The Aussie Man Who Used AI To Create A Cancer Cure For His Dog

**Ashlee Vance** (0:00)
Welcome to this special emergency edition of the Core Memory Podcast. This is Ashlee Vance, and we are, we're going remote today. We're not in studio because we've had to track down a man in Australia, a man and his dog.
Welcome Paul Conyngham to the show. Thank you for joining us.

**Paul Conyngham** (0:36)
Hi, Ash.

**Ashlee Vance** (0:37)
How you doing?

**Paul Conyngham** (0:41)
I'm, as mentioned, pretty tired. You're tired? It's been a hectic couple of days, yeah.

**Ashlee Vance** (0:46)
Yeah, as well. So if people haven't been paying attention, you're tired because I guess it was maybe US time last Friday, I think, or it could have been Saturday. A story came out in Australia about you using artificial intelligence technology to try to cure your dog's cancer. And this story caught the world's attention. And I suspect you've been in demand for the last couple of days.

**Paul Conyngham** (1:17)
Yeah, yeah. Yeah, I've been averaging about two to three hours sleep for the last three days.

**Ashlee Vance** (1:25)
Yeah, yeah. Well, okay, so many people will have seen, I think, some of the news stories and the coverage, but for those people who haven't, I apologize for making you repeat yourself. But do you mind giving us the basics and then we'll dive a bit deeper into the story?

**Paul Conyngham** (1:43)
Sure, sure, Ash. The Cliff Notes version is my dog, Rosie was misdiagnosed with cancer for about 11 months. So her original vet misdiagnosed the cancer. He said it was just a rash for 11 months. And then around the 11-month mark, I said, look, it was like bleeding through the skin like that. Can't be good. Let's remove it. Removed it, came back off to the biopsy, and it was cancer.
And then put it through a whole round of traditional treatments to try to bash it back. Because if you nip it in the boat early enough, with cancer, you can, for the most part, get rid of it. But unfortunately, Rosie had already passed that threshold. It had grown so far around her skin that if we removed that chunk of skin, she wouldn't have had enough skin left to close the wound. And that's essentially the trigger that I was like, okay, I'm going to try to see what I could do with my AI skills. I knew nothing about cancer at this stage, by the way. I was complete novice, just knew obviously that it's a very bad thing to have. And I'd heard about this. I've been practicing machine learning since about 2009 And through the community, obviously, followed AlphaGo and knew that the follow-on technology was AlphaFold, that a lot had been said about its ability to design cancer drugs.
So that's the path, essentially, that led me to going down this AI track.

**Ashlee Vance** (3:16)
Yeah. And I think Rosie is in the room, right?

**Paul Conyngham** (3:21)
She is.

**Ashlee Vance** (3:24)
There she is. Okay.
And so part of the story, I think, that caught everybody's attention was this idea of, like you said, somebody who is not coming from the biotech world, using these tools to create a personalized cancer therapy for their pet. And from everything I've read, you essentially did a sequence of Rosie's tumor, and you were able to match that against the gene sequence of her blood to begin looking for the variations between where her DNA should be and what had changed. And then you worked with some amazing, it sounds like scientists and doctors in Australia to then shape a custom mRNA vaccine. And so you're essentially delivering an injection into Rosie with this custom therapy that is then going to get her body to try to attack the tumor based on these DNA markers that you'd identified. Is that the gist of it?

**Paul Conyngham** (4:37)
Essentially. But just to backtrack a little bit for the audience that are listening in, when I essentially started down this path, I think I actually recorded the initial prompt somewhere. I should probably show that sometime. But I went into ChatGPT and just told it to list out a plan of how we can go through the steps of making some kind of drug or treatment to fight it. But also, before doing that, prior to that, I also used ChatGPT to build out a map. I used Obsidian, if you've ever used Obsidian, to build. It's like a mock-down note-taking tool. I went in and created an ontology, like the lay of the land of all the drugs that were out there, and what they do, and what their mechanisms are and stuff. I learned about this new, relatively new to me. I'm not an oncologist, so that's my disclaimer. I've been doing machine learning for a while, that's what I do. But yeah, I learned about immunotherapy. So the key thing about immunotherapy is that it's using your body's own immune system to fight the cancer. And to even backtrack this a little bit more, it's worth speaking a little bit about what cancer actually is. So if you rewind the clock all the way back to, I don't know, when life began type thing, there were single celled organisms floating around. And at some stage, they decided to become multi-celled organisms that work together to become animals and plants and fungi and everything. Cancer is when one of those cells in the organism sort of breaks away and starts doing its own thing.