Loading summary
Transcript481 lines
- [00:00]
A
Good news guys.
- [00:02]
B
You have some good news.
- [00:03]
A
You're invincible now is fixed all your health. That's name, name your problems. All of them.
- [00:08]
B
I guess I care too much. Is that cured?
- [00:12]
A
You'll be a sociopath.
- [00:13]
C
I work too hard.
- [00:15]
A
I just worked. You guys will be uncaring assholes, optimist robots going around.
- [00:21]
B
That is. That's brilliant.
- [00:22]
C
I've been looking forward to this.
- [00:25]
A
Look, I have another question for you guys and let's not be some podcast that just fucking. Just regurgitates the popular thing. Do you think cancer is good or bad?
- [00:34]
B
I think it's debatable.
- [00:36]
A
I would like us to at least steal, man. Because today I'm going to make an argument for how I can really help us in cancer treatment and prevention. But I don't want it to be one sided and we're just one wing of the spectrum. Okay.
- [00:48]
C
I think I'm kind of on that like raw milk, let the cancer grow sort of way.
- [00:53]
A
Yeah.
- [00:54]
C
It's like if it's, you know, it's natural. Have we thought about that?
- [00:57]
B
If the good Lord wanted me to fight my cancer, he wouldn't have given it to me. Okay, so that's true.
- [01:03]
A
He does give his.
- [01:05]
C
Hardest working, most caring soldiers his hardest answer.
- [01:11]
B
His most cancerous. Yeah. So, okay, there's a. There's cure. Well, that's a big bold claim.
- [01:18]
A
I actually going to make the argument that in our lifetimes will cure cancer because of AI.
- [01:21]
B
Oh, damn.
- [01:24]
A
It's that big. Dude, I'm that deep in the Kool Aid. Okay, for a week straight I've been jumping TechCrunch articles just like, yes, jamming it on the end of the super.
- [01:35]
C
Bowl over and over.
- [01:38]
A
I just have optimist robots in a revolving line dumping me with Gatorade.
- [01:43]
C
Well, before we hop into this, I think first things first, I think we're going to try something a little new this week. We wanted to try having one larger topic that we dive into a little deeper and, and see how that goes and then fit in a few smaller things on the tail end of the topic.
- [01:59]
B
Letting our topic talk for a bit, just letting us blow on things we've heard or seen throughout the week instead.
- [02:04]
C
Of we're going to start. I have been really excited about this topic because of all of the scrutinizing and negative things to be said about AI, a lot of which we've already talked about on the show. One of the things that has always excited me is the prospect of it improving just, just health for the average person. And life expectancy. Like the things that I imagined were AI helping you dig through research and data in a way that couldn't be dug through before. And that brings you to some sort of solution for a disease that we could never cure before. So those type of things I think.
- [02:38]
A
Should be we can give people new diseases or.
- [02:41]
C
And new ones, mix it up new.
- [02:45]
A
Like a patch notes.
- [02:49]
C
Here, stick. You're going to pop this neuralink in. It's going to give you malaria. But if you don't like that one, you can just toggle it. Your ability to experience all disease with none of the consequences.
- [02:59]
A
Well, it's like a Minecraft world. You can do a random seed or you get a predetermined, like the most popular trendy disease right now. You get that inserted into the brain.
- [03:08]
C
This is going to be this. Maybe this is like the rich, the weird rich people trend like 100 years from now. Do you know like the old, like older stories of in medieval times, like all rich people would like gorge on food and then like make themselves throw up so they can keep eating. This is going to be like that.
- [03:23]
B
100 thing to do is to wake up and get malaria for six hours.
- [03:28]
C
My morning routine, my malaria chip.
- [03:31]
A
10 minutes of malaria, 15 minutes of Alzheimer's. Then I go get three treatments of chemotherapy. So it'll be.
- [03:41]
B
Yeah, I'm excited.
- [03:42]
A
It'd be like a cool trip. Yeah. Damn, it's so sick. All right, well, that's the world I want to move to. So the way I'm going to break this down. So we're going to talk about AI in health care. I'm going to talk about what AI is currently doing in health care right now. That's chapter one. Then we're going to talk about what AI is probably going to be doing in the next couple of years. The exciting kind of ventures that are currently being developed that look really, really promising. And then chapter three is going to be, AI is going to fix all of our problems forever and we're going to have no more disease. So is this in, by the way? That's in 10 years. In 10 years guarantee.
- [04:12]
B
That's a Doug.
- [04:13]
A
That's a Doug.
- [04:13]
C
Doug.
- [04:14]
A
Probably ye.
- [04:16]
B
If you are sick in 2035, you.
- [04:19]
A
Are come to me get a refund for this, for this podcast.
- [04:24]
B
So, okay, my first question is this. When you say AI and healthcare solving all our problems, my understanding is that doctors are going to go to chat GPT. They're going to type in how fix cancer.
- [04:34]
A
Yeah, yeah. Okay.
- [04:34]
B
And it's Print out.
- [04:35]
A
This is a good thing. All right, so. So there's actually a lot to. To cover here, and hopefully I can make a compelling argument, both setting up the groundwork for, like, why this stuff matters and how healthcare works, and then how it can help. So it's very important to recognize right now everybody uses the term AI to mean everything. In the same way that you're like, oh, the Internet is good or bad. The Internet has a million different things. Right? So AI can mean many different things. We're not talking about ChatGPT in this podcast, really, at all. ChatGPT is one product with AI where you talk to a chatbot and you can get information, and that's cool and great. I imagine that there will be an increasing amount of, like, hospitals who have a ChatGPT agent that, like, screens you for certain things, like primary care. Maybe that would help primary care physicians, you know, alleviate some amount of workload, for example. But what we're talking about here is machine learning. That's fundamentally what AI is. So ChatGPT is a subset of AI. More broadly, we're talking about machine learning, which is basically the ability to get a network, a software product, to analyze a whole bunch of data and learn from it and then solve any problem. So the idea is very much like, can we create systems that can learn to solve anything really, really quickly? Not can we talk to ChatGPT and ask it for questions?
- [05:45]
B
Okay.
- [05:47]
A
Because that is not a good. So, yeah. And if you hear, like, right now, there's a lot of dumb things going on with AI, and in fact, I want to pull up. If you could pull this up on the thing, Perry, we are in the hype cycle. Okay?
- [05:59]
B
I love this chart.
- [05:59]
A
This chart, right? So the innovation is triggered, Right? And so the expectations are really inflated right now, and everybody's going, oh, my God, AI Incredible. Everything. And it's amazing. And so we're kind of at the peak of inflated expectations right now, where everybody's jamming AI into absolutely everything, and we're rapidly heading towards the trough of disillusionment, which a lot of you guys are already at, probably.
- [06:19]
B
I was going to say the investors are at the peak of inflated investors. Everybody else is, like, the commenters are dropping.
- [06:25]
A
Yeah, yeah.
- [06:26]
C
This week, I'll say in my car. I asked Apple Intelligence, like the new replacement for Siri, how to say a certain word in Swedish, and it couldn't handle that request, which didn't, you know, didn't leave.
- [06:40]
B
I've heard bad things about deep in.
- [06:42]
A
The trough, because, like, 15 minutes earlier you were up on that peak and the expectations were inflated. This is. I actually, I chatted with my sister about this as well. She is a registered nurse, so works in health care and primary care. I was talking about how some company that they work with was like, we're so happy to announce that we've introduced like a chat bot that can summarize your patient information. Or I was like, this can like reply, give automated replies to people who, to patients who reach out with you. And it's just the most like unemotional, awful thing. Like somebody's opening up about this trauma and their dog died and all this stuff. And it's like, wow, so glad to hear that you should take this medicine. So, as I'm sure everybody has seen, there's a lot of people right now, the most hype cycle thing is that every CEO in America is like, oh, I get stock to go up.
- [07:29]
B
And so they say 10 times.
- [07:31]
A
They say 10 times. And they literally. And you just got to watch the money go, yes.
- [07:35]
B
And then they're going to keep saying it.
- [07:37]
A
So I want to say that up front where I'm not claiming, and nobody's claiming that chat GPT is going to go solve cancer or anything like that. But what is really valuable is deep learning, which is again, the broader concept of AI, which is where you train a system to, to basically solve any problem, particularly specialized problems. So let's start with cancer. Let's talk about what Cancer.
- [07:56]
B
Always a good opener. Yeah, I do. At dinner parties or what you guys hear about cancer, about cancer. Let's open a little cancer discussion.
- [08:04]
A
So cancer, the reason cancer is so hard. And I had a family member who, past couple years who's been, who was dramatically affected by this. So I've been so partially focused on that for that reason. But also cancer, obviously extremely big deal. The reason cancer is hard to cure is because cancer cells are your body's own cells. And the DNA is mutated to where the cell now thinks it should just be growing and expanding constantly and not dying. So that's where a tumor comes from. It is the cells expanding repeatedly. And they're programmed to, to think that's what they're supposed to do. And they just keep expanding. They try to not die. And this goes on, on and on and on. And then that causes problems not only because you have a physical thing in your body somewhere, but also because it's growing so rapidly, it's consuming all this energy, potentially releasing enzymes that are causing all these different problems. So cancer in Particular is not like other. Let's say, you know, a bacteria where it's like this thing comes into your body. Cancer is your own cells that are just mutated and they're doing the wrong thing. And that makes it very, very hard for your body to correctly identify. This is a bad cell that is behaving badly because on the surface, it kind of looks. Well, this looks legit. This is. This is a totally legitimate.
- [09:13]
B
This is ours. Yeah, this one's ours. Yo, let it go.
- [09:16]
A
Yo, we're chill. There's. There's more complexity to this. Oh, yeah. By the way, disclaimer. I am not a biochemist, so as much as I typically claim to know everything about everything, there might be slight, broad generalizations here.
- [09:29]
B
You're telling us before the pod that, like, biochem or whatever was the class you wanted to.
- [09:33]
A
Your parents failed out of biology in high school. I got. That was the only grade I ever got where I got like a D minus on a test. And that was right when World of Warcraft released Blackwing Lair in 2006. Or seven. And that's the raid. It's a 40 man raid. You got to show up a couple hours every night. And I was not learning biology. So this is like the core problem with cancer. And so we basically have three ways to treat it right now. One is chemotherapy. Chemotherapy is where you put a drug in your body that is specifically meant to attack fast growing cells. Because one of the hallmarks of cancer, like, its whole thing is it just keeps growing and it doesn't die. Right. So if you get a drug that is able to identify cells that grow really fast, then that drug can go in and be like, aha, you're growing way too fast. You're a tumor cell. Kill it.
- [10:21]
B
Yeah, yeah.
- [10:22]
A
The problem is there's other parts of our body that grow really fast. For example, bone marrow. Ooh. And you do want that. Like, bone marrow is actually helpful. Also your hair. I mean, that's like, chemo has worked a lot on you. And then.
- [10:35]
C
And chemo is radiation.
- [10:36]
B
Did you say chemo has worked on.
- [10:39]
C
Radiation is different than chemo?
- [10:41]
A
Yes, we'll get to that.
- [10:41]
C
Right, right.
- [10:42]
A
I was attempting a bald joke, and I don't know if it really landed, but I think the audience is so desperate and hungry for bald jokes, they'll be like, nice one, Doug.
- [10:50]
B
I'm particularly bald today. You're actually a little.
- [10:52]
A
You really got him. That was clever. Way to work it in. So chemotherapy is like one of the main three ways we treat cancer. This is. You basically are giving yourself poison that attacks fast growing cells and you do a bunch of collateral damage in the process because you're attacking, you're killing off a bunch of your own fast growing cells.
- [11:09]
C
Yeah. My understanding is chemo is basically. It's, it's attacking your. Your like hedging that your body can tolerate the damage done to the rest of its systems while the chemo destroys the cancer.
- [11:19]
B
Yes.
- [11:20]
A
It's like you're pointing. It's like the end of Fight Club. You're like pointing a gun at yourself and you'll be like, I'll pull the trigger. I'll fucking do it, man.
- [11:26]
B
It's badass.
- [11:27]
A
And so, yeah, no, it's, it's bad. And so for people who have had, you know, friends or family who've gone through chemotherapy, the reason you feel like fucking terrible the whole time, why it's brutal, is like your stomach lining, for example, fast growing cells. So if you're killing your stomach lining, you're going to get nauseous and inflammation and indigestion and all these horrible things. Your bone marrow, which is helping, you know, run your immune system that's getting shot to all hell. So. So there's all these negatives and you can't even. Every person is kind of different with how tumors show up, which we'll get to later. So you don't even know the effect that a chemo is going to have unless you start it. So chemo, it's like one of our best things right now that we have. The second is radiation. Radiation is where you are applying basically high energy at cells. So most often what this means is you take like a gun and you shoot electromagnetic radiation at cells. And this is bad for cells.
- [12:17]
B
I can tell this is an American podcast because all of our metaphors are bad about how different treatments are like different types of gun.
- [12:25]
A
So chemotherapy is like a lot of little guns in your body that come around shooters.
- [12:32]
C
Your body can only take so much.
- [12:34]
A
Burger.
- [12:36]
C
Gets rid of the cat. Okay, okay.
