotides (ASOs): Drugs that reduce production of toxic proteins. Approved for spinal muscular atrophy; in trials for Huntington's and other conditions.

Regenerative approaches: Stem cell therapies that could potentially replace lost neurons. Early-stage research with many challenges (getting cells to the right place, integrating into circuits, avoiding tumors).

Immune modulation: The brain has its own immune system (microglia, astrocytes) that may contribute to neurodegeneration. Targeting this immune response might slow disease.

AI's contribution: Designing gene therapies, identifying optimal ASO sequences, discovering small molecules that modulate novel targets, predicting which patients will respond to which approaches.

Trajectory: Plausible that AI accelerates development of these modalities. Wild speculation would be actual reversal of established neurodegeneration—replacing lost neurons and restoring function—but this is not currently on the horizon.

Neuroprotection and Resilience

A different approach: rather than attacking the disease mechanism directly, protect neurons from damage or enhance the brain's ability to compensate.

Some people carry the genetic variants for Alzheimer's but don't develop symptoms, or develop them later than expected. Understanding what protects them might lead to treatments.

Possibilities:

Enhancing autophagy (cellular cleanup mechanisms)
Supporting mitochondrial function
Reducing inflammation
Enhancing synaptic plasticity and cognitive reserve

AI's role: Identifying protective factors through analysis of resilient individuals. Discovering compounds that enhance neuroprotection. Optimizing combination approaches.

Trajectory: Plausible as a complementary approach. Unlikely to provide complete solutions but might meaningfully slow progression.

Second-Order Impacts

If AI-accelerated research substantially improves outcomes for these three killers, what changes?

Lifespan Extends Further

Heart disease, cancer, and neurodegeneration are the primary causes of death in developed countries after about age 50. Substantially improving outcomes for all three would extend life expectancy significantly—potentially by years or decades.

This cascades into everything: retirement systems, healthcare costs (people live longer but eventually get other conditions), family structures (more generations alive simultaneously), and economic productivity (more years of potential contribution but also more years of potential dependency).

Healthcare Economics Transform

Current healthcare spending is heavily weighted toward treating these conditions, especially near end of life. If interventions shift from expensive late-stage treatment to earlier, potentially cheaper intervention, the economic dynamics change.

But "cheaper" isn't guaranteed. Prevention and early intervention might cost less per event prevented, but if everyone receives them, total spending could increase. The cost of widespread screening, continuous monitoring, and preventive treatment might exceed the savings from avoided late-stage care.

The answer depends on how the technologies develop and how payment systems adapt.

Inequality Risks Increase

If transformative treatments are expensive and limited, access becomes a critical question. Will AI-designed cancer vaccines be available to everyone, or only to those with the best insurance or most money? Will precision cardiovascular prevention reach underserved communities, or widen existing gaps?

The technologies themselves are neutral; distribution is a choice. Policy decisions made in the coming decade will determine whether these advances reduce health inequality or amplify it.

Psychological Shifts

Living with the expectation of dying from cancer or heart disease in your 70s or 80s is different from living with the expectation of avoiding these conditions. If the big killers become manageable, people's relationship to mortality changes.

This could be positive (less fear, more long-term planning) or negative (what do you die from if not these? What meaning comes from confronting mortality?). The psychological and cultural effects are unpredictable.

Risks and Guardrails

Overtreatment and Overdiagnosis

Finding more cancers earlier doesn't automatically improve outcomes. Some detected cancers would never have caused harm—indolent tumors that grow slowly or not at all. Treating these causes harm (surgery, chemotherapy, radiation) without corresponding benefit.

As screening becomes more sensitive, distinguishing dangerous cancers from harmless ones becomes critical. AI systems must be evaluated not just on detection rates but on outcomes: are the cancers they detect ones that would have caused harm?

Unequal Access

The most sophisticated treatments—personalized cancer vaccines, CAR-T cell therapy, comprehensive cardiovascular monitoring—are currently expensive and limited. If transformative treatments remain inaccessible to most people, the result is a two-tier system where the wealthy live and the poor die of preventable conditions.

This is a policy choice, not a technological inevitability. Universal coverage, price regulation, public investment, and other mechanisms can ensure broad access—if the political will exists.

False Hope and Fraud

When people are desperate—facing cancer, watching a parent decline into dementia—they're vulnerable to exploitation. Unproven treatments, miracle cures, and outright fraud thrive in spaces where established medicine fails.

As AI makes drug discovery and personalized medicine more accessible, the risk of poorly validated or fraudulent treatments increases. Regulatory frameworks must adapt to distinguish genuine advances from snake oil, without choking legitimate innovation.

