The AI drug revolution is real but the hype around it isn’t

If you listen to the brightest minds in tech right now, you might think human disease is just a software bug waiting for a patch.

​At the 2025 World Economic Forum in Davos, Anthropic CEO Dario Amodei—drawing on his background in biophysics—predicted that AI could condense a century of biological progress into a single decade, potentially doubling human lifespans. Demis Hassabis, the Nobel laureate behind Google DeepMind, recently floated a similarly audacious timeline, suggesting that AI could help eliminate all diseases within 10 years. Hassabis aims to shrink the decade-long drug design process down to mere months.

​I’ve spent my career straddling the mathematical elegance of artificial intelligence and the grueling, messy reality of drug discovery. So, when I hear these predictions, I get it. Silicon Valley loves a moonshot.

But some of the rhetoric implies that one day we might treat the human body like software—diagnosing problems, simulating fixes, and “debugging” disease before it appears. That is not how biology works. The human body is an extraordinarily complex, adaptive system shaped by millions of years of evolution, and it does not behave like code running on a computer.

​WHY AI CAN’T DEBUG BIOLOGY

An algorithm can beautifully solve the 3D puzzle of designing a molecule to fit a protein. But no AI can magically compute away the chaos of a living human immune system or guarantee a molecule won’t trigger unpredictable liver toxicity once ingested.

​We’ve seen this friction between tech-sector optimism and clinical reality play out brutally over the last year. Veteran “techbio” pioneers faced a harsh reckoning when it became obvious that AI-discovered compounds still hit the same clinical roadblocks as traditional drugs. Following mid-stage clinical failures, companies like BenevolentAI went through massive restructuring. Recursion Pharmaceuticals quietly pruned several clinical-stage programs in a defensive pivot.

​The hard truth? Algorithms have not yet repealed the pharmaceutical industry’s punishing 90% clinical failure rate.

THE SKEPTICS ARE WRONG TOO

If you stop reading there, you might think the use of AI in drug discovery is just another hype cycle. You would be wrong. The revolution is happening—it just isn’t the overnight miracle the tech billionaires are selling.

​The claim that AI makes drug discovery cheaper and faster is entirely true, provided you know where to look. In the preclinical phase, AI is fundamentally rewriting the rulebook. For decades, the pharmaceutical industry has suffered from Eroom’s Law (literally Moore’s Law spelled backward), where discovering a viable drug candidate becomes slower and more expensive every year.

​Generative AI is finally shattering that trend. We are now compressing the traditional three-to-four-year marathon of finding a viable preclinical candidate into a 13-to-18-month sprint.

​More importantly, these early-stage candidates are proving to be of vastly higher quality. Today, AI-discovered drugs are passing Phase I clinical safety trials at a rate of 80% to 90%—nearly double the historical pharmaceutical benchmark.

FROM A SINGLE BREAKTHROUGH TO PHARMACEUTICAL SUPERINTELLIGENCE

​At my company, Insilico Medicine, we recently hit a milestone that moves this out of the realm of theory and into reality. As published in Nature Medicine, our AI-designed drug rentosertib delivered positive Phase IIa results for patients with idiopathic pulmonary fibrosis—a devastating, age-related lung disease.

​Why does this matter? Because it is a global first. This marks the first time a drug with both a novel biological target and a novel molecular structure—both discovered entirely by generative AI—has shown a measurable clinical efficacy signal, actively improving lung function in living patients. We took that drug from a blank screen to preclinical nomination in just 18 months, proving in the real world that AI can smash the earliest, most expensive bottlenecks of drug design.

​Now that we’ve broken the initial bottleneck, the next frontier isn’t just building better isolated predictive algorithms. It’s what we call “pharmaceutical superintelligence.”

​Working with researchers at Eli Lilly, we recently laid out the blueprint for this in ACS Central Science. Imagine a near future where a lead scientist simply types a prompt: Design a drug for this specific mutation of pancreatic cancer. A central AI controller then takes over, deploying specialized sub-agents to autonomously find the target, design the chemistry, and validate the biology in one seamless, continuous workflow.

FINAL THOUGHTS

​Will AI overwrite the laws of nature and eliminate all human disease in 10 years? No. It won’t magically bypass the years of rigorous human safety testing required to put a pill in a patient’s hand.

​But what AI is doing is transforming biological discovery from a slow, bespoke artisan craft into a highly scalable, compute-driven engine. The precision medicines of tomorrow will arrive years earlier, cost a fraction of what they do today, and save millions of lives in the process.

We don’t need to achieve immortality by 2035 to recognize that as a total gamechanger.

Alex Zhavoronkov, PhD, is founder and CEO of Insilico Medicine.