Closing the data loop in AI-driven drug discovery
TL;DR
Drug discovery is a high-cost, high-risk endeavor that is under growing pressure from a market increasingly defined by first-mover advantage. Since the 1950s, the cost of developing new pharmaceuticals has roughly doubled every nine years—a phenomenon known as Eroom’s Law. Today, bringing a new drug to market takes an average of 10-15 years and costs….
Nauti's Take
The finding is useful because it does not blame the model: when development costs double roughly every nine years, the leverage sits in data from failed experiments — and that data currently goes nowhere. The limit is execution, since closed data loops rarely fail on technology but on ownership and formats between lab and development.
A strong argument for pharma teams, and poor material for quick success stories.
Briefingshow
The piece names a problem bigger models cannot fix: the most valuable data in pharma research comes from failed candidates, and those results are rarely captured or shared in structured form. As long as failures disappear into binders and isolated systems, models train on a flattering subset. That is less an AI question than one of incentives and data ownership across the industry.
Summary
Drug discovery is a high-cost, high-risk endeavor that is under growing pressure from a market increasingly defined by first-mover advantage. Since the 1950s, the cost of developing new pharmaceuticals has roughly doubled every nine years—a phenomenon known as Eroom’s Law.
Today, bringing a new drug to market takes an average of 10-15 years and costs…
Sources