Science Friday - Colorectal Cancer Rates Rising In Young People | What An AI Learns From A Baby

Episode Date: February 13, 2024

Colorectal Cancer Rates Are Rising In Young PeopleGastrointestinal medicine practitioners have noticed something strange in recent years: More and more young people are being diagnosed with colorectal cancer.It used to be incredibly rare for anyone under the age of 50 to be diagnosed with colorectal cancer. Cases were generally limited to people with excess weight who live a sedentary lifestyle. But practitioners are increasingly seeing people in their 40s, 30s, and even 20s without prior risk factors being diagnosed with colorectal cancer.Jennifer Fijor is one nurse practitioner who has seen this rise in cases firsthand at Virginia Mason Franciscan Health in Seattle, Washington. Jennifer has been spreading awareness about this rise on her social media accounts.Jennifer speaks with guest host Kathleen Davis about the warning signs of colorectal cancer, such as sudden changes in bowel movements, and how patients can advocate for themselves to get screened early.What An AI Learns From A Baby’s-Eye View Of The WorldThere’s a lot to learn in the first couple of years of a child’s life—not the least of which is how to talk. But little kids don’t sit down and study a vocabulary book. They soak up language from daily experiences, which are often filled with parents and caregivers saying things like “look at the kitty cat.” Scientists wondered whether an artificial intelligence model could learn about language using a similar strategy—not by being fed a curated set of pictures and words, but by eavesdropping on the day-to-day activities of a small child.They found that associating images and sounds from 60 hours of video captured by a camera mounted on a baby’s head could teach a computer model a set of several dozen basic nouns, such as “car,” “cat,” and “ball.” And the learning was generalizable, meaning that the computer was able to properly identify cars and cats that it had not seen before.Dr. Wai Keen Vong, a research scientist in the Center for Data Science at New York University and one of the authors of a study recently published in the journal Science, joins SciFri’s Kathleen Davis to talk about the research and what it can teach us about learning.Transcripts for each segment will be available the week after the show airs on sciencefriday.com. Subscribe to this podcast. Plus, to stay updated on all things science, sign up for Science Friday's newsletters.

Rates of colorectal cancer have been rising in young people over the past decade. And I remember back in the day somebody being diagnosed in their 30s and I was, you know, everybody was quite shocked. It was a travesty and it was something that was just not heard of. It's Tuesday, February 13th, but just like every day, today is Science Friday. I'm SciFRI producer Kathleen Davis. It used to be that cases of colorectal cancer were limited to sedentary, older people with excess weight. But it's more and more common for healthy people in their 40s, 30s, and even 20s to be diagnosed. We'll talk about that story in just a bit. But first, I spoke to Dr. Weakin

Vong, who ran possibly one of the cutest studies ever done, teaching AI about language via a camera strapped to a two-year-old. Let's take a listen. There's a lot to learn in the first couple of years of a child's life. Not the least of which is how to talk. But little kids don't sit down and just study a vocabulary book. They soak up language from their daily experiences, which are often filled with parents and caregivers asking questions like, do you want some water? Or saying things like, oh, look a kitty cat. Scientists wondered whether an AI model could learn about language the same way, not by being fed a curated set of pictures and words or the entire content of the internet, but by eavesdropping on the day-to-day activities of a two-year-old.

Dr. Weykeen-Vong is a research scientist in the Center for Data Science at New York University and one of the authors of a report on that effort recently published in the journal Science. Welcome to Science Friday, Dr. Vong. Great to speak to you, Catherine. So you trained your AI model on about 60 hours. hours of video. Where did all that video come from? Yeah. So this video was really the efforts of a number of pioneering developmental psychologists and cognitive scientists who kind of realized that we really needed data that

better matched what children really experienced to formulate better theories. And so about a decade ago, a number of researchers got together and started coming up with how to gather this data. How could we get something that resembled sort of the real experiences of a child and produced this sort of that mini camera that they were able to attach for a number of babies' heads and sort of record them over their course of their developing years. Okay. So I'm trying to imagine this for myself. This may be the cutest study ever done. So, I mean, what would it look like if I was looking at this baby with a camera? So the initial set of babies wore the camera on a sort of headband or a strap, But a second set of kids whose data is now being collected involves a sort of more sturdy and

firm helmet. It looks like a home video. You are seeing these kids playing with their toys, their parents are reading books for them. They're having mealtimes or they're out in the playground on a slide or a swing. You're really getting a real sort of first-person glimpse at. You know, what the baby is interested in is sort of like on a second to second level, really. It's incredible. Wow.

