We had been informed AI would treatment illness in years. That’s a fallacy.
As we speak, over 90% of medical drug candidates nonetheless fail. It prices greater than $2 billion and over a decade to carry a single drug to market. And now the business faces a brutal paradox, which is the promise of exponential progress caught inside a system ruled by diminishing returns.
That is pharma’s model of Eroom’s Regulation, i.e. the inverse of Moore’s Regulation, the place the return on drug improvement declines as science advances. This disconnect has created the “valley of demise”: the chasm between discovery and medical validation, the place danger is excessive, capital is scarce, and most concepts go to die.
Pharma’s response? Keep away from early danger. Look ahead to efficacy alerts. Let small biotechs take the primary leap. The consequence: a damaged innovation mannequin, a biotech echo chamber, and too few new solutions for sufferers.
Fantasy #1: Pharma is aware of the right way to develop medicine higher.
Not true. Drug analysis and improvement stays probabilistic, messy, and context-specific. Massive Pharma excels at scaling, not all the time at discovering. There’s no secret playbook locked inside company vaults.
Fantasy #2: AI can develop medicine higher.
Additionally false, or, at greatest, wildly overstated. As we speak’s AI has helped with discovery, however not with improvement. It may recommend new molecules or predict binding, however it could’t but navigate the complicated, high-dimensional path to secure, efficient therapies.
The present wave of hype assumes that AI can inform drug improvement simply by studying scientific papers. However as AI pioneer Yann LeCun has argued: true intelligence doesn’t come from studying alone – it comes from interacting with the world and studying from expertise.
Drug improvement is deeply bodily. It spans chemistry, immunology, toxicology, pharmacokinetics, medical design, and real-world outcomes. Optimizing this requires greater than intelligent language fashions, it calls for techniques that may sense biology in movement.
Pharma has tried to bolt AI onto present pipelines. However with out coherent, standardized, longitudinal, multi-modal organic knowledge, AI can not motive. As an alternative of a world mannequin for biology, we get piecemeal instruments educated on static PDFs or siloed snapshots.
If a automotive breaks down, would you learn the logbook or look below the hood?
Even when giant pharma needed to construct this, they might hit three partitions:
1. Siloed knowledge: Fragmented throughout departments, trials, and distributors.
2. No longitudinal integration: Incompatible codecs throughout the preclinical-to-clinical arc.
3. Lack of engineering tradition: Constructing bodily AI requires sequencing platforms, Machine Studying pipelines, cloud infrastructure, and fast iteration: issues pharma doesn’t do natively.
Too usually, invaluable samples (e.g., blood, tissue) are collected however by no means become usable knowledge. Or they’re analyzed after which locked inside PDFs, inaccessible to any AI.
Pharma brings capital, regulatory expertise, and medical entry. Techbio startups carry the engineering muscle, velocity, and studying tradition.
This isn’t a build-vs-buy resolution. It’s a rethink-the-whole-system resolution. The one viable path ahead is partnership, the place each side decide to producing structured, machine-readable organic knowledge, tied to outcomes.
We are able to escape Eroom’s Regulation and cross “the valley of demise.” However provided that we engineer our means out as a substitute of counting on business methods.
To construct a real-world mannequin of human biology, we should:
- Seize organic and medical knowledge throughout the complete improvement arc
- Make that knowledge multi-modal, longitudinal, and interoperable
- Prepare fashions on biology itself, not simply literature
- Design collaborations that align incentives round studying, not simply licensing
That is how we cease hallucinating and begin reasoning. That is how we carry intelligence to biology. That is how we give sufferers higher well being outcomes and treatment illnesses.
About Noam Solomon
Noam Solomon is the co-founder and CEO of Immunai, a biotech startup utilizing single-cell genomics and machine studying to find and develop novel therapeutics that reprogram the immune system. Noam, who started his research at Tel Aviv College on the age of 14 and earned his bachelor’s diploma in laptop science by 19, continued his tutorial journey incomes two PhDs at MIT and Harvard. Whereas at MIT, Noam met his co-founder, Luis Voloch, and launched Immunai in 2018. Below Noam’s management, Immunai has partnered with prime pharmaceutical firms and tutorial establishments to enhance medical trial success charges and remedy effectiveness.
