Medical AI isn’t the bottleneck to medical progress

Matthew Barnett, Tamay Besiroglu, Ege Erdil
August 4, 2025

Conventional wisdom in the AI field suggests that to advance medical progress, we should focus on applying AI directly to the medical field. People propose using AI to design new proteins, perform biological research, or identify disease targets from patient genomics. They suggest this is what will unlock the most gains to human health.

We disagree. In our view, the biggest barrier to medical progress right now has little to do with medical R&D. Instead, the problem is that our global economy is too small. Our world today simply lacks the infrastructure to support transformative, new medical innovation. When tackling Alzheimer’s and heart disease, the tools at our disposal are inadequate for the challenge.

To push human lifespans past 120, it’s not enough to accelerate drug discovery and make clinical trials faster or more efficient. Instead, we need revolutionary new technologies, like nanomedicine that repairs cells at the molecular level, comprehensive cellular reprogramming, or full tissue regeneration. But these technologies can’t exist without an economy that’s much larger to support them. We need entirely new industries that can supply advanced materials, manufacture at nanoscale precision, and distribute these innovations widely.

This industrial transformation, not medical research itself, is where AI will have the biggest impact.

The case of MRI machines is instructive. While they’re now a cornerstone of modern healthcare, MRI machines would have been impossible to develop and mass manufacture in 1925, no matter how much people spent on medical R&D. At its core, a contemporary MRI machine consists of a large donut-shaped magnet built from superconducting coils. To achieve superconductivity, these coils are cooled with liquid helium to temperatures near absolute zero. Yet until the mid 20th century, reliable helium liquefaction plants and widespread distribution networks simply didn’t exist. Moreover, MRI machines rely on extremely sensitive radiofrequency electronics to detect faint proton signals, a technology that was highly underdeveloped until the 1940s. Affordable computers capable of quickly performing Fast Fourier Transforms, without which MRI machines would be incapable of producing coherent images, didn’t come out until the 1970s. All of this means that, from the standpoint of someone in 1925, MRI machines required far more than laboratory research to build: they needed a larger and more advanced economy, with new supply chains, supporting industries, and specialized technologies.

Or take another example: antibiotics. While many assume that the key development that enabled widespread access to antibiotics was the discovery of penicillin in 1928 by Alexander Fleming, this is only partly true. In reality, antibiotics remained largely a laboratory curiosity until the 1940s when, as a result of massive WW2-related investments in industrial expansion, large-scale microbial fermentation suddenly became feasible. To culture enough penicillin to heal millions of sick patients, engineers directly leveraged expertise that had originally been developed to synthesize explosives, aviation fuel, and rubber at scale for the war effort. Without this expansion of general economic and logistical capacity, penicillin would have had a negligible impact on human health.

We think future medical technologies will look similar, with large impacts on human health only arriving after our civilization has the requisite economic infrastructure for supplying and coordinating the necessary inputs for realizing these technologies at scale. Building out this infrastructure will require progress on multiple fronts, but the key enabling force will be the deployment of highly agentic AIs that can substitute for labor across a wide range of work.

Creating AIs that simply perform medical research is necessary but insufficient. Future medical innovation will demand an AI workforce to handle engineering, quality control, administration, and construction of medical facilities and products. Yet what’s even more important is for AIs to build and manage the complementary industries that make new medical technologies possible. Just as MRI machines needed an existing network of helium liquefaction plants to become possible, medical technologies of the future will likely rely on external industries that supply advanced nanomaterials and the tools for precision biofabrication.

The interdependent nature of the economy suggests that future biomedical technologies will depend heavily on infrastructure originally developed for entirely different purposes. Refrigeration was originally developed to support the food industry, but later became a central part of distributing vaccines through cold chain logistics. Similarly, improvements in semiconductor lithography initially created to produce better microprocessors might later enable the precise fabrication of nanoscale sensors capable of detecting and repairing cellular damage directly within the human body. It is the broad deployment of AIs across the economy, rather than their narrow application in a specific domain, that will most rapidly accelerate the development of general-purpose infrastructure, enabling radical medical innovations.

The true potential for AI is not its use as a better medical researcher, but as a solution to our most fundamental economic constraint: the limited size of our global workforce. The future impact from AI will emerge from its dual nature: it scales like capital, yet functions like labor. By manufacturing trillions of GPUs, and employing AIs everywhere in the world, we will unlock a new economic regime where fundamentally new medical technologies become possible.

Yet scaling up the AI workforce is just one part of what’s needed. The other part of the equation is improving AI capabilities. Right now, AIs are too unreliable, too narrow, and struggle with long-horizon tasks. Simply scaling up current AI would not create the new economic conditions needed for transformative medical innovations: at most it would lead to incremental improvements across sectors.

To address this gap, Mechanize is obsessed with rapidly enhancing AI capabilities to accelerate economic growth. At the moment, we suspect the best opportunity for us is to build RL environments to aid in automating software engineering. Our current focus on automating software engineering isn’t driven by the belief that it will directly produce medical breakthroughs. Nonetheless, we consider this work important because of its broader, indirect impact.

Software is a multi-trillion dollar industry that’s deeply embedded in the modern economy. We anticipate that future AI SWEs will impact our world by dramatically expanding and improving our digital infrastructure. This digital revolution, in turn, will provide essential support for new industries, and more generally, economic growth: the real bottleneck to medical progress.

Want to help us build RL environments to initiate this economic explosion? We’re hiring software engineers.