So you fed 60 hours of this, what I would imagine is sort of a stream of consciousness data to this AI model, as you said, which could include things like going to the park or talking to a caregiver. What did this model actually learn? Right. So the way the model works is it's trained to learn language by associating what it's seeing with what it's hearing. So in cognitive science, the style of learning is what's known as associative learning, because you're not. you're learning to associate certain words with certain images or certain objects out there. And we're really the first to sort of apply these kinds of theories to, you know, this kind of representative subset of what children are seeing and hearing and sort of seeing what it

could learn. So what it could learn is basically a number of the sort of earliest object words or nouns that children also display when learning a language. So those are often the first words that children learn to speak. and we wanted to see if we could capture that aspect of learning from this data and this model. So what kind of words? I'm imagining like dad-da or mama or cat or dog, things like that? Exactly. Yeah, words like bowl, words like prim or car or chair.

Yeah, these objects sort of pop up sort of almost in every video that you see. And so they're the ones that are relatively straightforward to test for as well. Were there any words that were surprisingly hard for this AI to learn? Yeah, I think one of the, obvious failure loads we saw was that the model just couldn't really acquire the word hand, no matter how we tested it. Hand like your hand. Exactly, exactly. And one reason is that hands sort of appear all over the video data set. So, you know, because the kid is manipulating things or playing with things, they always show up in the video no matter what frame or what words

are being spoken. And so that is a bit of a challenge with how our model operates. And so we didn't find sort of a consistent learning with that word. Did the AI think that hand meant something else or vice versa? Yeah, exactly. So if you look through the data, which I did quite a lot while working on this study, the word hand really only gets spoken when the child was playing at the beach or at the sandpit. And so because of this association between words and objects that is built into our model, the model sort of mistakenly learns to associate the word hand,

with bits of sand instead. That is very interesting. There are all sorts of television shows that, you know, we here are supposed to be educational for young kids. Do you think that the model may learn more from watching Sesame Street or Peppa Pig than it would from an actual child? That's a fantastic question. I don't know.

I do know a number of other researchers from the Netherlands ran a similar study training a similar kind of model but on episodes from Pepper Pig and showed that it was able to learn a similar amount of words as our model. Obviously, children start watching television a little bit later in life. We're primarily interested in language

at the sort of very conception at the very beginning of life and sort of showing how that emerges from this blooming, buzzing confusion that every newborn is experiencing. I mean, to be clear, we aren't saying that this is necessarily how kids' brains actually work, but just that this approach is enough to start to train this AI model.

Is that right? That's right. This work speaks to this larger debate in cognitive science around nature versus nurture. Like what needs to be built in for humans to learn versus what can we acquire from experience. So our model, our work is really focused on that latter aspect, where we actually have very few elements or innate components to our model, and much of the learning is really driven by the video data that's being fed into our model. So I would say it's something like 5% built-in or innate elements and sort of 95% experience.

And I couldn't give you a range on people's opinions in the field of what the right split is, But, you know, I hope our work sort of shifts people's minds to realize that experience is really valuable and we don't really need that much built in on top of that. There are parts of language that aren't just the words, but things like tone. Do you think that you could train the AI to understand, you know, that it's different for, you know, a calm voice to say, watch out for the dog or a scared voice saying, watch out for that dog? Yeah, absolutely.

I think that's something in the future we'd definitely like to try. You know, like I mentioned before, most children's words are these common objects or these concrete nouns. But children also learn words like, uh-oh, quite early on in life, right? And the tone in which that is spoken sort of indicates they're recognizing something about the situation is surprising or a mistake has been made. So I think tone is a huge factor. You know, another thing that many other researchers have looked into is this notion of

child directed speech and recognizing that the way that we often talk to babies is, you know, different from how we talk to adults. We often accentuate certain words or certain phonines. And, you know, other researchers have shown that that also might be a pedagogical strategy to help with learning. If you gave this AI model more hours of video to work with, do you think that it would just get better at working with this set of nouns that it learned? Or would it possibly have a bigger vocabulary? I think the answer is probably both to those. We see that in children as well. We see that children sort of a little bit after the age of one really show this large comprehension boost in their sort of word recognition abilities. And we hope that our model displays, you know, a similar