Data Requirements and Privacy

AI-driven personalized medicine requires comprehensive data: genomics, imaging, clinical history, continuous monitoring. Assembling this data at scale raises privacy concerns.

Health data can be misused for discrimination, surveillance, or exploitation. Comprehensive health surveillance—even in service of better medicine—creates risks that require careful governance.

Research Integrity

AI can generate hypotheses faster than humans can test them. This could accelerate discovery—or flood the literature with poorly validated claims. If AI-generated research isn't held to rigorous standards, the signal-to-noise ratio in science could worsen.

Peer review, replication, and other scientific norms must adapt to an era of AI-assisted research without abandoning the fundamental commitment to evidence.

The Path Forward

Three diseases. Centuries of research. Millions of lives lost every year. Can AI change the trajectory?

Near-term likely (5-7 years):

AI-assisted cancer detection catches more tumors earlier
Personalized cancer vaccines enter clinical practice for some tumor types
Cardiovascular risk prediction becomes more precise and personalized
Blood biomarkers enable earlier detection of neurodegeneration
Drug discovery timelines for all three disease areas compress

Plausible (7-15 years):

Cancer becomes primarily a chronic disease—serious but manageable for most patients
Most cardiovascular events are prevented through precision prevention
Disease-modifying treatments for Alzheimer's and other neurodegenerative conditions reach patients
AI-designed therapies targeting novel mechanisms show transformative results

Wild (speculative):

Cancer is functionally cured—detected and eliminated before it causes harm in nearly all cases
Heart disease becomes rare through vascular regeneration and reversal of accumulated damage
Neurodegeneration is not just slowed but reversed—cognitive function restored
The three horsemen are tamed, no longer the dominant causes of death

The gap between plausible and wild is real. These are hard problems that have resisted decades of effort. AI provides new tools, but tools don't guarantee outcomes.

What's different now is the scale of attack: more hypotheses tested, more data integrated, more targets explored, more molecules designed, more patients stratified—all accelerated by AI capabilities that didn't exist a decade ago. Whether this is enough to break through barriers that have held for generations is the question the next decade will answer.

Endnotes — Chapter 4

World Health Organization (WHO) Global Health Estimates. Cardiovascular diseases are the leading cause of death globally, killing an estimated 17.9 million people annually.
WHO Global Cancer Observatory (GLOBOCAN). Cancer caused an estimated 10 million deaths in 2020, with incidence expected to rise as populations age.
Alzheimer's Disease International estimates over 55 million people living with dementia globally, with numbers expected to nearly triple by 2050. Parkinson's affects over 8 million.
The history of failed Alzheimer's trials is extensive. See the Alzheimer's Drug Discovery Foundation database for a comprehensive list of discontinued candidates.
SEER (Surveillance, Epidemiology, and End Results) Program data shows wide variation in five-year survival by cancer type and stage at diagnosis.
CDC data shows age-adjusted heart disease death rates in the US declined by approximately 70% from their peak in the 1960s to their nadir around 2011, though progress has stalled or reversed since.
Alzheimer's Disease International World Alzheimer Report provides global prevalence estimates and projections.
Aducanumab (Aduhelm) received accelerated FDA approval in 2021 despite contested clinical benefit. Lecanemab (Leqembi) showed modest but statistically significant slowing of cognitive decline in trials.
BioNTech and Moderna have reported early trial results for personalized cancer vaccines in melanoma and other cancers, with some showing improved disease-free survival in combination with checkpoint inhibitors.
CAR-T therapies have achieved complete response rates exceeding 50% in some relapsed/refractory blood cancers, with durable remissions lasting years in a significant subset of patients.
American Cancer Society data shows stage-specific survival rates across cancer types, with dramatic differences between early and late-stage detection.
Multiple studies have shown that personalized risk visualization can motivate behavioral change, though effects vary by population and communication approach.
Meta-analyses of digital behavior change interventions show modest but significant effects on physical activity, diet, and medication adherence, with high variability across programs.
The amyloid hypothesis has been challenged by the failure of most amyloid-targeting drugs to produce meaningful clinical benefit. Recent antibody therapies show biomarker effects but modest clinical impact.
AlphaFold has predicted structures for thousands of proteins implicated in neurodegenerative diseases. UK Biobank analyses have identified genetic variants associated with Alzheimer's risk and resilience.
Blood biomarkers including p-tau217 show high accuracy in detecting Alzheimer's pathology years before symptom onset, potentially enabling earlier trial enrollment and future prevention strategies.