- [12:40]
A
So radiation is a little different. What you're doing is. I mean, you can literally think of it as a gun kind of you are shooting a high energy beam of electromagnetic waves at a target. And radiation is not good for cells. It kills cells. Basically it fucks up the DNA in a cell and that makes it either kill itself or it will over time die. So radiation in the body is super bad. In the same way shooting yourself in the chest is bad. I've heard it is not good for you, but in theory, let's say you could point a gun just at the cancer tumor and just shoot that. Well, that actually could be great. So radiation, which is a thing you do not want in your body, if we kind of like have a narrow beam that we target directly at the tumor, can hit just the tumor, that actually can help kill off a bunch of those cells and prevent it from growing to the same extent. So that's radiation, as you can imagine. Also not ideal. We are shooting radiation beams into the body and just trying to avoid shooting all the other stuff. And then the third is surgery, which is if you are able to, in some cases, you just cut the tumor out. But that is basically like, you know, you have to get to a point where the tumor hasn't spread. If it's spread around the body, called metastasizing, you're kind of fucked at that point because it's all over the place and it's already spreading. And it's more about, like, a broad treatment, like chemo, that's going to have the best chance of attacking it, versus surgery, which is a very specific thing. So there's a lot of people who get, you know, cancer, for example, let's say pancreatic cancer, and maybe they're able to get a surgery and cut that out of the pancreas. Causes a whole lot of collateral damage. And if you're lucky, you got it before the cancer spread.
- [14:13]
B
I see. Yep.
- [14:14]
A
Yep. So if cancer spreads, it's really bad. And there's. Okay, quick. More terminology, and then we'll get to the kind of, like, treatment of it. So you have all these scans to determine if somebody has cancer. So you've probably heard of things like X rays, CT scans, MRIs, biopsies, ultrasounds. Basically, there's different categories of how we scan the human body, see what's in there, and then doctors are able to go, okay, this or this or this thing is happening.
- [14:37]
B
I watch House, and I'm kind of an expert.
- [14:39]
A
Yeah, the thing is, with cancer, it could be lupus, but that doesn't explain the headaches. It's so something like 99% of the time, it is lupus.
- [14:50]
B
That's what I've learned.
- [14:51]
A
Yeah, I'm kind of cancer, specifically, lupus is the key. So what we're going to keep coming back to is radiation. Radiation is, again, it's like really, really high energy waves like X rays are literally shooting protons, and that kills cells. So you don't want the body to get a lot of radiation. But we can use it to learn what's going on in the body, or to kill cells if we want to. So an X ray. If you go get an X ray, you're shooting radiation into your body. And basically the energy as it goes through you see that it is being absorbed by certain parts of the body, like a bone or whatever. And that allows you to see if a bone is broken.
- [15:22]
C
This ties my. My. My closest friend from college, he's studying to become a medical physicist right now. And I didn't know this was a job a year ago. I didn't know this existed. And they create. Their position is to create, like, the radioactive isotopes that these sorts of machines and treatments use. Yeah, and there's a really limited of these, like, residencies and positions for this role in the country. And he's studying that right now at university as well. I never thought about, like, they're. They're not like the X ray technician. They're not the person, like, performing the actual treatment. They're creating the radiator radioactive isotope that the treatments or the machines use to do these things.
- [16:01]
A
Things keep it grounded. He's making a gun.
- [16:02]
C
He's basically making a gun.
- [16:04]
A
He's making a gun.
- [16:04]
B
So he's trying to find safer ones, I assume. My understanding is that, you know, an X ray is a lower dose of radiation, but, like, if you do a bunch of X rays, that's also. That's. It's very damaging to your body.
- [16:13]
A
So this is the whole thing that it gets back to. So there's all these different ways of scanning your body. An X ray, we've all done that, right. It shows you bones. But if you want to get more detail, you do a CT scan. A CT scan is where you go in and your body, to put it simply, you get a bunch of X rays. So it's doing X rays on many, many different, different planes of your body. So you're getting a shitload. And from that, you can kind of make a map of the body. You get more detail. But the problem is you just got a shitload of X rays. That's a bunch of radiation in the body, which, again, is not good. That's gonna damage healthy cells. Radiation also, ironically, can create cancer cells. Because if you're messing up the DNA in cells, that can cause them to become cancerous. So the. One of the most powerful things we have, radiation, to learn what's going on in the body and get rid of cancer, also can create cancer if you're not careful.
- [16:56]
B
Is there a. This is how most superheroes are made. So is there a chance that you could get.
- [17:02]
A
So I don't know if you brought a lot. Like, what is it?
- [17:05]
C
Yeah. What's chat GPT got to say about that?
- [17:07]
B
About, like. Oh, we.
- [17:09]
A
So we ran a test to basically see if that would work on anybody. It was called Chernobyl. And we're still waiting.
- [17:14]
B
Still waiting for a Russian superhero.
- [17:15]
C
Right.
- [17:17]
A
And so far, I haven't really given.
- [17:18]
C
That enough time to stew.
- [17:20]
B
Ovechkin.
- [17:22]
A
And what's the. Who's the karate guy? He came out of Russia. Stevenson. Steven Seagal. Steven Seagal, also from your theory is.
- [17:32]
B
That he had a Chernobyl radiation, that.
- [17:34]
A
He was one of the test subjects.
- [17:35]
C
This is good. This is good.
- [17:36]
B
I didn't know that. That's a fact, though.
- [17:37]
A
Yeah, but. But most of the test subjects, again, do die because radiation is bad for you, and it's not good. All right, so, yeah. So the. The kind of main things that are relevant here for this conversation is CT scans. These are extremely common. It's basically a ton of X rays. You are exposing your body to a bunch of radiation. So you're trying to do this as little as you can while getting a lot of information. And a CT scan can actually show, okay, there's most likely a cancerous tumor there. Then there's MRIs, magnetic rib resonance imaging. Anyway, magnets. You get into an MRI machine and fucking pulls all your atoms with magnets, like, to your. To the edge of the cells, and then it whips them all backwards, and it uses that to figure out your body. It's crazy. So an MRI is actually even more accurate so you can see in more detail what the body looks like. In fact, let me show a CT scan because we're going to get into this. If you can pull up this. Perry. So a CT scan looks something like this, right? So this is like a slice of a ct, where it's basically one, you know, one image of your body that's like, okay, this is what's there.
- [18:37]
B
Okay.
- [18:38]
A
So. And you might have seen this with, like, an X ray or something like that. It kind of. It kind of looks like that. An MRI is even better at particularly soft tissue, meaning a tumor or something like that rather than bones, because the tumor is just, again, a mass of these cells that keep growing, or at least a malignant tumor is. And so an mri. What's really cool about an MRI is there's no radiation. You get a person into an MRI machine, and it's crazy, and it's kind of scary and claustrophobic. And it's expensive and it's slower and it's harder to do, but there isn't radiation. So MRIs are like, great.
- [19:09]
B
I did not know that. That is cool.
- [19:11]
A
Yeah. So radiation. So radiologists, including the one I spoke to for this is like, MRIs are kind of ideal. CT scans actually have less detail and put the body through a bunch of radiation, but it's easier to get people through. Cheaper and easier, cheaper, easier. So then if you are, for example, like a hospital or the NIH in Britain, it does actually matter. Right. If you can get through twice as many patients with a CT scan versus everybody get its MRIs, then that's better. Or often what happens, you get a CT scan and then you get an MRI for additional information. There's other supplementary things like a PET scan. So that allows you. I didn't know. You do a CT PET scan, which is where you give the body radioactive sugar and then that gets eaten by the cancer. Because again, the cancer is growing so rapidly that it's draining your body of resources. And so you can kind of use it to find where the cancer is. But then the ultimate thing is you do a biopsy. This is where you take a physical sample of the tumor and scientists look at it under a microscope and go, yes, this is cancer. So in general, the simple way to think about this, there's kind of three levels to this. If you, if somebody thinks you might have cancer, you generally get a CT scan. So this is a whole bunch of radiation, but it's going to show is, does it look like there's tumors in your body somewhere? You might additionally get an mri, which is going to be more detailed, but again, it's slower, more expensive, but it gives you more information. And then the ultimate you do or do not have cancer is a biopsy. That's where they grab the actual physical cells and look at them.
- [20:35]
B
They have to find it first. They have to. Have to do a different scan.
- [20:38]
A
Yeah. So you don't start with a biopsy because that would mean they're just randomly. What if you just suck and stuff?
- [20:42]
B
I just feel it and I tell the doctor, like, biopsy right here or whatever.
- [20:45]
A
Yeah. So you could. And that's what an AI will do. It will go just touch you every morning and look for lumps and just be like, biopsy.
- [20:54]
C
Your robot assistant will give you a pat down and be like, this is like biopsy.
- [20:58]
A
And you're like, oh, your boobs are feeling a bit big. And you're like, no, I've just been Gaining weight. It's like no biopsy time. And it just starts with a turkey baster.
- [21:05]
B
Grabbing cells would not want Elon Musk, optimist, to decide whether my boobs are too big. I need to biopsy, but we don't want to.
- [21:13]
A
You're supporting cancer today.
- [21:14]
B
All right, Broke cancer should be in the village, guys.
- [21:16]
A
Be sure to leave a comment. Who do you think won the debate? Pro cancer or anti cancer today? All right, so this is the groundwork for all the stuff we're going to talk about. Cancer. Fast growing cells. We have three. We have these different ways of scanning it and then we have a couple. Not great, but they sort of work and help people ways of treating it, which is chemotherapy, radiation, surgery, all with this weird dynamic where radiation is super bad for you, it kills cells and it can cause more cancer, but we can use it to kill bad stuff and to learn about it. Right, so understood. That is the whole dynamic. So let's focus in on. Imagine you are crafted robot from our community who is a trainee clinical science at the NHS in the uk. Nhs? I said nis.
- [21:56]
C
Nih.
- [21:57]
A
National Health Care Service in the us. In the uk. For people who are not aware UK has a national health care system. So basically this is somebody who is involved with the treatment of cancer patients with radiation. So this is a person who is actually. You have a cancer patient who comes in, they're like, I have prostate cancer. And he is part of the team that's going to make sure the machinery works. Kind of like what you're saying, the machinery works. They're going to use the information from the scans to determine like, what does this person look like. They're going to figure out a plan of where to shoot the radiation gun, which again is literally a machine that's going to shoot a beam of like deadly laser into your body and then, and then, you know, basically follow through with that treatment plan. So the process is basically what I just described. You have a patient come in, you have some degree of scanning and the first thing you need to do, you have some degree of like information from the previous scans that determined that the person even has cancer. And the first thing you need to do is create an outline or a contour of their body because each person is different and you need to know if you're going to be shooting radiation into their body. You got to know exactly where the tumor is and like where their organs are. Right. They're a unique combination of people.
- [23:03]
B
That sounds important.
- [23:04]
A
Yeah. So what that means in practice for somebody like Kraft is that he looks at hundreds of images like these because again, a CT scan is usually just like many, many X rays essentially. And he has to combine these together to make a map of the person's body that then they will go, okay, we can shoot radiation from this angle and with this intensity in order to get the maximum impact.
- [23:25]
B
Okay.
- [23:25]
A
But the first step is this auto contouring or contouring where you are taking these, these X rays basically of the human body and turning them into something like this. Like zoom in on. Oh, okay, there we go. So you can see something like this is you're actually mapping specifically how this person's organs are arranged in very precise detail. And this matters again a lot if you're going to shoot radiation into their body.
- [23:49]
B
This looks like in a video game if your lungs were like more damaged than your.
- [23:53]
A
Yeah, no, it looks like when you go into like the stealth mode and you can see the enemies. So this is a, let's say average three hour process to take all of these X ray images from the CT scan, the CT images and turn them into this outline of the body. And this is like one of the main steps that's needed in order to get a treatment plan for this patient. This is really hard and it takes a lot of time. If you think about treating thousands of cancer patients, that is a huge bottleneck of needing a trained, you know, radiotherapy oncologist technician who is able to make one of these plans using these images. Not only is it hard, it's actually kind of subject subjective. You have to make educated guesses about what goes where. It turns out a lot of parts of the body are near each other. If you're trying to target the prostate, for example, there's, you know, there's intestines all over the place. It's not always clear what is where. Right. And so you making this map of the human body. I think one of the things I wanted to express here is that I think there's an expectation or kind of feeling that doctors are all great and we need to make sure we match that. But doctors are actually subjective in a lot of ways. You're often. I talk to doctors for this.
- [25:04]
B
They're human.
- [25:05]
A
One, they're human. And two, so much of what's going on in medicine is like, you're like a detective, right? You get all these clues and you're trying to figure it out. Right. It's not like you don't just get to check a box and be like, yes, it is that. There's maybe like 50% of stuff where you're like, okay, I'm using my best guess to figure this out. And anybody with a chronic disease or anything can attest to this, where, you know, different doctors will say different things. So this is an example of one of many elements of medicine where there's a subjective element to this, where you're trying your best, but the technician comes up with this outline over the course of three hours. Then the doctor will come in, confirm it's all good, and then they say, okay, now let's come up with the treatment plan to shoot this person with radiation. Okay, so this whole process, right, if a person feels like they maybe have cancer, they go into the hospital, they get scanned with a CT scan or an mri, they get this imaging and a doctor will say, okay, it looks like you have cancer. They then get a biopsy to confirm that it is 100% cancer. At this point, maybe they're recommended for radiation. They go into the radiation team, which is where like crafted robot would be participating in this. They use the scans to build an image of their body. From there they build a plan to shoot radiation into the body. And then over, let's say 20 sessions, it varies, this person is coming in, laying on a table and being shot with radiation in the exact way that they're reporting how they're.
