When an Australian tech entrepreneur with no background in biology or medication mentioned ChatGPT helped save his canine from most cancers, the story unfold with the sort of validation Massive Tech has lengthy craved: proof that AI will revolutionize medication and tackle one among its deadliest illnesses. The truth, as typical, is extra difficult.
The model of the story that made the rounds on-line, first reported by The Australian, was comparatively simple. In 2024, Sydney-based Paul Conyngham realized that his canine Rosie had most cancers. Chemotherapy slowed the illness however didn’t shrink the tumors. After vets mentioned “nothing could possibly be executed” for the Staffordshire bull terrier-shar pei, Conyngham said “I took it upon myself to discover a treatment.”
Conyngham mentioned he used ChatGPT to brainstorm therapy concepts. The chatbot surfaced immunotherapy as an choice and pointed him towards consultants on the College of New South Wales, who then genetically profiled Rosie’s most cancers. He then used ChatGPT and Google’s protein structure AI model AlphaFold to assist make sense of the outcomes. With the assistance of UNSW professor Pall Thordarson, he pursued a personalized mRNA vaccine tailor-made to Rosie’s tumor mutations. Thordarson informed The Australian he thinks it’s the primary time such a therapy has been designed for a canine.
A couple of weeks after Rosie’s first injection final December, Conyngham mentioned her tumors had shrunk and he or she’s doing higher, even chasing rabbits within the park. They’ve not disappeared totally, although, and one tumor didn’t reply in any respect. “I’m below no phantasm that this can be a treatment, however I do consider this therapy has purchased Rosie considerably extra time and high quality of life,” Conyngham informed The Australian.
That nuance was misplaced because the story unfold. Newsweek ran the headline “Proprietor With No Medical Background Invents Treatment for Canine’s Terminal Most cancers,” whereas the New York Submit declared {that a} “Tech professional saves his dying canine through the use of ChatGPT to code a customized most cancers vaccine.” On social media, many accounts hyped Rosie’s case as a “treatment” and an indication a brand new period of customized medication had arrived. Some, notably OpenAI president and cofounder Greg Brockman, ought to have undoubtedly recognized higher, and others, like Google DeepMind CEO Demis Hassabis, did and shared it with out hype. Elon Musk joined in too, eager to level out that xAI’s Grok also played a part — a element that was absent from a lot of the unique protection.
The story additionally provides AI far an excessive amount of credit score. Not solely was Rosie not cured of most cancers, it’s not clear the mRNA vaccine was accountable for her enchancment. The customized therapy was administered alongside one other type of immunotherapy often known as a checkpoint inhibitor, designed to assist the immune system goal tumors, making it troublesome to know if the vaccine had any impact in any respect. One of many scientists concerned, Martin Smith, said the workforce is performing exams to test the immune response.
ChatGPT didn’t design or create Rosie’s therapy; human researchers did.
Nor was the vaccine itself generated by a chatbot. ChatGPT didn’t design or create Rosie’s therapy; human researchers did. At most, the chatbot served as a analysis assistant serving to Conyngham parse medical literature — spectacular, however a far cry from the breakthrough implied.
Stories are additionally obscure on AlphaFold’s function. David Ascher, a professor and director of biotechnology applications on the College of Queensland in Australia, informed The Verge that the mannequin “might contribute structural hypotheses about proteins, however it’s not a turnkey cancer-vaccine design system.” Official steering, he famous, additionally warns that AlphaFold is just not validated for predicting the consequences of some mutations and doesn’t mannequin “a number of biologically necessary contexts” both.
Grok’s contribution is even more durable to pin down. On X, Conyngham wrote that “the ultimate vaccine assemble for Rose was designed by Grok,” nevertheless it’s not clear what meaning in apply or what inputs the mannequin was given. Ascher mentioned Grok would realistically fall into a lot the identical class as ChatGPT: a device that “might assist with literature search, summarising papers, translating jargon, suggesting workflows, drafting code or paperwork, and serving to a consumer assume by choices.” A helpful function, however hardly what designing a most cancers vaccine suggests.
The “AI made this” framing ignores this large human effort, with out which the “AI’s output would have remained simply textual content on a display.”
All in all, Ascher mentioned Rosie’s case “is healthier seen as an uncommon, extremely particular proof of risk than a template strange individuals can readily reproduce.” It wanted “substantial” professional labor, he mentioned, “not only a chatbot and some prompts.”
That distinction is very necessary in medication, the place success relies upon not simply on producing believable data, however on the professional, bodily work of manufacturing, testing, and delivering precise therapy. Alvin Chan, an assistant professor on the Nanyang Technological College in Singapore who’s constructing AI for biomedical and drug discoveries, informed The Verge the “AI made this” framing ignores this large human effort, with out which the “AI’s output would have remained simply textual content on a display.” In Rosie’s case, AI is healthier understood as a device for sketching a blueprint than because the creator of the therapy itself.
The entire thing carries a faint whiff of a PR stunt that’s onerous to shake. Daring claims constructed from questionable foundations utilizing obscure strategies comfortably match contained in the world of tech fundraising. mRNA vaccines — very like the broader promise of customized medication — stay largely unproven as most cancers therapies in people, not to mention canines, and whereas the case could also be actual, it feels too tidy and conveniently glosses over the tens of 1000’s of {dollars} and important experience required to show the thought right into a viable therapy.
I reached out to Conyngham asking for a chat on X however haven’t acquired a response but. His profile says “Ending Most cancers for Canines” and links to a Google type describing his “dream to make this course of one thing everybody might have entry to.” The shape asks whether or not your canine has most cancers, whether or not you’re a researcher or scientist who needs to become involved, and whether or not you might be an investor.
I believe it might be a mistake to dismiss Rosie’s story as fully meaningless. AI might not be changing the lab anytime quickly, however it’s making science extra accessible to strange individuals. Nonetheless, that’s not the identical as making care extra accessible, and few sufferers — or pet house owners — have prepared entry to the world-class consultants, specialised gear, and substantial funds wanted to show that data into actual therapy.
