Key Takeaways
- Despite massive AI investment, drug discovery costs keep rising while output declines
- AI excels at pattern recognition but struggles with creative hypothesis generation
- Human intuition remains essential for breakthrough medical discoveries
Five years after AlphaFold’s celebrated protein-folding breakthrough, the promised flood of new AI-discovered drugs has failed to materialize. Despite billions in investment, artificial intelligence has not reversed the steady decline in pharmaceutical productivity known as Eroom’s Law.
The Drug Discovery Paradox
While computing power follows Moore’s Law—doubling every two years—drug discovery moves in the opposite direction. Eroom’s Law (Moore spelled backward) describes how the number of new drugs per billion dollars has fallen for decades. Today’s superior technology hasn’t reduced costs; it’s made drug development more expensive than in the 1970s.
Quantity vs Quality Problem
The fundamental challenge isn’t generating hypotheses but improving their quality. AI systems can produce billions of molecular ideas, but most lead nowhere. Algorithms multiply quantity but cannot enhance quality through intuition or imagination—this remains a human capability.
Why AlphaFold Succeeded Where Drug Discovery Fails
AlphaFold triumphed because it solved a bounded problem with clear rules and extensive training data. Scientists had already mapped over 150,000 protein structures through decades of work. The AI essentially aced a well-defined exam with a predictable syllabus.
Drug discovery, however, resembles spotting cricket talent in village grounds or predicting India’s next prime minister. There are no fixed patterns or reliable manuals—randomness dominates.
Historical Breakthroughs Came From Chaos
Most major drug discoveries emerged from accidents and messy experiments:
- Penicillin: Found when Alexander Fleming forgot to cover a petri dish
- Insulin: Discovered through chaotic dog experiments
- Paracetamol: Originated from laboratory misidentification
- Metformin: Initially studied for influenza before diabetes application
The Regulatory Reality
Modern ethical standards require stringent testing and multi-phase clinical trials. While essential for safety, this caution slows discovery. Even when AI proposes promising molecules, the path to prescription remains a marathon of paperwork and risk assessment.
AI’s Proper Role in Medicine
AI excels when humans provide clear questions and verification methods. It will continue transforming drug screening, clinical trial design, and repurposing existing medications. However, expecting AI to single-handedly create new cures misunderstands its limitations.
As history shows, from insulin to paracetamol, every medical breakthrough began with human minds willing to explore beyond existing data.