- [26:16]
B
And they're checking on it and seeing if they.
- [26:18]
A
Okay, so this is an example right now where this company, this is a sperm bank in Germany called Siemens, is using GI right now to speed up this process.
- [26:30]
B
He's joking, to be clear.
- [26:33]
A
Well, I don't speak German, so I. If you imagine there are thousands and thousands of cancer patients coming in every, you know, every week, every month, whatever it is. If you are the entire country of the UK or any part, any healthcare, wherever you want, you have all these cancer patients coming in. There's two reasons why the speed of this matters. One, if you cut out a three hour process that a technician has to do by hand, that is a crazy speed up. And right now this program, Siemens, is being used in the UK by the NHS and is taking that three hour process and making it 10 minutes. That is a crazy reduction. Not only does that mean a technician is now has way more resources to spend on additional patients or focusing on reviewing the, let's say, you know, 95% of this that's done and they make sure that it all looks good and check it off and everything. Not only that, as a reminder, it's really fucking important to get cancer patients into the Hospital. As time goes on, the cancer is growing and potentially spreading. So if you get somebody into like it's a very common story that you get diagnosed with cancer and you can't start treatment for weeks, sometimes you can't get a scan for weeks. Right. And this is terrifying if you're like, there is maybe a malignant tumor that might spread and make me ineligible for surgery and kill me. Every day matters. And so if you imagine that this kind of process gets us from, we can see, you know, let's say 100 people a week to 150 or 200, or you get patients in a week earlier than you would have otherwise. This is potentially life saving stuff. This is a big deal.
- [27:59]
B
Just a question process. As someone who also played World of Warcraft in high school. So, okay, let me just simple question. What you're describing is it is taking the images from the CT scan or the mri.
- [28:11]
A
Yes.
- [28:12]
B
And instead of a human putting them all together into a composite, it is using machine learning on a million composites.
- [28:17]
A
Yes.
- [28:18]
B
To determine what the likely best combination is.
- [28:21]
A
Yes.
- [28:21]
C
Yeah.
- [28:22]
B
And now the doctor has access to this much quicker.
- [28:25]
A
Yes.
- [28:26]
B
Okay.
- [28:26]
A
So basically that step and it is exactly that. So when we talk about AI, so right now there's an AI system by Siemens that is doing this. Generally the way you make an AI system like this is you need a whole lot of data from what humans did in the past where they figured out here's what a person's contour looks like. And they then feed that data into one of these systems. And by doing that tens of thousands of times, the system can now extremely quickly look at a new patient and go, Okay, I get 99% of what this is. So that, that is fundamentally what's going on. Something we'll come back to. You need a lot of data to start this off for. For most of these services. In this case, it requires, you know, the data that we've had from a long time of doctors doing this by hand. But now this machine can do it.
- [29:08]
B
The same way a doctor would learn. Right. A doctor would.
- [29:11]
A
Yes.
- [29:11]
B
Look at one of these composites or whatever and see what a previous successful doctor already did and learn from that is just a sped up advanced technological solute. But it's the same kind of.
- [29:23]
C
Yes, this is, this is really cool because I think naively I. When I think about medical solutions because of machine learning, I imagine somebody in a lab like pouring over data and a bunch of different scientists around the world like collecting data that the AI then analyzed and then reduces into some sort of like vaccine or super serum and. And then that's the cure for cancer. Like in my head, that's the, that's the.
- [29:53]
A
We will get to that today.
- [29:54]
C
That's what AI could do. And in 10 years. Within 10 years.
- [29:57]
B
Within 10 years, as you promised.
- [29:59]
C
But in this case, it's so cool to see. It's like, no, this is just an automation of an existing process that takes a really long time that humans have to do themselves right now that we can just streamline. And it's not creating some sort of miracle treatment that doesn't exist, it's just making the existing treatment way more effective.
- [30:16]
A
Yes.
- [30:16]
C
And faster to get to. Which is.
- [30:18]
B
And reducing mistakes, I assume. You know, one study I saw a while back that is super stuck with me is the idea that, you know, judges have spent their whole life studying law. They're probably more knowledgeable the law than any of us. Judges impose harsher penalties before their lunch break than immediately after because they're hangry.
- [30:37]
A
Right.
- [30:38]
B
That's a real, it's a proven or longitudinal study that when they're hangry, they, they're more so like humans just have these kind of especially things they do every day.
- [30:47]
A
We're human, like we're just human.
- [30:48]
B
They're going to make these errors or.
- [30:49]
C
Even talking to friends that are in, that are nurses or are in residency or are doctors, you know, being at the tail end of your 20 hour shift as a resident, dude, it's insane. What, what? Because I've asked my friends before, I'm like, what are you bringing to the table at hour 19?
- [31:07]
A
I don't want you to see me at hour.
- [31:09]
B
That's what I want.
- [31:10]
A
Yeah.
- [31:10]
B
I want a fresh.
- [31:11]
C
But that is, that is the phase that you're interacting with, with your medical staff a lot of the time. And I, that's not necessary. That's not their fault. I mean, the system is constructed in a way that forces that to be the case. But. Yeah, not. I wouldn't want to have my exhausted MD trying to piece together all the scans at the end of a long day when something could potentially. Yeah.
- [31:37]
B
You know, life or death for you and like for him. You know, one of the things we've.
- [31:40]
A
All felt terrible, you wake up, you're exhausted, you got a bad night of sleep. Whatever happened, maybe you're a little sick or something, but you still go into work and you kind of power through it. Imagine you're a brain surgeon, right. Like that person is. Their entire life is in your hands that morning. And it's like, it's crazy how much depends on the performance of a person and how variable it is. One of the things with this family member of mine who passed from pancreatic cancer over the past like two years, they got to a point where chemotherapy had gotten them to where they're eligible for surgery. So eventually they did surgery. But we talked to three different surgeons who gave three completely different answers about whether or not there should be this surgery for the pancreatic cancer. Right? Yeah. So again, it's like even when you, these are like three of the top doctors in the country, even when you talk to them, it's like people have different perspectives of what's going on. So one of the things that crafted, mentioned who again does this radiation oncology work right now, is that like different, as you can imagine, different people have different specialties in the medical field. So some people in this process of treating cancer patients with radiation might be a little more skilled at one area or better at detecting certain types of things. Right. And one of the benefits of AI and we'll get to downsides, but one of the benefits of AI is that you're kind of pooling all that information together. Right. If you can have the scans from the guy who's insanely good at, you know, at finding pancreatic cancer or the angle that other doctors can't, and you combine that with the doctors, way better at contrast or whatever it is. Right. And you combine those skill sets together rather than you like you have a source of knowledge that pools all of it rather than what people right now have to do, which is you basically reach out to doctors and you try to get the best one, Hope you get the best one. Right?
- [33:14]
B
Yeah, I think that.
- [33:15]
C
Oh, do you know what sort of feedback has there been on this so far? Because I imagine this is similar to self driving in that the obvious benefit is that by automating this, you could potentially make it way more effective, safer, cheaper, et cetera. Right?
- [33:36]
A
Yes.
- [33:36]
C
But much like self driving, any accident that is highlighted is, even if it's like by the stats, maybe more effective than the human, it's replacing it becomes a gigantic story or fear that any accident or failure is happening at all. Right. So it's not seen as an adequate replacement for the human equivalent until it's totally perfect.
- [33:58]
A
Yes.
- [33:59]
C
So has there been, you know, what is that, for example, the one we're looking at right now? What is the effectiveness of this so far? What's the pushback?
- [34:06]
A
Yeah. So. And we're going to keep like Right now we're just talking about what AI is doing currently in 2025.
- [34:11]
C
Yeah.
- [34:12]
A
And then we're going to get way crazier over the next like 30 minutes. But yeah, right now. So there's a lot of hesitation for crafted who is doing this radiation work and talked about this his experience, it's faster and there's less errors actually than humans doing this. So it seems to largely be extremely positive. There are of course long term questions though of like, how do you train somebody to be good at this in the first place and be able to come in and review that end process if you weren't grinding and doing that first thing. There's ways around this, but every industry is asking this right now. For example, coding, same question. You're like, should a person learn the fundamentals of computer science or what if they just use an AI and never learn them? Right. So this is very interesting. Like how do you teach new people in the field to be able to oversee AIs? There's a long, long concern around that.
- [34:58]
B
It's very useful when you have an expert paired with AI who can make things quicker and easy, but they have the expertise to check it. But I do understand the the issue with new people in the industry. It's like they don't get that expertise.
- [35:10]
C
And they're still why can't I just type it in into my TI84?
- [35:14]
B
That is an interesting.
- [35:15]
A
I'm asking you to do it. Yeah. I mean, so with software it's like I have a computer science degree. I went through learning the fact that foundation when I use an AI, I can check off everything it's doing. And then I have friends who are learning from scratch with ChatGPT and they barely know any fundamentals at all. Right. And so there's a very interesting broad question about learning. So that's part of it. I think there's also a lot of resistance talking with my sister who's again a nurse practitioner or nurse. Resident nurse. Oh my God. Rn, resident nurse, registered nurse, which that's like right below being a doctor basically. Cool.
- [35:48]
C
Wait, I'm sorry. So go ahead. It doesn't really matter.
- [35:51]
A
Point being she works in primary care.
- [35:53]
B
Okay.
- [35:53]
A
And so she described a lot of her and also my mom is a physician. So a lot of doctor people.
- [35:59]
B
Doctors in the family.
- [36:00]
A
A lot of doctors and family.
- [36:01]
B
You went into tech though?
- [36:02]
A
Yeah, my.
- [36:03]
B
Okay, you're kind of a black sheep.
- [36:04]
A
My parents are doctors, my brother and brother in law are doctors or you know, in the healthcare field. My cousins are doctors. My aunt and uncle are doctors and me and my brother play video games.
- [36:14]
B
You said, I don't want to help sick people.
- [36:16]
A
No. Well, so I actually was like thinking about it and then World of Warcraft came out. I was like, I want to make video games. And now, hey, who's having the bigger impact on the world now? We're podcasters.
- [36:27]
B
We're the most important.
- [36:28]
C
We are solving it right now, saving lives.
- [36:32]
A
And so she, she talked about how there's a lot of, A lot of hesitation for, for multiple reasons. One is people in the medical industry are having the same experience like we talked about earlier that everybody else is having, where some executive barges in and is like, you're going to, you're going to use this AI chat bot and it's just not going to save us so much money. Right, right. You know this, use this to reply to your patients. They're just like, oh my God, no. So there's a lot of resistance that is justified. There's a lot of resistance in the medical industry generally to change things. There's a lot of old crotchety doctors who just don't. And my dad experienced this when he was trying to get doctors to switch to telemedicine. So much resistance because these doctors have been in the field for 30 years. Didn't want to do it. There's like, no, don't want it. I don't want to. They need to come into the office. And it's like, dude, this is clearly going to be helpful for many patients. So there's, it's partially just old guard kind of vibe that people being averse to change. And it's partially. It's not there yet. Right. It's it for most of them, it's not actually adding value yet.
- [37:26]
B
I'll say a third thing is with all things, somebody financially benefits from the current way of doing things. Like somebody is an expert.
- [37:35]
A
Right.
- [37:36]
B
That can get paid absolute top dollar.
- [37:38]
A
Yes.
- [37:38]
B
To be compiling these things. And if it becomes fast and cheap and easier, that person loses an economic advantage. And so that a lot of resistance comes back to that. In most fields, I don't know about medical field, but in a lot of fields, some people are resistant because they have a monopoly on some skill that allows them to extract extra wealth that is attacked.
- [37:58]
A
And the last piece is the, the healthcare industry in a lot of ways is not. You're not a robot that's just doing this stuff. Maybe, you know, maybe a little more if you're a technician behind the scenes, but like, if you are for Example, a prim care physician. A large portion of what you're doing is also kind of being a psychiatrist. Right. You are also determining like, what's going on. You know, the example I've been given of like, if you're going in, you're going to treat somebody who has trauma and you walk in the door and they start sobbing versus a person who's like drunk, versus a person who's the most lays off air person who has no issue. They just immediately it's old woman who takes her shirt off and she's like, all right, go ahead. And you know, like, those are all very different situations. Yeah. And so, yeah, I won't give all the details, but I heard some interesting examples. But, you know, a large portion of the job is about being, you know, being tuned into the human element of what's going on. It's not an algorithm to fix everything. You got to be conscious in there. And that's both, that's both a, you know, a blessing and a curse to the industry. The blessing is people can really be human and, you know, carefully cater what they're doing to what a person needs. And on the other side, not everybody can do that, you know, and it's almost unfair for us to treat the healthcare system as like, you should listen to all my woes and be, you know, responsible for everything. Which is just. Is one of the reasons that so many people in primary care feel so overwhelmed. And everybody I've talked to feels like it's overwhelming right now.
