Silicon Valley Promises AI Will Cure Disease. Drugmaker CEOs Aren't Buying It
Claims from Silicon Valley that AI will cure most diseases within a decade meet strong skepticism from people familiar with drug development. Commenters note that while AI already helps with tasks like target selection, protein structure prediction and trial design, the true bottlenecks are messy biology, slow and expensive clinical trials, and regulatory and financial incentives that don’t reward outright cures. Many see real potential for AI to improve efficiency and decision-making in pharma, but regard promises of near-term, sweeping medical breakthroughs as hype rather than grounded forecasts.
AI’s Role in Drug Discovery Today
- Commenters note pharma already uses AI/ML as a “co‑pilot”: target selection, candidate design, patient stratification, and analyzing complex data.
- Many argue this is valuable but far from “AI curing disease by itself.”
- Several stress that AI need not cure disease to be impactful; even modest improvements in candidate quality, trial design, or paperwork could matter.
Bottlenecks: Trials, Biology, and Complexity
- Repeated point: drug discovery is not the main bottleneck; clinical testing (6–10 years) and high failure rates (often cited ~80–90%) are.
- Biology is described as messy, non-deterministic, and multi-scale; we lack models that can reliably predict whole‑organism or population effects from molecules.
- Long‑term safety and variability across humans cannot be skipped, even with perfect simulations.
Hype, Timelines, and Mischaracterized Claims
- Several criticize claims that AI will cure “most human diseases in 5–10 years” as wildly ambitious or misleading.
- Some distinguish softer claims (AI “could speed up” medicine) from stronger ones (“cure most disease”), arguing they get conflated in media.
- Others compare this to past tech hype (crypto, NFTs), seeing another overblown promise.
Data, Simulation, and “AI-Shaped” Problems
- Lack of comprehensive, labeled clinical trial data (success and failure) is seen as a core limitation.
- Some view AI‑driven prediction of trial outcomes or better stopping rules as plausible, even at small improvements.
- Others argue input–output mappings in human biology are too complex to be an easy fit for current ML, unlike math or code.
Ethics, Incentives, and Bureaucracy
- Disagreement over whether bureaucracy mainly protects safety or needlessly slows progress; some invoke the idea that systems’ “purpose is what they do.”
- Skepticism about pharma incentives: claims that companies prefer chronic management over cures, and that regulation and profit shape which “solutions” are acceptable.
- Some hope AI might surface taboo or politically uncomfortable biomedical strategies (e.g., germline interventions), while others warn AI itself is shaped by politics, bias, and centralization.
Automation, Wet Labs, and Risk
- Interest in closed‑loop systems where AI designs experiments and automated labs execute them; seen as both promising and potentially dangerous (e.g., accelerated bio‑risk).
- Multiple commenters stress that no conceivable “fully automated lab” can remove the need for human trials and long-term observation.