ChatGPT and Dog Cancer: Between Hope and Complex Reality

A Captivating but Misleading Tale

An Australian tech entrepreneur recently made headlines by claiming that ChatGPT played a crucial role in treating his dog's cancer. This story, which quickly circulated in the media, fueled the notion that artificial intelligence could transform medicine. However, the reality behind these claims is much more nuanced.

The story begins with Paul Conyngham, a resident of Sydney, who discovered in 2024 that his dog, Rosie, was suffering from cancer. Despite chemotherapy that slowed the progression of the disease, veterinarians ultimately declared that there was nothing more to be done for the Staffordshire bull terrier-shar pei. Faced with this deadlock, Conyngham decided to seek alternative solutions.

The Role of ChatGPT and Experts

Conyngham turned to ChatGPT to explore treatment options. The chatbot suggested immunotherapy as a potential avenue and directed him to specialists at the University of New South Wales. These experts conducted genetic profiling of Rosie's cancer. Conyngham then used ChatGPT and Google's protein structure AI model, AlphaFold, to analyze the results. With the help of Professor Pall Thordarson from UNSW, a personalized mRNA-based vaccine was developed, targeting Rosie's specific mutations. Thordarson claimed it was likely the first application of such a treatment in a dog.

An Improvement, but Not a Cure

A few weeks after the first injection administered last December, Conyngham observed a reduction in Rosie's tumors, and she appeared to be in better shape, even chasing rabbits in the park. However, not all tumors responded the same way, and one remained unchanged. Conyngham clarified that this was not a cure but a significant improvement in Rosie's quality of life. This nuance has often been lost in media coverage, with some headlines even claiming that Conyngham had invented a cure for his dog's cancer through AI.

The Limits of AI in This Case

The story also tends to overstate the role of AI. Rosie was not cured of cancer, and it is uncertain whether the mRNA-based vaccine is the cause of her improvement. The treatment was administered alongside a checkpoint inhibitor, another form of immunotherapy, complicating the assessment of the vaccine's effectiveness. Martin Smith, one of the scientists involved, indicated that tests were ongoing to verify the immune response.

ChatGPT did not design or create Rosie's treatment. At best, it served as a research tool, helping Conyngham navigate the medical literature. This is impressive, but far from the medical breakthrough that some have suggested.

Tools, Not Miracle Solutions

Reports on the role of AlphaFold are also vague. David Ascher, a professor at the University of Queensland, explained that the model can help formulate hypotheses about protein structures but is not designed to create cancer vaccines. Official guidelines specify that AlphaFold is not validated to predict the effects of certain mutations and does not model several important biological contexts.

The contribution of Grok is even more uncertain. Conyngham mentioned on X that the final vaccine was designed by Grok, but did not specify how or with what data. Ascher compared Grok to ChatGPT: a tool that can assist in literature research and writing, but not a vaccine creator.

Essential Human Collaboration

The narrative "AI did this" overlooks the enormous human contribution without which "the AI output would have remained just text on a screen." Ascher emphasized that Rosie's case is a unique proof of concept, requiring considerable expert work, and not an easily reproducible model.

This distinction is crucial in medicine, where success relies on the combination of plausible information and expert work to produce, test, and administer real treatments. Alvin Chan, a professor at Nanyang Technological University, noted that AI is better understood as a tool for sketching a plan rather than as the creator of the treatment itself.

A Seductive but Misleading Story

The entire affair appears to be a well-orchestrated communication operation. Bold claims based on vague methods fit perfectly into the world of tech funding. mRNA vaccines, much like the promise of personalized medicine, remain largely unproven as cancer treatments in humans, let alone in dogs. While the story may be true, it conveniently omits the high costs and expertise required to turn an idea into a viable treatment.

Conyngham did not respond to my requests for comment on X. His profile states "Ending cancer in dogs" and links to a Google form describing his "dream of making this process accessible to all," asking if your dog has cancer, if you are a researcher, or an investor.

Although Rosie's story is not insignificant, it shows that AI will not replace the lab anytime soon. It makes science more accessible, but that does not mean that care is, and few patients or pet owners have access to the experts and resources needed to turn this information into real treatment.