- [39:15]
B
Yeah, I've heard that as well.
- [39:17]
A
So this is some of the what's going on right now. So now we're going to start to accelerate. All right, this is one example we've talked about and now talking about a lot more stuff. Turns out this example with radiation of helping speed up the process, cutting that three hour thing down is being replicated in a lot of different areas. This next one, if you pull this up, Perry Paige, AI so this is a company right now, a healthcare medical company that is helping speed up cancer diagnosis. So again, we talked about how biopsies is where you actually get cells from the tumor and then you analyze it to go, okay, what type of cancer is this? Confirm is it cancer or not? This software does that same thing. It is looking at an image that it's been trained on many, many, many biopsy images. It now knows how to recognize cancer in biopsy images extremely fast. So doctors can get the sample, feed the images of it to this AI. And now it's able to diagnose exactly what's going on much quicker. This is FDA approved for prostate cancer cancer detection. They're expanding to other cancers. Again, this might sound like, eh, why is that a big deal? But if this speeds up how quickly you can start cancer treatment. If you're a patient who, like, if you're given, for example, pancreatic cancer, which you very well could have a couple weeks to live if you don't start treatment, this is a big, big fucking deal. If this cleans off an hour or.
- [40:31]
B
Two hours, does this mean when I start logging into places and I have to do a captcha, it's going to be four cancer cell images and I have to pick which one pancreatic are they going to use?
- [40:42]
A
It's crazy. Like tumor cells in the body all look like letters that are a little distorted. Three or four.
- [40:47]
C
I just need to get in my yacht. I just want to log in because.
- [40:51]
B
I'm always picking the one that has the bus in it. And I know they're using me for training Doug. And I just, I, I picked the wrong one or it's half a bus. It's not right. And I get, I hate it.
- [40:59]
C
I hate it. But okay, I see this. The cool thing with this is I imagine outside of cancer, anything that is image diagnoses based, right? Like I just sprained my ankle pretty badly four weeks ago.
- [41:10]
A
Yes.
- [41:11]
C
Any, anything that is a, like a break, a sprain, like your fracture, a torn ligament, anything that could be identified through imaging. You could compile a database of different like breaks or issues and then have the tool reference the database and then pull an immediate diagnosis. That is amazing.
- [41:36]
B
Pretty cool.
- [41:36]
C
I mean, I imagine there's some sort of maybe like downside here with protection of your medical data. Something like that. Like we have this, you know, we have counterpoint.
- [41:49]
A
I'm a tech bro, don't care about your data. Okay, now we will talk about data. Yeah. That's a key part of this.
- [41:55]
C
Yeah. And I wonder how that we can.
- [41:57]
B
Am I crazy?
- [41:58]
A
That's a good question. We'll talk about that.
- [42:00]
C
It's interesting to see that this would clash with some level of modern like, like HIPAA regulation, right? This, this would clash with that. And the data or the pool of data to reference might not be possible to build with the current regulation. And who gets to own that data? Private companies? Is the government. Who gets to take advantage of that? But I'm just thinking about my experience walking into an urgent care four weeks ago, trying to figure out how bad my ankle sprain was. Right and how long this was going to take to heal what I should do. It's like if I and I got an X ray and so having that be verified quickly through a giant system of all the people have been imagining millions of other people. That sounds amazing.
- [42:42]
A
Yeah. And that's a mild thing. Right. You know, so I'll rapidly go. So you've picked up on the pattern which we were working towards. This is all image based stuff. Right. So let me hit with a couple more. This is viz AI. They basically once a stroke patient comes in. So stroke is where there's blood flow blockage in your brain in some way. This is like every minute that goes by your brain is dying faster. And more like it is super time sensitive to treat somebody with stroke. You get them into the hospital, you do a scan and then you need to, you need to look at the scanned images of their brain and go, what type of stroke is this? What's a reasonable treatment? And if that is typically 30 to 60 minutes viz, I can knock that down to a couple of minutes. Right. And that is a huge deal. If you're talking about a five hour process knocking out 45 minutes and allowing the treatment to start faster might literally be the difference between you have your legs functioning or not. Right. This is a big deal. Again, this is in. This is serving 45,000 health care providers right now. It's been acknowledged by the government. Some discord folks suggested one from Super Levis in the discord Singapore. I think I have it up here. No, that's another thing. Okay, well I'll just list out. Singapore has been like their National Health Care Innovation center has been funding AI projects that are like doing eye scan really, really quickly. And so now they're able to do mammograms and eye scans way faster than they were before with higher accuracy and get more treatments through. This is currently happening in Singapore and a bunch of areas. Denmark released a three year study that shows that with breast cancer radiology or the breast cancer analysis, if they use AI systems to again basically analyze the image that come from these scans that people are doing to see if you have breast cancer, that it is dramatically speeding up. Basically how they're able to detect these things and how quickly.
- [44:24]
C
It's always Denmark in Singapore, bro.
- [44:26]
A
Yeah, I'm sick of them.
- [44:27]
C
I'm sick of them.
- [44:29]
B
I am sick of this.
- [44:31]
C
With the Nordics, dude, it's like, oh, what is this cool breakthrough? Where, where is this happening? Oh, interesting. It's the same three, you know, we.
- [44:38]
B
Added a burger where you can have the donuts instead of the buns. And I don't ever see that coming out of Denmark.
- [44:44]
C
That's actually next week's episode 45.
- [44:51]
A
All.
- [44:51]
B
The KFC double chicken.
- [44:53]
A
Next week is live at the state fair. We're frying everything, we're eating everything, and we're giving a middle finger to Denmark. All right, so we can quickly chat about data. It's exactly what you said. Right. So to summarize everything that's been going on right now, when you're treating diseases, conditions, some of which are super time sensitive, you have to do analysis of imagery that you're scanning from people. And if you can get it, if you can get an AI to, you know, if you give them a couple hundred thousand examples, it can be incredibly good at doing this. You have a doctor come in and oversee the final result, check it off, say, yes, that does look good. It's not like we're removing doctors from the situation, but you're speeding it up dramatically. The question, of course, is like, where's that data come from? So that's not clear right now. Talking to Crafted with the nih. Nhs. Nhs. With the NHS in the uk, it's like you got to get companies that have gathered all this stuff to pull it together, and that can be really challenging. In the United States, we have a hippo that you also can't. You can't just like share this. Big data, big hippo, he's got all the.
- [45:52]
C
That would be surprisingly effective.
- [45:54]
B
You have to get past this hip to get to the file.
- [45:57]
A
That's the. We've installed him as the guard and that way only like a really good doctor can get in there and grab files.
- [46:03]
B
Yes. The trick him with fire.
- [46:05]
A
Yeah. So that's a question. Not only is there patient privacy of like, okay, are you down? Let's say you went and got cancer, you know, scans for your body, Are you down for it to just be. Go fed into some AI for some other company? Yes, I would be.
- [46:18]
B
But a lot of people, why wouldn't you? I can see, like, if I write War and Peace and Mark Zuckerberg steals that, scans it and puts it into his fucking meta thing so he can try not to write. I get being upset with that, but I don't understand why if a doctor correctly identifies my cancer and then we anonymize it so nobody knows that it's me with cancer and it feds into a million other things so they can find other camp. Like what?
- [46:39]
C
I think there's just this. This Strong sense of, you know, bodily autonomy. It's like this idea of anything that is yours, especially your body specifically, being used to be in a way that could be monetized. Especially, especially maybe losing the autonomy itself, but especially losing it at the behest of monetization is something that people don't like. I think when you think about the tangible benefits, like, to me, one thing I hate when I move and then I try to go get like a checkup or like a new primary care physician, is that I feel like we have to walk through all of this basic background information again. We have to start with all these new questions. And as somebody who's moved a lot and had to do that over and over and over again, it's like, it would be so helpful if you, as my new doctor, could just pull up a page of all of the information about my health from all the previous times I had come in so you have a more specific picture of what's going on with me. And also if it's notes from previous doctors, probably better than the picture that I can present to you as the doctor right now. And in the same way, it's like, well, I would want to volunteer my information to be a part of a pool of data like this. I think maybe similar to something like signing up to be an organ donor on your driver's license. Right.
- [47:52]
A
That you basically agree I am okay with this image being used for training.
- [47:55]
C
Yeah. There probably has to be an opt in aspect to that. My pushback to that would be like, well, what securities do I have over that information being abused? But to pretend like I'm really concerned like that about that personally right now, it's like, well, I think I'm already like, you know, scrolling through terms of service when it comes to medical apps in my life and signing them over already in different ways. So to pretend like I'm suddenly worried about this now when I haven't really been before would be kind of making that up. I would just hope that, you know, the regulation that comes alongside this is like preventing my, I don't know, the pictures of my ball cancer from being spread out publicly. What if you got, I don't know.
- [48:38]
A
If you check a box or something that says if this is used by an AI, they have to Photoshop my dick to look bigger, make it a little. And that way if the data does.
- [48:46]
C
Look good, it's like you've really touched on the main issue. I was kind of.
- [48:50]
A
Yeah, we read in between the line.
- [48:53]
B
Cuts the heart of it. Wait a Minute I do see, actually I just saw an issue. Okay. What if, what if somebody, what if an insurance company buys this data and then can de. Anonymize it or whatever. Now you have higher rates because they know you have pre existing conditioner and there it is.
- [49:07]
C
Just realized that's why they pay you the big bucks.
- [49:09]
B
I just realized how, how I would, I was like, how would I abuse that? Okay, that's okay.
- [49:13]
A
Yeah, no, there. And it's in. Another question is like, okay, let's say, say you're. You're seaman. Right. Which makes these auto contouring things. Yeah. Again, this, this is being used right now to help with radiology and help it go faster. This is great.
- [49:24]
C
Yeah.
- [49:25]
A
I don't know. I should have asked and maybe I forgot, but crafted. I don't remember asking. Where does that data come from? Right. And so really to, to build tools like this, you want companies pooling all their stuff together. But if you are a company like Siemens, you're incentivized to keep the data for yourself, right?
- [49:41]
C
Yeah.
- [49:42]
A
You don't necessarily, if you're to make your product better. Right. It's a, you know, if you are the company that has been pooling all this data, let's say for stroke analysis and then you have the ability to create your own AI and another company is like, hey, let's all share so we can all make products.
- [49:55]
B
You're like, fuck you.
- [49:56]
A
Yeah, you might be like, fuck you. It's also, it's interesting with healthcare. I think this is by far the easiest and most likable way to apply AI in our world. So many other areas that are a lot messier with this one, I think most people are like, you know what if less people can get jobs in health care, but cancer is 10 times easier to treat, we're okay with that, right? I think most. Same with the imaging.
- [50:19]
B
And cheaper, right?
- [50:20]
A
And cheaper, right, Right.
- [50:21]
B
Like people really want it to be cheaper.
- [50:22]
A
And again, we'll keep going forward. This is going to keep advancing of like how AI is going to do more and more and more. But I think with health care specifically, unlike some of the other areas like writing or creativity or all of this stuff, it's a little easier to be like, okay, for the public good, let's pool this data. Let's make, let's make an AI able to diagnose this stuff. Because preventing somebody from dying of cancer, getting that rate down even 10%, saving, you know, 10,000 more lives a year, that's worth it. And maybe that means the Government steps up and is like, you have to share this data.
- [50:51]
C
Okay.
- [50:51]
B
You know how when you like go to a website that you haven't been to before and you can log in with Google or you can log in with Apple. What if there was like a government thing with your data and I could go to any hospital or somewhere and I could log in, I could give them permission and it goes to a database that has my previous things. And what if it's like the login of.
- [51:11]
C
I. I know it might, I know it might sound like a joke, I know it might sound like a joke, but I think this is actually how it works in Estonia. I think this might actually be back to our roots.
- [51:22]
A
This might be a bit of an Estonia part situation.
- [51:25]
B
Reminds me of the Wire, episode four, actually.
- [51:27]
C
The Wire, Estonia. Those are pretty much the only tools I have in my mind. No, I think they, they have, I think they might have a system where your, your, your digitized medical data is like more widely, more widely available.
- [51:41]
B
Like I still own it, but I could log in and they could give them permission.
- [51:45]
C
I don't know if it's literally like a lot.
- [51:46]
B
I don't, I don't like an ability where I can give them the, the unlock key.
- [51:50]
C
Yeah, I mean that. I think what we're asking for here and what, what you are too is what is the way that this, like these private companies have a motivation to develop solutions for healthcare while not abusing the fact that we're contributing very personal data in order to build these databases. And, and also the biggest data. Well, it's not the biggest of the three of us. I don't know about that, but fucking megabytes, dude.
- [52:20]
B
We'll say medium data, okay.
- [52:22]
A
In Microsoft Word. It's going like 15 pages down. It just keeps going.
- [52:26]
B
You need like half a data scientist to some. I might need a full team squad.
- [52:33]
C
It's not a oneman job.
- [52:34]
A
Well, if we take my dick and every other porn star, we can make a penis analysis AI that can do that quickly. That way I'm not adding too much burden to bring down the average. Well, I'm bringing up the average.
