Remaining 84 Back to PAICON
PAICON
The Remaining 84 ·
 A movement by PAICON

What happens when AI ignores 84% of the world?

When AI does not see you, it cannot serve you. Remaining 84 is PAICON's commitment to rebuild health AI with true global representation at its core. Built for 100%, rooted in the 84% medicine forgot.

Join the movement ↓ Order the book →

◆All profits from this book are dedicated to NGOs supporting orphaned children worldwide.

Remaining 84, by Dr. Manasi Aichmueller Ratnaparkhe
“
This is not a book about diversity as a social obligation. It is a book about what medicine misses when it builds its foundations on a fraction of humanity, and what becomes possible when it decides to build them for everyone.
Dr. Manasi Aichmueller RatnaparkheCEO & Co-Founder, PAICON
The problem

A narrow few are teaching AI to treat everyone.

Caucasian African Hispanic E.Asian S.Asian SE.Asian
Global Population 16% Caucasian

This is the world. Ancestry shapes disease biology in every region, and no single group comes close to a majority of the people medicine has to serve.

This is the data AI learns from. Training sets are overwhelmingly Caucasian, and every other ancestry is largely invisible to today's models. The Remaining 84 has been receiving medicine built for someone else.

So what does that actually cause?

84% 84% of people are underrepresented in health AI data Visibility

AI does not see them

Decades of narrow research are now amplified by the models built on top of it, and those models decide how every disease is detected, dosed and treated, at planetary scale.

•Trial data skews Western
•Missed and late diagnoses
$5B $5B* lost across 15 drugs in five years Cost

Exclusion gets expensive

We traced 15 drug failures from the last five years: five billion dollars lost, most of it spent repeating trials whose evidence collapsed once the drug reached patients outside the original study group.

•Phase 2–3 failures, heavy losses for pharma
•Regulators now ask for proof

* Download the whitepaper to see the source and full method.

78% 78% of genetic studies use European-ancestry cohorts Why it matters

Biased data becomes a barrier

AI trained on narrow data produces biased outputs that do not generalize: money burned, clinical decisions that get real people wrong, and an eroding trust that snowballs, making every next attempt to bring AI into healthcare harder than the last.

•Wrong for patients, costly for pharma
•Each failure sets adoption back further
The answer · PAICON's vision

A data-first approach that covers everyone, by design.

Medicine was built on a narrow slice of humanity. We are rebuilding it for everyone.

Our mission

Redefine innovation through inclusion

Every model we build is validated across ancestries and geographies before it reaches a clinic, so its accuracy holds for the patient in front of you and not only for the population it was trained on.

Our approach

Representative data first,
then the model

We begin by building representative clinical and molecular datasets across regions, then train on them. Nothing is retrofitted with a correction factor once a model is already getting people wrong.

Our promise

Progress leaves no one behind

Equity is written into the pipeline, not appended as a report at the end. If a diagnostic underperforms for a population, we treat that as a defect and fix the data behind it.

Key opinion leaders · From ByteSight

Why representation matters, in their words.

“
More genetic diversity sits in Africa than in every other population combined, yet it is the least represented in the data.
” ▶ Listen to the episode
Prof. Segun Fatumo Prof. Segun FatumoProfessor & Chair of Genomic Diversity, QMUL
“
My cancer test came back a variant of unknown significance. It meant: I am Latina, and my genome is not in the databases.
” ▶ Listen to the episode
Dr. Catalina Lopez-Correa Dr. Catalina Lopez-CorreaChief Global Strategy Officer, Genome Canada
“
A drug proven in a white, middle-aged man may simply not work in a woman of African descent.
” ▶ Listen to the episode
Dr. Christian Tidona Dr. Christian TidonaCEO & Co-founder, BioMed X
Campaign 2025

A 4-part series on where exclusion begins, and what it is costing.

Go deeper

The Remaining 84 whitepaper

The full evidence base behind this page: how the 16% problem shows up across genomics, clinical trials and
AI training data, what it costs patients and pharma, and the data-first framework PAICON uses to close the gap.

The science behind it

Understand the evidence.

Representation in medical data is not an opinion, it is a documented scientific problem. These public, peer-reviewed and institutional resources explain why diverse data matters for everyone.

Science · 2019 Dissecting racial bias in an algorithm used to manage the health of populations Obermeyer, Powers, Vogeli & Mullainathan
A widely used clinical risk algorithm systematically underestimated the needs of Black patients, because it learned from biased historical data. Read the source ↗
Cell · 2019 The Missing Diversity in Human Genetic Studies Sirugo, Williams & Tishkoff
As of 2018, roughly 78% of genome-wide association study participants were of European ancestry, limiting how well genomic medicine generalizes to the rest of the world. Read the source ↗
JAMA · 2020 Geographic Distribution of US Cohorts Used to Train Deep Learning Algorithms Kaushal, Altman & Langlotz
Most US clinical deep-learning datasets came from just three states, leaving the other 47 with little or no representation in the data. Read the source ↗

Links open external, publicly available sources. PAICON is not affiliated with these publishers.

Let's connect

Join the movement. Help close the gap.

Joining is not a signature, it is data. The 84% only enters medicine when cohorts, trials and clinical archives are built to include the people they will treat.

Contribute a cohortBring an underrepresented population into a harmonised, research-ready dataset.
Co-design a studyPlan representation into your trial before recruitment, not after a failure.
Take the case to your boardUse the book and the whitepaper to make representation a funded priority.
Bring an idea, or a needHave an idea that could move Remaining 84 forward, or a data or insight need of your own? Come talk to us.
Book a demo
Send a message
Back to top ↑