qualitative effect if we were able to feed in more data. Now, that's something we're actually already working on. So stay tuned for that. Okay. And so far, as you said, been pretty focused on these concrete nouns, but I mean, what's next? Do you do like color or verbs? How do you expand this language understanding to the next level? Yeah, absolutely. I think going to verbs, to color words, to adjectives, to prepositions. I mean, we really want, and in some shape or form to be able to explain all of the different elements of language that children eventually acquire. You know, in this work, we really focused on this multimodal aspect of sort of linking language to vision, because that is really where children get started. But, you know, later on, it seems like children might have other mechanisms available, perhaps similar to the next token prediction objective in language models that might enable sort of acquiring some of these more complex words.

I've looked into this question a little bit for verbs. you know, it's quite funny because you think if a child is learning a word like run, if you look in the videos when the word run is spoken, you know, it's because the kid is running and the camera is really just bobbing up and down. So, you know, feeding that into the model, it's going to have a very different notion of what running is, I think, for most people. Right. Well, that's all the time that we have for today. Dr. Wei Keen-Vong is a research scientist in the Center for Data Science at New York University. Thank you so much for taking the time to talk with us today.

Thank you for all of your lovely questions, Kathleen. It was great to speak to you. Medical providers have noticed a strange trend over the past few years. More and more young people are being diagnosed with colorectal cancer. It used to be really rare for anyone under the age of 50 to be diagnosed with colorectal cancer. But practitioners are increasingly seeing people in their 40s, 30s, and even 20s with this type of cancer. So what's driving these cases and what can people who are concerned about this do? Joining me now to talk about it is my guest, Jennifer Fior, Gastroenterology Nurse Practitioner, at Virginia Mason-Franciccan Health, based in Seattle, Washington. Welcome to Science Friday, Jennifer.

Hi, thanks for having me. It's a pleasure to be here. So nice to have you. So just to be clear, what are we talking about when we say colorectal cancer? So colorectal cancer refers to a cancer of the. colon, which is like the large intestine, or the rectum, which is the very end of the colon, prior to the anus. So it is a cancer that's present within that area. And as far as cancers go, how common is this? So colorectal cancer is becoming an increasing trend. Right now in the United States, colorectal cancer is a third leading cause of cancer-related deaths in men, the fourth in women, but it is the second most common cause of cancer amongst men and women combined.

Wow. So when did it become clear to you as a nurse practitioner that rates of colorectal cancer were rising among young people? If you look at the data, you would actually see that it may suggest that there's been kind of an increase in this trend since the mid-90s. In my practice, though, I've kind of seen it pick up over the last seven years or so. Personally, I've been in GI since 2008. And I remember back in the day, it was really rare to have a colorectal cancer in somebody under the age of 50. I do remember somebody being diagnosed in their 30s, and I was, you know, everybody was quite shocked. It was a travesty, and it was something that was just not heard of. And I even had a friend who was 38 when he was diagnosed with colon cancer, and he was an Asian male

in 2015, and that was even a shock to us. But it seems to becoming more and more prevalent over the the last decade or so. And that's even why they've kind of rolled back the colorectal cancer screening guidelines from 50 to 45 more recently. Do we have any firm answers as to why young people are increasingly affected? I think this is a really difficult question to answer. And I would say it's multifactorial. And there's a lot of theories right now, but no definitive one size fits all answer. most of it kind of comes from our diet and lifestyle, particularly our Westerners lifestyle. We typically have a diet that's really high in processed foods, red meats, high fat foods. Most Americans typically don't get enough fiber, fruits, vegetables, whole grains.

I've read one study that showed the average American gets about 10 to 15 grams of fiber per day when the general recommended goal is anywhere from 20 to 35. So that's one reason. another reason may be some environmental exposures, for example, chemicals, pollutants, or even like pesticides. We've also seen an increase in early childhood obesity, especially in the setting of things like technological advances in social media, so like computers, cell phones, things like that, which has created some physical inactivity. Additionally, there's some theories to suggest that antibiotic exposure, especially the early onset, like prenatal years and, leading up to adolescent period of life will disrupt the gut microbiota and can make a pro-inflammatory state which may make you more susceptible to things like colorectal cancers down the line.

