Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

How generative AI is transforming life sciences

In life sciences, generative AI shows tremendous promise—and carries equally tremendous risks.

Man and woman looking at a computer screen

[Images: Adobe Stock/Gorodenkoff]

Profile photo of Travis Jones

Chief Operating Officer · Logic20/20

BY Travis Jones

AI is hardly a new concept, with origins dating back to Alan Turing’s work in the 1950s or even earlier. But in November 2022, a lightning bolt struck AI’s story arc and changed everything. OpenAI introduced generative AI tool ChatGPT to the public, and people couldn’t get enough: within two months, the app had topped 100 million monthly active users. 

Generative AI is also nothing new, having been a discrete area of AI research since 2014. What ChatGPT has done is deliver generative AI into the hands of average users. As a result, a surge of interest in uncovering new use cases began to swell across industries, and the life sciences sector is no exception.  

Now that the initial buzz has subsided, life sciences organizations are pondering the long-term ramifications of this new chapter in AI’s journey.

Is generative AI a game changer or a shiny object that may turn out to be more trouble than it’s worth?  

Generative AI technologies can drive much-needed efficiencies in pharma and other life science organizations—as long as they’re implemented prudently and with a keen eye toward medical ethics, regulatory compliance, and the trust and dignity of patients. 

THREE GENERATIVE AI USE CASES IN LIFE SCIENCES

Research

One of the most prominent illustrations of generative AI’s value in medical research is Google DeepMind’s AlphaFold, developed to solve the “protein folding problem” of how to figure out a protein’s shape based on its string of amino acids. As shape determines how a protein interacts with other molecules (such as those of therapeutic drugs), this understanding is critical for life science research. 

Legacy approaches to determining protein structures often consume years of research. Trained on the sequences and structures of 100,000 known proteins, the latest AlphaFold system can predict a protein’s shape “down to atomic accuracy” within minutes. 

This breakthrough is enabling the medical community to better understand diseases and develop drugs to fight them. For example, the Drugs for Neglected Diseases initiative (DNDi) is using AlphaFold to advance drug discovery for often-overlooked conditions that impact disadvantaged communities, such as Chagas disease and leishmaniasis. 

As John McGeehan, director of the Centre for Enzyme Innovation at the University of Portsmouth, affirmed, “What took us months and years to do, AlphaFold was able to do in a weekend.” 

Medical Imaging

Generative AI can create synthetic medical images for training machine learning models while preserving patient privacy. NVIDIA’s RadImageGAN, trained on a dataset of 1.3 million images, generates synthetic images reflecting a broad range of medical conditions, including stroke, heart attack, and cancer. 

Another “talent” of generative AI is improving the quality of medical images by reducing noise and sharpening contrast. In a recent study, researchers used the technology to integrate data from computed tomography (CT) and cone-beam computed tomography (CBCT) scans in measuring tumor shrinkage in lung cancer patients undergoing radiation. The sharper images give physicians a more accurate idea of how much tumors shrink from week to week, enabling them to plan the next radiation dose with greater precision while sparing healthy tissue. Results of the study showed an improvement in tumor shrinkage predictions, and radiation-induced pneumonitis and lung damage declined by up to 35%. 

Drug repurposing

In addition to supporting development of new drugs, generative AI is being used in repurposing existing therapies. These models can be trained with massive datasets of drugs and learn to rapidly emulate clinical trials in identifying new uses for existing medications. 

A research team at IBM, for example, recently used generative AI to uncover two new potential treatments for dementia that typically accompanies Parkinson’s disease: rasagiline (an existing Parkinson’s drug currently indicated for treating motor symptoms) and zolpidem (a widely used treatment for insomnia). Using de-identified electronic records from the IBM Watson Health network, the AI-driven analysis “unraveled therapeutic benefits of two drugs in decreasing the population-level incidence of dementia associated with Parkinson’s.” 

CHALLENGES AND BARRIERS TO WIDESPREAD USE IN LIFE SCIENCES

As inspiring as these use cases are, generative AI faces challenges on its journey to widespread use in life sciences. Medical data tends to be highly complex, access to data may be patchy, and regulations such as HIPAA and GDPR restrict its use. Many patients are also hesitant to trust “black box” technology with decisions about their medical care, and implementing generative AI will require considerable transparency if patient trust is to be preserved. 

Finally, and most critically, generative AI is not perfect. In December 2022, OpenAI CEO Sam Altman admitted, “ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness. It’s a mistake to be relying on it for anything important right now. It’s a preview of progress; we have lots of work to do on robustness and truthfulness.” In life sciences, where wrong answers can place lives at risk, this is no small consideration. 

BEST PRACTICES FOR IMPLEMENTING GENERATIVE AI IN LIFE SCIENCES

As with any technology, a methodical, collaborative approach to implementing generative AI is the best course of action. I recommend integrating these best practices in any exploration of generative AI use cases for life science organizations: 

  • Good data hygiene is essential. Develop and enforce firm quality control standards and establish validation protocols for AI-generated content. 
  • Consult regulatory bodies and legal experts during the development process to explore and address all potential compliance issues. 
  • Bring healthcare professionals, researchers, patients, and other stakeholders into the development process to ensure generative AI is implemented ethically and in alignment with the values of the organization. 
  • Establish human-in-the-loop oversight of AI-generated decisions. 
  • Build a governance committee with authority over generative AI–related processes and the ability to exercise that authority. 

Once again, we find ourselves living in “interesting times” regarding technology. In life sciences, generative AI shows tremendous promise—and carries equally tremendous risks. For organizations that seize the opportunity while also giving due attention to risk management and ethical considerations, the future can be bright indeed. 


Travis Jones is an entrepreneur, leader, and chief growth officer at Logic20/20, a business and technology consulting firm.