- [52:48]
C
But what if, if you, you want to thread the needle of having those things not also allow the average person to be charged more for healthcare or get made it more inaccessible? The goal of these things is to increase efficiency, reduce cost. I, I mean if I were we there, there's a section of. Ah, man, I'm going back to it. There's a section of abundance at the end when they're talking more about medical advancements and innovation and it's talking about the government using push and pull incentives specifically in the, the medicine field, the medical field, the medicine to get private companies to like explore and take on more risk to things that would benefit, you know, benefit society.
- [53:39]
A
Right.
- [53:39]
C
And I think one of the main examples it uses is the COVID vaccine MRNA and it talks about the government resources that were used to initially fund and allow the scientists who basically discovered MRI marina as like a treatment for things or a potential treatment for things. And then later on when we made Covid vaccines which were made by private companies, the government, even the US Government. Right. We didn't get charged for Covid vaccines. They were distributed and they were given out for free.
- [54:10]
B
Well, that was because FAUCI wanted to put a microchip.
- [54:13]
C
That was because of the Fauci microchip. Yes.
- [54:15]
B
To track me and I was willing.
- [54:17]
C
To accept that I take the microchip in order to be less damaged by COVID 19 I see. And I knew the trade off I was making and a lot of people don't know that. But in this case it's similar. Right. Is you want the, you need the underlying technology to be developed and the solutions to for these health care problems to be created. But at some point you want some sort of like government ownership or regulation over these things so that they aren't abused. That's what I would be looking for. Right?
- [54:47]
A
Yeah.
- [54:48]
C
I mean that's the. If the NHS is working with a company like that, that's a national health care system that is cooperating with a tool like this to make things happen.
- [54:57]
A
I'm concerned it will be harder. With the US health care system, everything is so driven by profit. Where the NHS can as the the country be like we are going to do this. It's still challenging. From what crafted was saying of like some of the forward facing AI stuff. Another thing somebody rumble Badger in Discord gave some great thoughts on this, has also worked in this field on like, you know, developing AI tools for this type of stuff. He, he listed an article that was mentioned that basically shows to put it quickly, you have these ECG systems that check a person's heart rate for arrhythmias. So again one of these specialty, you know, fields where it's like can we help a person with this certain type of issue? And a person, a human has to go through this vast amount of data, basically how their heart was beating over like multiple days. Now they have an AI trained on all that stuff. It was the humans were 14 times more likely to miss a diagnosis compared to an AI. Right. But the AI would say people had a heart arrhythmia a little more often than a human was. So it's like it was better in some ways, worse in others. And that's. It seems like a lot of the AIs are like that. It's not a silver bullet that solves everything, but it does. It gets you a lot of the way there, but it still needs some human oversight. But what he was saying is like, it's like he knows the company that contributed the data to that AI and he's like, that is not even what the data was really meant to be for. Like, technically that's kind of correct. Also, the company that does that has been dubious in its quality from his perspective. So it brings up this whole thing around the quality of data actually matters a lot. And there's different devices. So if, you know, it's. Not everybody is formatting things exactly the same way. So the way in which you bring all this data together to train an AI, where it comes from the quality of the people who is generating that data. What if you had a shitty doctor who kept misdiagnosing things and then that goes into the pool? Right? Yeah.
- [56:43]
C
And this was from the. This was one of the big notes I had from discussion in the discord when the topic was teased was how difficult data is to come by in the medical field. Good quality data is not as widely available as us just scraping the Internet for images. And then.
- [56:59]
A
And it's more important to get it right. It matters if the consequences are higher. ChatGPT makes some shit up, you know, okay, whatever. If this thing says you don't have a heart arrhythmia, you know, if, if the auto contouring system is like, oh yeah, this guy can take a shitload of radiation and then you get blasted in your pancreas with too much. It's like, dude, this is, this is a big deal. And it gets to then government oversight, where again, I will just reiterate for people who seem to not realize this, I am in favor of regulating AI. I'm not. I don't advocate for AI to just be this like wild west craziness. There has to be regulation around AI algorithms broadly, but particularly in the medical field. Right. Because holy shit. Right? Yeah. If Siemens comes in or this other company comes in and says, you're a child, it misdiagnoses my stroke and says I have a.
- [57:43]
B
Sorry, don't let the adults talk on this. One, because this baby over here.
- [57:48]
A
Dude, I said, that's like the eighth time I've said Siemens. It got funny.
- [57:51]
C
He's giggled every time, dude, it just comes back. It just. It just hits every time. Because I'm so sorry.
- [57:57]
B
I.
- [57:57]
C
But I was listening and then you derailed.
- [57:59]
B
I did I drill it one more time because I can't move on. The character you created of Guy who believes every Fauci conspiracy theory but got the vaccine anyway is so funny in my mind. He believes all of them, but he's like, that's a fair trait.
- [58:13]
C
Dr. Fauci's evil, bro. But the microchip is worth the trade off. He's tracking me right now because it's.
- [58:19]
B
Cracking me up in my brain. Okay, we're back, Doug.
- [58:21]
A
I'm sorry, we're back.
- [58:22]
B
Sorry, sorry.
- [58:23]
A
So there's a lot of challenges and complexities around the data of these things. So to summarize everything we've talked about so far, AI can help with image analysis that's being done right now in a bunch of different fields by a bunch of countries. I've talked to people for whom they're like, this is helping. This is better. This is potentially saving a lot of lives. It's a big fucking deal. If somebody that you love gets diagnosed with cancer and this gets them starting treatment a week earlier or a month earlier, this is a big, big deal. It's easy to be like, is going to be great. This is saving lives right now, and we'll continue to do so as it gets better. So I think this is a really big deal. There's a lot of challenges though, with regulation, data, all this other stuff. So now let's get into the crazier stuff, which is forward facing. Okay, so this is all happening right now, but the real holy grail is like, okay, well, you know, it's great that we, you know, are going to get better at treating cancer with radiation, but why do we let it get to the point that a person has cancer spreading everywhere before we even start treating it? Right. Our whole healthcare system. And there's like this article I read from the. Oh, is it Doug's about to go.
- [59:23]
C
Full minority report on cancer?
- [59:26]
A
Yeah. This is the American Hospital association. And really the gist of this, and this is what my mom specialized in for a long time in the healthcare industry. Preventative medicine. We have, there's. Right now we have like a sick system, which is basically, if you get sick, you show up. But it's way, way, way easier to treat people if you do it in advance. There's lots of examples of this. Like, you know, before your heart fails, if you start exercising and taking care of it earlier, that's going to prevent all these issues down the road. So not only for things like, you know, like smoking is an obvious one, but less obvious one. Alzheimer's fucks up your brain, right? Like it's your stops brain functioning correctly. And Alzheimer's is brutal in the world. We basically, by the time you recognize that Alzheimer's is happening, it's too late. Like, it's already going and it's already done damage. If you could detect it earlier, which we don't really have the capability to do right now, you could stop it. Even our medicine for Alzheimer's right now is mostly about stopping the progression. But if it's already done a lot of damage, you can't do very much. The person in my family was diagnosed with pancreatic cancer. It's one of the most deadly. Basically, everybody's dead within a couple years. And the biggest reason on top of it just being brutal is because you don't catch it until it's really developed. So unlike some other cancers like breast cancer, which it's easier to find early on, and maybe there's a lump or whatever else, you don't notice a pancreatic tumor until it's really causing problems. And by that point, not only is it in an area that's hard to operate on and it's near all these other organs, it also likely has spread. So very often with pancreatic cancer, it's like when somebody realizes they have it, they have a few weeks, like, it's often that serious, and then you're lucky if you make it like another two years or whatever. So preventative health care is hugely is a huge opportunity to not only massively shift how we treat people in general, but has a huge impact on the number of people whose lives could be saved, the number of people, like, the amount of money and resources it takes.
- [61:20]
B
This is a horrendously scary graph that you're showing here, by the way. I don't know if this is visible on screen.
- [61:23]
A
Yeah. So number of chronic health conditions. Basically, this whole article is about how the amount of chronic health conditions has dramatically increased in America.
- [61:31]
B
Can you go back every year that that number is millions of Americans? So 171, 25, 164 million Americans with chronic medical conditions.
- [61:42]
C
You scroll down what were the most common conditions?
- [61:44]
B
That's. That's.
- [61:45]
A
And then half the country. Yes. And then one of the One of the things that noted here, hypertensive arthritis. In the last between 1987, 2002, 2/3 of the growth in Medicare spending was accounted for by 10 chronic conditions. Like the majority of the cost, I don't want to say majority, but huge percentage of the amount of money that we spend as a country every year, at least in America. And, but this is relevant for every country on earth goes towards treating people who are already sick with chronic conditions. And if you could stop those things from happening in the first place, this is maybe what allows Medicare to even continue to exist in 10 years because it's going to go bankrupt at this point, right? This is, this is a huge deal. But preventative health care is really hard. Some of it's easy, right? And for example, you know, women going to get mammograms, right? And you have preventative screening for things like cancers. But imagine that right now all three of us are like, this is scary. I don't want cancer. I want to go in and get a scan, scan my whole body, see if there's cancer anywhere. Well, that's not feasible. When you go and you get a CT scan or an MRI like we talked about earlier, you don't want to CT scan your entire body. That's radiation. You don't want to fill yourself with radiation, just maybe in case you have a problem, right? MRI machines, expensive, difficult, slow. You don't want to MRI your entire body. That's not protocol, even if you convinced a hospital to do it, right. There's fields of somebody who looks for tumors in the brain, specifically tumors in the pancreas, specifically tumor in the lungs, tumor in the, in the abdomen, whatever, right? So you would have to then have a variety of specialists, all individually look over this giant amount of data that you produced about your own body to be able to tell you whether you maybe have cancer right now. And you probably don't. Doing that one time is completely infeasible right now. Doing that every six months, basically impossible. But I'm sure you could pick up from that. Well, if it's about processing data, that is exactly what AIs can do. There is no reason why long term, down the road you couldn't get a full body scan or some kind of imaging or information about your body every six months. And an AI has been trained on all the different potential diseases, problems, things that could come up, early onset Alzheimer's or early tumor markers or early hypertension or whatever it is, right? All these things that could be scanned, that is completely infeasible right now, which suddenly you can imagine. Wait a minute. If we had an AI trained on this and we expanded the resources for scanning people regularly, this could massively solve what we're dealing with and reduce the number of people who are just going to the hospital because they're fucked up, you know, and get that down to. Most people are preventing things before they happen. So there's two examples of this happening right now. One is gallery. So this is a blood test company. So their mission is basically to do that. They want to do a blood test that can scan for cancers and let you know if you're at risk for cancer and if they're seeing early onset cancer. This is in clinical trials right now. So it's not like out in the world, but there's like 100,000 people in this, using this right now in the UK specifically to get data to show and refine this. And this is an AI analyzing data about the bloodstream. Specifically, it's looking at fragments of DNA that come from cancer cells and using that to identify what we never would have been able to before and say, hey, you actually have this, this and this cancer that is starting to happen. Go into chemo now.
- [64:59]
B
So it's hundreds of thousands of people taking regular blood tests. The AI is trained on all their data and knows that if your blood looks like this, it's very similar to someone else's blood who had cancer that looks like this. And we've made this connection and that you can.
- [65:11]
A
Right.
- [65:11]
B
You're more. Okay.
- [65:12]
A
And a kind of thing to note here is it's, it's. In this case, it's not even just, oh, it's doing the work that a human could have done. Analyzing CFDNA is not, from my understanding, not possible right now. That's too.
- [65:23]
B
So this is new, right?
- [65:24]
A
And this is basically like, this is a data source that we didn't even have the ability to really use in a meaningful way before. Certainly not at scale. But given that AIs can basically learn to be nearly as good as a human and then apply that knowledge at 10,000 times the speed of a human, well, suddenly this is a data point we could actually use that could save literally tens of millions of lives or get people started in cancer treatment early. Maybe that person who's about to die of prostate cancer, they learned about it five years earlier and they were able to apply treatment earlier. Right.
- [65:53]
C
You know, I can't say that the exact, you know, because I, I'm not familiar with all the specifics, but my, my mom Passed away when I was like 1 years old from cancer and she was like 29. And I wonder, you know, I, I, I have a great life, I have a great mom. I'm not asking to change the past or anything, I'm just curious. Like there's, and, and I have two other family members who have died of cancer. And I wonder, you know, is, is this like the difference maker for, for a lot of people, like just them catching it earlier because she, I, you know, her, and then her dad, my grandfather, who I never met, they both died before they were like 40. And that's, and from like something that was just rare and like hard to find and late stage and that's, that was it.
