Pragmatic Optimism Needed as AI Advances in Drug Discovery, Experts Say

As artificial intelligence rapidly advances, some researchers warn of potential dangers and existential risks. However, a contrasting perspective emphasizes the technology's capacity for positive transformation, particularly in fields like drug discovery. "As the conversation around AI grows darker, we risk becoming so consumed by what might go wrong that we stop paying attention to what is beginning to go right," write Alex Karnal and Brian Kreiter in a recent analysis.

Drug development has historically been a challenging and lengthy process, with a high failure rate for new therapies. On average, about nine out of 10 drugs entering human testing do not succeed, and bringing a single medicine to market can take over a decade and cost billions of dollars. This traditional model often starts with a biological hypothesis, followed by the search for a molecule to affect it, and then extensive testing, which frequently reveals the initial hypothesis was incorrect.

AI is beginning to reshape this paradigm by accelerating the pace of scientific inquiry and discovery. Researchers are using AI to design novel antibodies against difficult medical targets and to rapidly move from initial concepts to demonstrating therapeutic capabilities. For instance, one team used an AI-driven experimental system to develop a new cancer-treating medicine in primates in under six months, a process that would conventionally involve years of work by multiple specialists.

The impact of AI extends to clinical trials and the development of personalized treatments. The recent announcement by Merck and Moderna regarding a personalized cancer vaccine for melanoma, which significantly delayed disease progression, highlights this potential. The authors suggest that growing clinical evidence, combined with AI tools and automated labs, can reduce uncertainty and lower the costs and timelines associated with pursuing new therapeutic approaches. This can potentially encourage more investment and competition in areas like cancer research, leading to a faster rate of promising ideas being tested.

Despite these advancements, the authors caution against complacency, noting that no AI-enabled drug program has yet received regulatory approval. Biology remains a complex field that continues to challenge even the most advanced technologies. Models can be flawed, drugs can fail, and patient responses can be unpredictable.

However, the emergence of AI in drug discovery is seen not as a final destination but as a crucial shift in direction. This improved ability to identify promising avenues for research is a significant source of optimism. It suggests that AI is beginning to provide better ways to navigate the complexities of fighting disease and improving human health, a field where progress has historically been incremental.

The authors conclude that the children of today will inherit a world shaped by AI, facing both anticipated disruptions and unforeseen challenges. They advocate for responsible leadership that acknowledges these risks while embracing the potential for progress. The story of AI's early successes in areas like understanding diseases, designing medicines, and tackling rare or complex conditions offers a hopeful glimpse into a future where the baseline of human health may be significantly improved.