Did ChatGPT Really Cure a Dog’s Cancer? It’s Complicated

OpenAI co-founder Greg Brockman magnified an extensively shared story over the weekend about a pet dog treated with a tailored mRNA cancer vaccine established with assistance from ChatGPT, drawing attention throughout tech and AI neighborhoods.

The case centres on Rosie, a seven-year-old Shar Pei owned by Australian AI specialist Paul Conyngham.

According to posts distributing online, Rosie had actually been provided just months to live before getting the speculative treatment, which Conyngham stated was established with help from the AI chatbot.

” Back in 2022, I observed odd swellings on her head,” Conyngham composed in a November 2024 thread recording the journey from the start. “What the veterinarian considered as ‘simply warts’ wound up being late-stage cancer.” Veterinarians approximated Rosie had in between one and 6 months left and informed Conyngham there was absolutely nothing more they might do.

The account spread quickly after Brockman shared it with his numerous countless fans, triggering protection throughout a number of innovation outlets.

While the treatment itself appears real, the function ChatGPT is credited with in establishing the vaccine has actually been discussed, with some scientists questioning just how much of the procedure might reasonably be managed by a big language design.

Pressing ahead

Conyngham stated he didn’t quit on Rosie. Rather, he chose to develop a research study pipeline out of customer AI tools. He began with ChatGPT, utilizing it to create a master plan.

The design informed him he required genomic sequencing, one sample of healthy tissue and one from the growth, and pointed him towards particular organizations and devices.

” The most paradoxical thing is that in a previous chat session with ChatGPT, it stated that I must try to connect to Elita or Dr. Martin which I must utilize an Illumina device,” he composed at the time.

So he followed that lead.

A director at UNSW linked him to Dr. Martin Smith, head of the Ramaciotti Centre for Genomics, who accepted series Rosie’s genome for around $3,000.

10 days. Thirty-times depth in healthy tissue, 60-times in growth: the greater pass rate required to separate the anomalies driving the cancer. The Centre returned 320 gigabytes of raw information.

Genomic details is revealed in strings of the letters A, T, C, and G, so professionals basically wound up with a stack of 700,000 double-sided pages filled with just those 4 letters, the University of New South Wales reported in June of in 2015. That was Rosie’s genome, her biological finger print.

He then concentrated on c-KIT, a protein well-documented in the released literature on mast cell growths in pet dogs.

Utilizing Google’s AlphaFold, he designed Rosie’s variation of the protein and compared it versus the healthy standard. It looked incorrect, altered in manner ins which matched what the literature anticipated. He then looked for existing substances that may assault c-KIT or proteins comparable to it, and discovered one: a drug currently in usage in the U.S. to deal with a various cancer in human beings.

” We took her tumour, sequenced the DNA, we transformed it from tissue to information, and we utilized that to discover the issue in her DNA and after that establish a remedy based upon that,” Conyngham informed the Australian Today Program on Saturday. “ChatGPT helped throughout that whole procedure.”

AI’s real function

However, there’s a huge space in between ChatGPT discovering a remedy for cancer and ChatGPT helping in research study.

Conyngham ultimately gotten in touch with Prof. Palli Thordarson, Director of the UNSW RNA Institute. “Prof. @martinalexsmith carried out the DNA/RNA sequencing to transform Rosie’s tissue into raw information,” Conyngham published. “Prof @PalliThordarson put together the mRNA vaccine,” he included another tweet.

Thordarson validated this in his own thread: “Happy with @UNSWRNA to have actually been included & & making the mRNA-LNP for Rosie,” he composed on X on Sunday. “The crossway of RNA innovation, genomic & & AI presents a chance to alter the method we do medication and make gain access to more fair.”

However Dr. Smith wasn’t a male behind a ChatGPT screen. He was a teacher running a university RNA institute, doing what his laboratory was developed to do.

And when Conyngham recognized the last vaccine construct– the particular molecular plan that would be encoded into the mRNA– he exposed which tool created it. Not AlphaFold. Not ChatGPT. “The last vaccine construct for Rose was created by Grok.”

That stated, he acknowledged in a different post that “Gemini did a lots of the heavy lifting too.”

ChatGPT was utilized to sort through clinical documents and recognize scientists who may be able to assist. The chatbot indicated the Ramaciotti Centre and recommended sequencing devices fit to the job, working mostly as a tool for browsing the research study literature. That function can be beneficial, however it varies from creating a vaccine or carrying out clinical analysis.

AlphaFold, a deep-learning system from Google DeepMind, forecasts three-dimensional protein structures from amino acid series. It’s not the very first design trained on biological information: other open-source efforts like Ankh or AlphaGenome deal with comparable facilities.

Conyngham utilized Alphafold to design Rosie’s c-KIT protein. The rendering brought a self-confidence rating of 54.55, which UNSW structural biologist Dr. Kate Michie openly referred to as low.

She kept in mind that AlphaFold “can get things incorrect” which substantial laboratory work is required to confirm any output. Dr. Smith, the UNSW genomics director, validated openly in the exact same thread that AlphaFold was not, in reality, utilized for the mRNA vaccine style at all.

Dr. Thordarson bewared about the framing, too.

” This might not have actually treated Rosie,” he composed on X. “Bought time for sure, yes, however a few of the tumours didn’t react.”

His group is now inspecting whether those growths altered in a different way, which would discuss why parts of the treatment worked and others did not. The vaccine likewise did not operate in seclusion.

” The treatment needed co-administration of a checkpoint inhibitor,” Thordarson kept in mind, “most likely to be with all individualized cancer vaccines.”

Using AI for cancer treatment has actually not constantly been a history of success.

In 2017, internal IBM files exposed that Watson for Oncology, marketed as a system that might suggest cancer treatments much better than human oncologists, was producing what its own engineers flagged as “hazardous and inaccurate” suggestions.

MD Anderson Cancer Center deserted the job after investing $62 million on it. IBM sold Watson Health in its whole in 2022.

The Rosie case does not fall under the classification of AI failures. Nobody was damaged, the underlying science is developed, and the scientists included have actually acknowledged qualifications.

The mRNA platform itself is supported by scientific research study. The anxiousness lies more in how the story has actually been framed. When AI tools get credit for work performed by researchers and research study organizations, it can blur public understanding of what the innovation in fact does.

The scientists who carried out the sequencing, produced the vaccine, and handled the security procedures run the risk of fading into the background.

The episode provides a pointer that AI can help with jobs such as browsing clinical literature, however it stays far from changing the knowledge and facilities needed to create and produce medical treatments.