- [66:40]
A
So at that point, again, you got limited options. And so I experienced the same thing very recently. For the last couple years I've been going through the, the logistics of that and it's fucking brutal. I mean it is so emotionally draining and just, and obviously for the person going through cancer treatment, it's just absolutely fucking brutal. And the idea that you could prevent that from happening or at the very least you start somebody treating, you know, a pancreatic tumor when it's just starting, right? And maybe chemo, a round of chemo every six months is able to keep it abated before it really grows where you're able to hit it with radiation, with a smaller amount of radiation. Some of the things I wanted to talk about, there's not really time. Another area that we're looking at that, so this is from again crafted robot who works in radiation in the NHS is right now we shoot a beam of radiation at the tumor, right? But you have to make it a really wide beam because if a person has like if a person's body has shifted from the previous time they came in and they did the scan, let's say that day their bladder is more full or they've been working out or whatever, something has changed in the body, body those millimeters matter and so they have to like make the beam of radiation wider in order to ensure that the changes in the body they're still going to hit the tumor. Right? And there's multiple factors that basically make them keep widening the thing to account for possible changes that have happened. And if an AI, and this is what they're starting to look into now and could easily happen in the next five or 10 years if an AI is able to provide the information to, for example, the day a patient is screened in an MRI machine, they get out of the MRI machine or maybe stay There. And the AI does a full analysis and recommends a treatment plan that can then be reviewed by doctors within two hours and the treatment happens that day. The beam of radiation, again, imagine you're shooting a gun into your body at the tumor. Like you can make the bullet a lot smaller. Right now they have to use giant ass bullets and you're filling the body with a ton of radiation. That again can literally cause cancer and cause all this damage. It is super bad. And the more refined you can be and more targeted, this is a huge, huge difference. And so this is like very feasible in the next few years. And then these other areas, we're talking about a preventative medicine light gallery. There's also a startup that I was looking at called Superpower. Their whole thing is they want to have this six month. Every six month you get a lab test and IT is analyzing 100 different biomarkers and gives you this comprehensive look of here are all these risks. And again, it's AI processing this data and being able to do this. And again, this isn't feasible for every person on earth. Even if you 10x to the amount of doctors in the world, it's still not feasible to do this at scale. But if suddenly the time that's taken and the sheer amount of data analysis gets reduced to like almost instantaneous, this is opening up all these new incredible things. There is a real world where in five to 10 years, many of the debilitating diseases that have killed so many people we know in our lifetimes, you now know five or ten years ahead of time, could start treatment way earlier. And in some cases it never grows to a deadly point.
- [69:31]
B
It sounds incredible. I guess here's my feedback.
- [69:33]
A
We haven't even gotten to AI curing all disease, by the way. That's chapter three. We're almost there.
- [69:38]
B
All right. I want to give my feedback is that in a, in a public health care system like the nhs, I can totally see how this works. I'm already seeing. I guess I have fears, like if I'm taking a blood test every six months in an environment where my employer is paying for my health insurance and they find out that I now have early stage pancreatic cancer or something, their incentive is to drop me from their coverage, you know, or there's some. I guess that's my. This feels like it works really well. If we've all agreed to subsidize health care now, I'm.
- [70:16]
A
It's. This is all concern. It's more concerning in America's.
- [70:19]
B
It's just one Yeah, I understand America. I mean, this is America's flaws. Nothing to do with AI.
- [70:23]
A
There is, there's some. So, you know, if, and so again, I think there's going to need to be a lot of government regulation with all this stuff. And I personally am somebody who would rather have a national health care system. Not stoked about the free market for, for medicine, but just a counterpoint. There is like, ultimately health insurance providers would rather have healthy people. Right. And so if broadly they. The government says, hey, you have to apply these preventative screenings to all these people, and that means, yes, 10% of your, you know, insured clients turns out they're high risk for these things, you are not allowed to drop them based on this. But the other 90% are now far less likely to need medical, expensive medical support. That might be a net positive for an insurance company. Right. So in theory, the overall net gain of people being healthier and overall cost dropping would in theory allow a health insurance company to actually spend so much less on the average patient.
- [71:15]
B
Just because preventative is so much cheaper.
- [71:16]
A
Right. Because preventative medicine is so ridiculously impactful, but just not feasible in many ways right now in our healthcare system. If you could introduce that into the system, bring the average cost necessary for a person way down because you've prevented all of these different issues in theory, that works out. I'm not going to pretend like the health insurance industry is some.
- [71:35]
B
No, man, I hate the health insurance.
- [71:36]
A
Right?
- [71:36]
B
Yeah. But I'm thinking it through.
- [71:38]
C
You hate them so much, you might even take action.
- [71:44]
A
It's a meaning.
- [71:45]
C
Who knows? I, Well, I, it's interesting because when we're looking at, I think when we're talking about preventative measures to combat something like cancer. Right. That is very different than the graphs we were looking at before. If you look at the chronic diseases that were listed out, I mean, it was showing like chronic disease in the disease in the United States. Right. We can, we could be honest. Obesity, smoking, most of these are from.
- [72:15]
A
Being very preventable stuff. Yeah.
- [72:17]
C
And, and, but those are not like genetically, oftentimes not genetically predispositions or have strong genetic predispositions towards those things. Those are like environment and like behavior conditions. Like the. One of the reasons that Japan's healthcare system spends so little money compared to like the US's. Right. It, you know, there are a lot of factors at play.
- [72:46]
A
Bushido spirit.
- [72:48]
B
That's what it is. It's mostly the Bushido spirit.
- [72:50]
C
But even compared to maybe other like nationalized or like public health care systems, Right. Japan spends not that much money. And one of the reasons is that they have a really healthy population to begin with. They don't eat burger and they, they eat less burger. And I want.
- [73:09]
B
Is that really living, though? I'm posing.
- [73:12]
C
Are they all strapped?
- [73:14]
B
I'm just posing.
- [73:15]
A
I'd rather fight on my. Yeah, what is it? I.
- [73:18]
B
With my mouth full of burger, I.
- [73:20]
A
Rather die on my feet than live on my knees.
- [73:23]
C
I just see like the application in an American context. I feel like from what you're saying is so much of the, the care after the fact is. Is what can be helped. Like dealing with the fallout of these conditions and helping speed up the process of like, going into the doctor. That feels like the, The American proposition also applies everywhere else as well. Right. But there are. It's. If it's like, if I just choose to like, you know, not exercise and eat a bunch all the time, and I, you know, the AI isn't helping. Helping me there, and that is such a strong amount of this disease. Or if I'm addicted to cigarettes, I'm dealing with nicotine addiction.
- [74:06]
B
Okay.
- [74:07]
C
I just, like, I do feel like the. So many of the conditions that people are dealing with and the health problems we're dealing with and the scale at which we are dealing with them have preventative measures. As far as AI catching them, I feel like this is a smaller piece of the puzzle than the part you were saying earlier. It's like, to me, the appeal is like curing, like curing the incurable, approaching things like cancer that have. Or helping call out things that are like, genetically more likely to be dealt with. Like later in my life that might be popping up. Like, those are the huge things to me. But AI isn't going to solve, like, I don't really feel like AI is going to solve the amount of people being diagnosed with heart disease because that's tied to other things.
- [74:53]
A
No, no, it will not. And I'm not claiming.
- [74:55]
B
I mean, it could tell you have an increased risk for it or something and maybe.
- [74:58]
C
Yeah, there are genetic factors in whether. In those things. Still, I'm not, I'm not ignoring them.
- [75:04]
A
So something from this article that's relevant to what you're saying. It was a quote I thought was interesting. Research suggests that clinical prevention services reduce disease, disability and death. For example, counseling all smokers. Just counseling. Counseling all smokers on a regular basis could save roughly 70,000 lives a year. Screening all persons age 50 and over with a fecal occult blood test and a sigmoid, a scopi could prevent 18,000 deaths per year.
- [75:28]
B
Sort of a sigmoid myself.
- [75:30]
A
Yeah. And so I think the, you know, part of this is just again, preventative medicine is so helpful of counseling people. And I'm so.
- [75:38]
C
Even if, like, if I were to give an example, even if I were addicted to cigarettes. Right. You by like talking to me or interview intervening so much earlier, using machine learning in some capacity might still provide a benefit down the line for.
- [75:52]
A
For me personally, I'm making the argument one, preventative medicine is extremely important and impactful. If we get to problems earlier, that's a huge deal. I am mostly focusing in this conversation on like serious disease, like Alzheimer's, cancer, things like this. There's then a second part of this which is okay, America objectively has horrific obesity problems that are causing enormous.
- [76:14]
C
And a lot of other places too.
- [76:16]
A
I can absolutely imagine a world where, you know, you have an AI health person that checks in with you once a month, that the doctor oversees. But that way people who are more receptive to this stuff or need encouragement or need more frequent guidance or whatever else can be really helped by that. In the same way that somebody who gets a personal trainer is more likely to go to the gym every week. Right. If they have some obligation. So I imagine there's some world like that. I'm certainly not claiming though that AI is going to fix. Yeah, that's obviously that gets to. You're totally right. That gets to much bigger questions around what the quality of our food in America and our diet and the way.
- [76:52]
C
All these things are incentivized cities. Like there's so many things that a.
- [76:55]
A
Lot of stuff there.
- [76:56]
C
Absolutely.
- [76:57]
A
So not making claims in that specific stuff. This is more like the hospital medical system. How do they treat serious disease? And that preventative is going to be really impactful for certain one of the deadly diseases users.
- [77:08]
C
Basically one thing you. I think maybe that fits into the more of the first chapter of what you were talking about too, I saw was similar to law firms. This is something that just helps with like bureaucracy and paperwork. Like you can use things that just make that part of the process more efficient. And a lot of what you might spend doing as a healthcare professional is like writing up, you know, or going through the meeting with the patient you just had summarizing everything, needing to communicate information to other doctors. Like all of those things that you spend time in between actually dealing with.
- [77:42]
A
Patients, like half your time.
- [77:44]
C
That's. That's so much of the time spent. Yeah. So can you tell us what about the cures for all diseases. The only 10 years away.
- [77:54]
A
You know all this shit we've talked about for an hour, 15 minutes, none of it matters. We're going to fix it all, okay? The real Holy Grail is we used to even need to treat cancer because we cure it all right? That's the Holy Grail. Okay, Back to my naive vision of what this is.
- [78:09]
B
I'm listening.
- [78:11]
A
And I'm giving the. The Doug Doug stamp. No diseases and everybody is a perfect life in 10 years. And it's all thanks to AI with.
- [78:19]
C
No regulation and no asterisk.
- [78:22]
A
Right, and no asterisks. And there's no nuance.
- [78:24]
B
2035. Okay.
- [78:26]
A
All right. So why is disease hard to figure out? Why is it hard to cure disease?
- [78:31]
B
Answer the question.
- [78:32]
A
Yeah, yeah, Doctors are lazy. All doctors are lazy, by the way.
- [78:37]
C
Still my first answer and my second, my third. My third. Is that presumably just because disease changes in so many ways over time. Like there are so many things mutate, like whether they be like bacteria or viruses or your body cancer, cells change. Things are just constantly changing. Things are not necessarily the, the same thing in every person all the time. You're not finding one all stuff.
- [79:06]
A
A my next student. What is a protein powerhouse of the cell? That's not correct. Okay, so I'm going to super simplify things. You know, disease, some kind of problem is there's some kind of, let's say, reaction process happening in the body that you don't want. Proteins are like the building blocks of biological life. Basically, they're the workforce. And you can think of a protein as a literal 3D structure. Imagine like making origami. And you, you know what? I actually. This fidget toy. I forgot about this. Like, imagine that a protein is a whole bunch of amino acids, okay? And you fold it into some weird shape. And now that's the shape of the protein. And based on its shape, that actually that dictates largely what it's going to do because it'll go around the cell or the body or whatever, and if it finds other molecules or DNA or proteins or whatever else, it can click into that and they, they lock in like a lock in a, in a, in a door, in a key. And then, and then they start making magic. They start porking and they do all.
- [80:10]
B
This cool stuff, okay? So when I drink a protein shake, that's what's, that's what's going on. That's what's going on in my body. Okay?
- [80:16]
A
So a of lot, a lot of biological function, disease comes down to Proteins. It is a major part of the biological process. And so if you are trying to do. To fix a disease, let's talk about malaria. Malaria is. I didn't even know this before right now. It's a parasite. It's like a little bug that gets in your body. I mean, not a bug because it's a single cell, but it's a parasite that gets into your body. So malaria kills something like 100,000 people a year. 200. It's an insane amount of people.
- [80:42]
C
There's some insane stat. Like, everyone. There are more people that have died of malaria in history than, like, the amount of people alive right now.
- [80:51]
A
Oh, I'm sorry. Sorry. 627,000 people a year. My bad. I was a little low. That was in 2020, so.
- [80:58]
B
Wow.
- [80:58]
A
It is. Malaria is wildly deadly. This kills literally hundreds of thousands of.
- [81:03]
B
People, mostly children under five, mostly children.
- [81:05]
A
And mostly in Africa. This shit is fucking brutal. And we don't really hear about it anymore because for us, I know it's almost like smallpox or something. We don't really deal with it much in the. In the west, but this is a big, big, big deal.
- [81:17]
C
Also, there's been, like, pretty substantial the. My understanding is that number is actually a lot lower than it was even 10, 20 years ago.
- [81:24]
A
Makes sense because we currently have, like, okay, vaccines that are like, 30% helpful. So. So let's start with malaria as. As our grounding point of, like, how do we cure all disease? So malaria is this parasite. It's. It's transported by mosquitoes, and mosquitoes, you know, bites you. The malaria gets in and it starts basically growing and developing in your bloodstream and your cells. But it's. It goes through this whole life cycle in your body, and it changes rapidly. So let's say you want to design a drug. And just to simplify it, imagine, you know, you know, what the. The parasite roughly looks like, and you're like, we need to design a drug, a molecule, something that's going to go into the body and it stops the thing. It recognizes that this is a parasite, and it stops it. But right now, the parasite has so many different forms that it goes through. It goes so rapidly, it hides in blood cells. It's extremely hard to understand the physical nature of the parasite to the point that you could design something that attacks it.
