Ten years and $2.5 billion — that’s what it takes, on average, to bring a new drug to market in the United States. Artificial intelligence promises to supercharge this process, drastically reducing the time and costs of bringing life-saving therapies to market and lowering patient costs.
I have spent nearly two decades in software and technology. I’ve been working with others toward the inflection point the world’s reaching — when advancements in technology and science mean that data science can finally keep up with the mounds of data that science produces.
What does that mean for the effort to create life-saving or quality-of-life-improving therapeutics? With the power to use massive, complex data, researchers can predict how drugs will interact, their toxicity and their potential inhibitions. Researchers can also identify potential successful compounds far more quickly and cost-effectively.
This isn’t just theoretical. Biotech startups such as Relay Therapeutics and Recursion Pharmaceuticals have reported success in clinical trials of drugs developed through AI-powered processes. These are first-in-human trials — having passed laboratory and animal studies. These drugs are now being offered to patients.
It’s thrilling to see the potential of AI becoming a reality. However, I can’t ignore the challenges it poses.
Every AI expert will have an opinion on the most pressing concerns. Here are just a few of the questions I’m thinking about:
Quality and accuracy are paramount for scientists from the earliest stages of drug discovery through human trials. With AI’s propensity to “hallucinate” now so well-documented in the various large language model systems, how will we ensure that insights that inform the development of real-world treatments are accurate?
The ethical considerations of AI are numerous. How do we harness the power of genetic data while protecting people against potential harm? For one example, imagine if health insurers could know — before considering coverage — if someone had specific gene signatures. They could decline coverage or make it more expensive based on information that a patient may not have consented to make available.
When will it be appropriate to remove the human from the loop — if ever? As industries like transportation march toward full autonomy, healthcare is rightly approaching with caution. Even companies building AI to diagnose conditions without physician input call their products an aide to physicians, not a replacement for them.
I don’t foresee a world where technology can operate without human ingenuity and creativity. Still, I do wonder what will happen when AI can handle tasks that used to fall to early career scientists and technicians.
Researchers are excited to leave behind the drudgery of data wrangling, analysis and annotation, but schooling needs to change to keep up.
Intellectual property is the lifeblood of the pharmaceutical industry. When AI is generating novel drug candidates, who owns that intellectual property? These questions are being played out right now among lawmakers. Can AI be responsible for patent infringement, and if not, what happens to the patent system? What will the answers to these questions mean for the incentives that underpin drug discovery?
Answering these questions will require stakeholders with conflicting interests and incentives to find common ground without ignoring the voices of the scientists these systems are meant to empower.
We stand today on the cusp of a radical revolution in how life-saving drugs are brought to market and used. If we approach this crossroad thoughtfully and strategically — if we work together to shine a light on concerns and implement the requisite safeguards — the AI revolution and the researchers and scientists using it will change scientific discovery for the better.