- [82:15]
B
Okay.
- [82:16]
A
But these guys who are developing a malaria vaccine realized that they could make.
- [82:20]
C
A tiny bus and it would have a bunch of. It, have a class of school children, and then they'd go in and explore.
- [82:26]
B
Your body and find them kind of like osmosis.
- [82:28]
A
Oh, I thought you were sending. We would send a bus to Africa. And I was like, that won't help.
- [82:32]
C
I was talking about the magic school bus. But there is a surprising amount. There is a surprising amount of children's media where tiny ships are.
- [82:39]
B
Yeah, go into.
- [82:41]
C
That's interesting.
- [82:43]
A
Okay, so let's say, you know, we are humans and we want to understand what the hell's going on with malaria. So you start looking at this malaria parasite. It's tiny. And you realize that there's all these proteins around the body of the parasite, right. There's a shitload of them. And you can eventually look at and understand that on the gametes of the parasite, which is basically the dick and balls, that there's a certain protein Socrates in the rat. Let me actually, let me get the number so it.
- [83:13]
B
So it sounds d the dick and balls.
- [83:17]
C
Is this one also made by Siemens?
- [83:21]
A
Well, they're making siemens on the PFS4845 protein. So basically, imagine you're looking at the parasite and you're looking at his dick and balls, but it's blurry and you're like, okay, we know that there's a protein there, right, that's on the body of the parasite. This is the protein it uses to reproduce, Right. If you could, in theory, put a condom on the dick and balls, it would no longer be able to reproduce. Right? Now it's incredibly hard because this thing has all these different life cycles in the body that it's going through. But if you could stop that. It just can't reproduce. Rather than saying, hey, maybe we can stop it at this section of the liver, you say, okay, what if we can identify that? Stop that specifically. And again, if you think about a protein, like a physical shape and you need to find. You need to design a key that goes in, binds to it and blocks it. You can stop them from getting laid.
- [84:10]
B
They've done this for actual mosquitoes, right, don't they? Like, they irradiate a batch of only males, so they're infertile and they send them out in the world. They fuck. But then they. There's no babies produced.
- [84:19]
A
So I want to get higher crops. Some my thoughts on this, but somebody in the comments is going to reply and be like, as a mosquito penis biologist, what Doug said was wrong and I'm not going to speculate.
- [84:29]
C
I would love mosquito penis biologists.
- [84:31]
A
If there's any mosquitoes.
- [84:33]
B
Penis biologists, they certainly have mosquitoes.
- [84:35]
A
It's the same. It's the same guys who have student loans in Australia and live in Vienna. They're studying Mosquito.
- [84:42]
C
We're looking for just a miracle listeners.
- [84:46]
A
No, no, no. It's one guy and we're asking that guy to go expand your horizons a little bit. Stop being selfish. So this is like fundamentally, one of the core things that goes on with drug discovery is you are trying to figure out, like, what is the protein, what is the molecule, what is the thing that is causing a function that is bad. Here's another example in a tumor in the body, right? This tumor cell again is a cell that just keeps dividing rapidly. And the problem is that it looks like one of our own cells. So it's hard to determine that it's bad in the first place. But technically it's got like, it'll over express certain proteins, it'll over express certain antigens, meaning that literally on the surface of the, of the cell, it has more of these proteins like embedded into it. Like, it's like one of those, you know, candies with like nerds on the outside.
- [85:31]
B
A nerd's gummy cluster.
- [85:32]
A
It's like a nerds gummy cluster. So think of, think of the cells as.
- [85:35]
B
Are those dangerous to eat? Is that a disease now?
- [85:38]
A
Well, they're, they're proteins. Just don't put them near your gametes.
- [85:42]
B
Well, okay, in your case, I wasn't thinking about it. My gametes are so big they wouldn't even. Yeah, you wouldn't even notice a nerd.
- [85:49]
C
I want to hear about.
- [85:51]
A
Bro, can we move on from your game? You brought it out, right?
- [85:53]
B
You brought up my gametes.
- [85:55]
A
Okay, so, so this is like fundamentally what goes on in a lot of drug discovery is you understand a function is happening or like this is what a thing looks like that, you know, is bad, but you need to teach the body how to recognize it. But this is very hard to determine the structure of a protein. You have to do something called X ray crystallography or something like that. I don't exactly remember. There's some old bullshit, some bullshit, I don't know, some sciencey shit. But essentially, up until the past few years, the only way that we could figure out how to figure out the structure of a protein is by this extremely meticulous kind of guess and check process where you're crystallizing the protein into a certain structure in a lab. And then you can be like, okay, this is pretty much correct, but that is extraordinarily hard to do. And so I actually want to go to the. Oh, this Is.
- [86:41]
B
Back in the day you could submit your PS3 to help folding at home where they would do this. You could have your PS3 running all day to help test and check these crystallized.
- [86:57]
C
What?
- [86:57]
B
Yeah, yeah. I mean this sounds like alphafold which is like some new.
- [87:00]
A
Yeah. Which is.
- [87:00]
B
We're gonna way better way of doing it. But you used to be able to.
- [87:03]
A
Yeah, this is, this is here. If you pull this up like this is the potential structure of a protein, right. This shit. This is crazy, right? These are these crazy ass. These structures again that all kind of like fold together in this crazy way. And so the way we had of figuring out structures like this is this extremely tough laborious process that you're talking 6 12, maybe 24 months to figure out the structure of a protein. Introduce AlphaFold. AlphaFold came out. Alpha 2 is in 2020 I believe. The newest version, AlphaFold 3 came out last year. This was so impactful that they won the Nobel prize in chemistry for this. AlphaFold is a product of DeepMind which is a subsidiary of Google. So this is basically coming from Google and this is kind of like one of their labs, almost like Bell labs where they're just like cranking out crazy AI shit. And their goal is literally to cure all disease. And so what they have been doing over the past few years is building a deep machine learning system that can learn to figure out just from the ingredients that go into a protein, which is like a list of amino acids, understand how it would fold into crazy ass shapes like this. So it just gets the bare. The kind of like this is what it's made out of and can actually determine this shape. Which are crazy. These are, you know, these crazy things. The way it was able to do that is because of there's like the protein structure data bank which was a giant pool of data all these companies have put together. But they were able to make the system actually just from some basic information about the protein, like the amino acids, determine what this thing actually looks like in detail. So you can imagine now going back to the malaria example, if you've been trying to make this malaria vaccine and you know that dick and balls has the protein that if you stop that you cure the whole thing, but you don't know what it looks like. There's all these challenges that arise with trying to figure out the exact shape of the protein on those gametes. AlphaFold comes in, you can just feed it the amino acids. This did literally happen. This is going on with this malaria vaccine and basically a years long problem of Trying to figure out what exactly this protein is so how they could design a fix for it so that something can come in and stop it is suddenly done in like weeks or days. It was something ridiculously short. I don't know if they said the exact timeline here. I don't remember. So this is like a massive breakthrough. Suddenly. One of the key pieces of the puzzle is now visible. We have the structure for it and it was all made in a few minutes by a now open sourced system. It's shared with the for non commercial use by Google that can dictate and has now it's something like hundreds of thousands of protein structures. Oh no, no, hold on. It predicted 200 million new protein structures in one year. So it's not like, oh, this is cool, it's kind of helping with malaria. It's like the type of things that were holding us back from understanding processes and understanding how to attack certain systems went from a year long process to now basically instantaneous. The ability to now start pursuing drug discovery and drug cures and figure out what's going on in a system just went from like an extremely slow specific thing to like, anybody can do this anywhere. Not only that, AlphaFold3, which is what came out last year and won the Nobel prize, it doesn't just predict the protein structure anymore. It also then simulates how that protein structure will interact with, with other proteins or with molecules or with DNA. So if you're a drug designer, if you are, for example, trying to, you know, come up with a drug that's going to, let's say, attack cancer cells or Alzheimer protein or something like that, before you had a blurry image of what the protein looked like, you came up with an image. You came up with an idea of like, okay, here's my image, here's my idea for what a structure would look like that's going to come in and attack that thing. And you kind of guess and check. You then spend six months crystallizing your idea of what might work in a lab, then put it together with the, you know, with the target and see what happens. If it doesn't work back to the lab, you're back to another like one year iterative loop. Now AlphaFold3, you do that digitally in minutes. So a drug designer going from years for every single iteration of a test to see what might help with malaria or what might help with cancer, or it might help with any of these other things, is now in minutes able to put it through alphafold and say, here's my idea for what might work on attacking the malaria parasite. It simulates, it goes. That didn't really work. And you adjust from there and you're doing cycles at literally I don't know what the math on that is.
- [91:19]
C
I mean that's the magnitude is insane. You're saving so much time in the time scales.
- [91:25]
A
You're talking about unbelievable amounts of time. And not only that, you're making it now so many more people can participate in trying to come up with cures for things. You don't have to be a lab that has X ray crystallography.
- [91:35]
C
I see how you're arriving at your 10 year conclusion. Collusion. Because this is such an escalation at which you could, you can search for these answers dramatic of how we approach.
- [91:46]
A
These things that it is like almost unbelievable.
- [91:48]
C
Yeah. That is, that is so sick. I, I, I mean do you, come on, you might need to steal this villain chair for a second. I don't.
- [91:56]
B
Sure. Yeah, I will.
- [91:57]
C
Yeah.
- [91:58]
A
Yeah. Well the parasite, the malaria is like a living being and maybe it should have a right to kill 600,000.
- [92:03]
B
We asked the malaria what he thinks about this.
- [92:05]
A
Yeah, that's true. And there's more I'll kind of go on. But yeah, initial thoughts for you guys.
- [92:10]
B
Here's my. I mean I actually think this is incredible. I really have no negative thoughts but I'm in a steel man or whatever. I'm gonna try to get you to.
- [92:15]
A
Yeah, yeah.
- [92:16]
B
Okay. So this is from DeepMind. Google website.
- [92:18]
A
Yes.
- [92:18]
B
They have an incentive to make it seem as cool and dope as possible.
- [92:21]
A
Yes.
- [92:22]
B
I've seen many technological advances on company websites that sound fucking incredible. And then we flash forward a few years and nothing's fundamentally changed.
- [92:30]
A
Yes.
- [92:30]
B
And we feel like okay, where did that go? Where was that thing that was supposed to do X?
- [92:34]
C
Yes, that, yeah, that I think that skepticism is super, is super fair, right?
- [92:41]
A
Yes.
- [92:41]
C
The incentive structure to present these things as world changing. Of course solutions as soon as possible is something I totally understand. You know, for I, I have read, you know when a blog post about some new blockchain thing and this is how it's going to change the way we interact with each other. And I get that part of it. But to me the part that is more believable about this is like maybe AlphaFold isn't quite as effective or there isn't all the negative aspects of what they're explaining on this post. Maybe that's true. But the basic idea of using machine learning to shave down the time it takes to analyze data, to cut down on something like revising your approach to curing or attacking like the malaria parasite.
- [93:34]
A
Right.
- [93:35]
C
I, however, whatever the type of disease, whether it be a parasite, a virus or something else that just hold, that core idea holds very true to me. Like, that is my understanding of machine learning, even before AI was like a big craze, is like, this is the type of functionality it would be particularly good for. It can just cram all of this time. It takes a human to do something into something so much shorter and we can more efficiently trial and error our way through solving diseases. And that part of it seems incredibly believable to me. And that is the part that I have faith in is like, maybe it's not alpha fold at the forefront of who ends up being the critical piece of what it takes to attack these problems, but this is an example or a first mover in this. And dude, this is, this is kind of what I was hoping for, is like, this is the type of stuff that I think is so, so sick. It's so much more. It's like I would give up being able to ask my phone passing questions about language or research projects or how, how I feel today for, for this. This is like, to me like the most important thing.
- [94:49]
A
Yes, I by far agree. And actually Dariel Amadai, who I'll talk about briefly later because we've been going on for a while, that's also really what he says is like, look, when you get into medicine and biology, this is the most clear cut of like, AI is going to dramatically improve our lives because we're not talking about, oh my God, lawyers are going to have this fucking AI chat bot. We're not talking about ChatGPT, we're not talking about AI, Art, all this stuff that has a lot more complexity. This is like, we're probably going to cure so many diseases. Demis, the CEO of DeepMind, who's running this and again won the Nobel Prize for chemistry last year for this. Like, there's broad scientific acknowledgment of how incredibly impactful this is. On average, it takes 10 years and billions of dollars to design one drug. We could reduce that down from years to maybe months or maybe even weeks, which sounds incredible today, but that's what people used to think about protein structures. And there's all these examples. Like he's the one who said it'll, you know, in the next five to 10 years we'll have systems that are capable of not only solving important problems or conjectures in science, but coming up with them in the first place, it'll revolutionize human help. I think one day we'll cure all disease with the help of AI. And he said, I think that's within the reach within the next decade or so. I don't see why not. So some of the smartest people on the planet who are winning Nobel prizes for their work are like, this is so dramatic of a leap forward in drug discovery and testing and improving that we might literally cure 95% of the things that have killed billions of humans. Right. And if like, and again, if you think of your whatever, close family member and if they come down with some disease and then this allows us to cure Alzheimer's or cancer or whatever else, like, holy shit, that's a big, big deal. I would, I would give, I would also give up everything else with AI if I could have this to address what you were saying briefly. Yeah, there's absolutely. This isn't some holy grail. First off, acknowledge, want to acknowledge. Alpha Fold is not perfect. It's just really good. So even in this post, right. They talk about how it helped the humans, but it was this major leap forward that probably would not have been possible before. Right. So it'll continue to get better though, over time.
- [96:47]
C
Yeah.
- [96:48]
A
And then the other thing is again, you know, I think the government needs to get involved with this stuff as much as I generally do actually have faith in Google. They've shown to be well intentioned with a lot of things like this. I wouldn't trust them. I don't think they should be the ones, you know, hallmarking this stuff. And so it gets back to the abundance thing of I think the government should be incentivizing centralized development of this stuff. But also it's incredible that private companies, this is a private company that did this and has been working on this and they're continuing to. I listened to a podcast by the CTO of one of the kind of twin subsidiaries that work on this stuff. And he's like, our goal is to have like five to six more of breakthroughs of this scale soon. And we think that will basically, you know, 1000x drug discovery. And there's a lot of just thinking around that. Like another from dario, who's the CEO of Anthropic. It's my guess that powerful AI could at least 10x the rate of discoveries, giving us the next 50 to 100 years of biological progress. In 5 to 10 years, like the rate is just going to skyrocket. But also it's challenging. Yeah. And I mean, I'm so, I Also want to give just really quick people who are like, Doug, you're way too optimistic about AI. This is why, by the way, this is the stuff that gets me excited. It's not ChatGPT. ChatGPT is really cool, I think, but this is the shit where I'm like, oh, my God, we might eliminate cancer. Let me actually quickly go through a cancer scenario. So right now, a, you know, early stage treatment for cancer is called T cell therapy. It's actually car T cell therapy. So again, the core problem with cancer is it's from your body. It's your own cells that have mutated and they're. They're dividing. They're creating too much of themselves. So the body isn't good at recognizing this is a bad thing that needs to be taken out. So what you in theory could do, though, is you could take T cells, like, you know, immune system cells out of the body that are normally supposed to attack cancer cells. And you train it to look for specifically what the cancer cells look like. You put it back into the body, and then they go in. It's like you're. You like, pull out soldiers from the body and you like, send them through like seal camp, like boot camp. And you train them specifically to go fuck up.
- [98:55]
B
Sounds bad.
- [98:56]
A
Tumor cells. And then you put it back in. That is working right now. I don't know where. Oh, this is car T cell therapy. So it's called chimeric antigen receptor therapy. It's named after Hunter. Hunter. And it's literally what they do is they grab a T cell. You can see here on the left. If you pull it up, grab a T cell from your body, they reprogram it. Basically add a little thing to the outside. They like jam a little key on there. Then you send it back into the body and that key fits into the tumor lock. And fucking the body goes, oh, shit, all right, we gotta cure this. And it goes and like expands and works. And this is working right now with leukemia and lymphoma. And it's full on, like curing some of these incredible.
- [99:33]
C
You to hear something bad about this. Not bad that the therapy is. The therapy is very cool, like the functionally. But this is something I learned about car T cell therapy about a year ago because I was talking to my doctor friend about what are some up things in the medicine industry. And the five. So there's five pharma companies that have the. Basically that do car T cell like, treatment. And the five. There's like five different types of the treatment that you can do depending on, I think, the type of like cancer you're fighting and the type of treatment that you need.
- [100:05]
B
Okay.
- [100:06]
C
And these five pharma companies agreed that each one of them would only do one of the therapies. So they have a full monopoly on each of the therapies and they can charge as much as they want. I've learned that. And that's how I learned about CAR T cell therapy.
- [100:21]
A
Yeah, that's, that's some. Yeah.
- [100:24]
C
I mean this is, this is, it's awesome.
- [100:27]
A
Yeah. So let's talk about this. This is. Right, it's cool, right? It's the idea of you take immune cells out of your body, you program them to attack the tumor and then you put them back in. But this has the same problem as any other cancer treatment, which is your, the cancer cells look like healthy cells. And so this same treatment will go attack, it goes into your bloodstream and it helps blood cancer, but then it also attacks healthy parts of your bloodstream. And my understanding is like, you can survive this, but then you need blood transfusions and whatever else because it's killing a bunch of healthy stuff. So even with this, this crazy ass therapy where we're pulling out the soldiers of the immune system, we're training them to look for a specific thing and putting it back in, conceptually that could solve almost anything, right? If you teach the immune system how to go look for this certain type of thing, that's what a vaccine is, right? You're teaching it, okay, this is what to look for. And we're manually fucking grafting it like Elden ring onto the T cell's body and then putting it back in. But cancer is insanely hard. And this might work for blood cancer, but you know, if you have again like pancreatic or prostate cancer or whatever it is, every cancer, not only is it going to be different, they're like morphing. They have all these different ways of expressing themselves. They mutate and so they have different looks. One of the things you can do is they over express meaning on the surface of it. They have way more proteins or antigens than they should, but that's hard to detect of like, oh, it's more than another cell. Right. All you can kind of do is say like you do or do not have this thing, but it's hard to tell. Oh, you have too much of this thing. And so the same type of treatment is not really feasible right now because we don't have a way to realistically design the super, super precise type of soldier. In the immune system that can do all this stuff. But again, that's not impossible. So imagine a world in 10, 20 years where you get cancer and they pull a biopsy of the cancer cells, they look at the tumor cells, and while a human couldn't do this, an AI analyzes the tens of thousands of different proteins on the surface of the tumor and goes, we've actually figured out a pattern. Then once determining that concrete pattern that makes the tumor cell distinct from any other types of cells in the body, you then have the AI design a molecule using tools like AlphaFold to say, we're going to design a type of T cell receptor that is specifically going to fit on this lock, even though that's insanely complex. But we have the data and ability to do it. Because it's an AI, we put that onto the T cells, put that back into the body, and it fully kills the cancer.
- [102:52]
C
Dude, that is.
- [102:53]
A
So that's feasible within our lifetimes. Absolutely. And the challenge here is like, the immense amount of data complexity and processing and the amount of time, but if all of these tools are able to be incorporated into an AI system that for. Can, for example, actually crystallize these CAR receptors and put them on the thing, so the timeline is also reduced. Like, this is. We could cure cancer straight up. And you can imagine this happening for everything else. All the complexities of diseases that we currently deal with, like Alzheimer's, where we kind of don't really know what's going on. If we can crunch the numbers on what's going on with a sample of neurons, an AI could identify a specific way to actually attack it. And this, because of AlphaFold last year, is this isn't fantasy stuff. This is like, yeah, we see a path to this right now in our lifetimes, in 10, 20, 30 years.
- [103:42]
C
I mean, okay, this is. This is the update I was looking for. Something that, you know, what is the concrete progress that's being made now even at the tail end? And what we could call chapter three here, where there are companies that are sort of edging in this direction or introducing technology that is chasing what you're talking about is so significant to me because even in my version of what I was hoping for coming into this is, I thought a lot of these things were further away. I didn't. I didn't know any of the companies that were already working on a lot of these solutions going into this. And yeah, this is. This is to me, the holy grail of like, what machine learning or AI is supposed to be about. This is what you're. This is what it. The end game of it. It's not to like, replace humanity. It's not to create funny pictures of Spider Man. It's to help protect people from disease and suffering.
- [104:40]
A
And then put a funny image of Spider man onto the T cells that goes into your body.
- [104:44]
C
The proteins actually got a tattoo spider. So this is. I mean, this was really wild and it makes me really hopeful. I do think a lot of the brief downsides that we pointed out along the way. I'm curious what other people have to point out here, especially people that are involved in the medical or research industry around these topics. If you have anything to contribute more to this discussion, I'm curious what you have to say. I think to me, most of the downsides that we're talking about doesn't really have to do with the functionality of the technology. It doesn't have to do with, oh, what are the, like, am I worried about it taking over the human soul or something? Like, I'm not worried about that.
- [105:26]
B
No, no, no.
- [105:27]
C
I'm worried about basically just monetization and people's access to it. And if we were to treat this in the best light and imagine a scenario where these technologies are developed distributed through systems that are at least better than the American healthcare system. But, you know, even we talked about the NHS a lot in this episode. The NHS has a ton of problems right now. They're like one of the, you know, one of the most struggling health care systems in like Western Europe. So out. If we can under, if we can view this through the lens of these systems are going to be improved, we're going to make those things better and this is what is going to be deployed through it. That's awesome. The idea that I, I mean, one of my biggest fears is that I really, really don't want to Alzheimer's, I, I think and I got way more.
- [106:21]
A
Scared about cancer the past couple of years. I'm like, oh, fuck, I would really not like that. And everybody in my cancer die. Everybody in my family seems to die of cancer. Alzheimer's with both seem brutal. And it's like, yeah, and very real world that in 10 to 15 years that's gone.
- [106:36]
C
That being on the horizon for like you, your loved ones, like the kids you have, and the idea that that's something that those people don't have to worry about anymore, that's something that worth chasing. That's like the great, the great human.
- [106:48]
A
Achievement or even aging. Starting to understand, like, why does like, you know, there's like tortoises that live like 200 years. We don't, there isn't. We don't have to die at 100 years. So like, as this stuff starts to expand, there's a very real world where we understand better, like why do cells age and can we program that out of them? Is there a DNA change that we can make?
- [107:06]
C
And then South Korea won't collect collapse.
- [107:08]
A
That's true. We keep every South Korean alive and preserved.
- [107:11]
B
They look like K pop stars for.
- [107:12]
A
200 years in like a small booth with a Starcraft and a BTS album. And we just preserve them like, yeah, like the dire wolves in the, in the compound.
- [107:22]
C
So I, I think that's about time for the episode this week. This is, we tried something new. We wanted to delve into like a much larger topic for the duration of the episode.
- [107:33]
A
Also, this was way long, longer than I thought. I thought this would be 45 minutes. So we would have time for other stuff.
- [107:37]
C
So I think in the future, like, we'll, we'll try to make sure that the, the main topic, if we keep doing this leaves a little more room to talk about like other small room. Like me and Brandon brought like very small time things this week to kind of fill the gap of what we thought would be 90 minutes. But this was really compelling and you broke it out in like a really interesting way. I think it's really important that and kind of how you clarified at the beginning. I feel like because AI is such a powerful buzzword right now, people like, because they feel negative about that and then the idea of attaching it to something sensitive like health care, there's this immediate like, ooh, yeah, none of what.
- [108:14]
A
We talked about is fucking chat bots that your doctor is, you know, like, that's going to be a lot of obnoxious stuff in, in health care that may end up being really great, but in the short term it's going to be really obnoxious. But the, the good stuff is actually happening as well. Yeah.
- [108:26]
C
And this is, this is really cool.
- [108:28]
A
Really, really quick. Just some resources if people are interested in more. Dario Amadai I think I've talked about this before. CEO of Anthropic has a blog called Machines of Loving Grace where he talks about his theory on why biology, health and neuroscience will be impacted by AI in a positive way, which also talks about how AI will basically create researchers that help generate new discoveries like CRISPR or MRNA vaccines. And that's why they'll continue to be really impactful beyond just image analysis and Then the other one if you want to freak out. AI 2027. I read this the other day. It's basically a map of how the next couple years of AI development could go. And this single handedly made me much, much more scared of AI. So if you want to be kind of freaked out and undo all of the positive good feelings you maybe developed in this episode, yeah, this article, Dario's article will make you feel like, holy shit, we might actually achieve utopia in our lives. And this one will make you think, I have two years left to live. I need to go eat all my favorite foods tonight.
- [109:23]
C
I want to tack something on the end here. Kind of the last villain chair I could add to the episode. I didn't bring this up earlier in the episode because as far as I understand, before AI or machine learning had hit more of a mainstream, this is already something that was happening. But you reminded me because you mentioned crispr. I think there is something to be said here about when we're catching diseases in advance, where you're doing analytics on embryos and you start making choices about the babies that can be born and can't be born and then those decisions like kind of stumbling away from diseases and more into things that are like cosmetic or choices of how you want your baby to look. I think that's like a really interesting topic to me. The reason why I didn't bring that up in this episode is I actually don't think that's tied specifically to AI because it was something that was already kind of happening prior to these things being mainstream. And I do think that's a potential big downside or thing to look out for when we're talking about these sorts of technology. I just want to bring that up at the end and maybe it's something we get to talk about on the Patreon patreon.com lemonade stand. If you want to join us for an extra episode every week, it comes out on Mondays. Or if you want to join our second tier, we have a book club that our first episode is going to come out pretty soon and the second book is already announced. And we also have been doing a kind of a third bonus show where we respond like more directly to comments and things we feel passionate about. So if you want to check those things out, join the discord. Talk about things there at you got anything to.
- [110:52]
B
That's it. That was great.
- [110:53]
A
Hey, drop a comment below. Who do you think won the debate? Cancer or humanity?
- [110:58]
C
I think I'm locking it in for cancer. See you guys later.
- [111:02]
A
Bye, everybody.
- [111:03]
B
Thanks.