Of course, there's other environmental exposures, tobacco, I'm sure we all know that that one's not good for us, and then excessive alcohol intake as well. In fact, I just did a TikTok on how excessive alcohol intake can disrupt the gut microbiota. So it's, you know, something to definitely pay attention to. Wow. And it's interesting that you talked about those environmental factors potentially playing in here just last week on the show we talked to a cancer researcher about quote unquote hidden carcinogens like air pollution. I want to ask you about the quote unquote typical patient for colorectal cancer before, you know, all these young people started getting diagnosed. I mean, who was sort of like, okay, this is sort of a normal patient. that you would see. I mean, the expectation is always going to be somebody over the age of 50, somebody who typically leads an unhealthy lifestyle. Tobacco, as I mentioned, was a definite risk factor, fairly obese, sedentary lifestyle, just poor diet in general. And that is still a part of

this. But there are some things that we're seeing that, you know, may play some other contributing factors where they're having young, healthy patients come in. And that's kind of the more concerning thing, I think, for most people. So what kind of symptoms should people be looking out for? Oh, that's really important. So at this point, the hard part about colorectal cancer is that sometimes it doesn't rear its head until it's ugly. But with that being said, if you are under the age of 45 and you are not somebody who is

quote unquote eligible for screening colonoscopies, the big things to look out for these red flag symptoms or warning symptoms of change in bowel habits. rectal bleeding. And when I say change in bowel habits, I want to clarify, change in bowel habits over like two week period, three week period, four week period, not like I had an episode yesterday and it changed today. But, you know, anemia of unknown causes, especially in somebody who's a non-menstruating person, bloating, unintentional weight loss. These are all signs that you should at least discuss with your primary care doctor or your gastroenterologist if you already have a relationship with them for further evaluation.

So you mentioned some preventative screenings. Are colonoscopies typically covered by insurance? So a screening colonoscopy, meaning I have no symptoms. I am just looking for the possibility of polyps or growths inside the colon that could potentially become a colon cancer or, of course, colon cancers. Those are usually covered by insurance companies. Currently, the CDC recommends adults beginning at the age of 45 undergo colon cancer screening tests. This can be done via colonoscopies. But that's not the only test to detect the possibility of colon cancers. However, it is not only the best at detecting but preventing colon cancers, and that is the key. So there are some other options for those who are afraid of colonoscopies, such as the at-home stool studies, like a fecal occult test.

And now they're even coming out with blood DNA tests that might pick up on some detection. With that being said, the diagnostic colonoscopies, however, those are the ones where we're performing for not routine screening, meaning I have symptoms like those ones I had just mentioned. So if you're under the age of 45 and you are experiencing these symptoms, it's really important to talk to your doctor, as you may still qualify for a colonoscopy. However, the difference is this is called a diagnostic colonoscopy. What that means really is based upon billing. A diagnostic colonoscopy is subject to out-of-pocket costs, and this is dependent on the insurance company, your copay, you know, what you've already determined for out-of-pocket expenses that year. So it does depend on each individual,

but these are not routinely covered by insurance at 100%. Mm-hmm. So what advice do you have for patients who are freaked out about this? Well, there is a decent reason to be aware. I wouldn't say to freak out quite yet. There are some things that we can do to prevent colon cancer, and that's kind of mitigate those risk factors. So the best advice I can give to people is remain healthy overall, keep good gut health. What does that mean good gut health? I know that's a buzzword for everybody. But what I tend to mean with that is make good dietary and lifestyle choices.

So high fiber diets, rich in fruits, vegetables, whole grains, and a good way to kind of keep that gut bacteria as diverse as possible is by eating the rainbow, making sure, as I mentioned before, to eliminate tobacco, limit out. alcohol, limit those red meats, high fatty foods, highly processed foods. And of course, you know, make sure you get adequate sleep, try to reduce stress. I know that's tough this day and age. But of course, things like that. And then exercise is also important. And the other thing is avoiding unnecessary medications or supplements. Unfortunately with this, there's no magic bully. But one important other thing to know is your family history. Family history plays a large part in this. there has been, you know, increase in these genetic cancers. Our genes haven't changed, but a lot of the stuff that we do or what's in our environment has been able to turn on those genes. So we've

been seeing increase in these hereditary GI cancers such as Lynch syndrome, FAP. Actually, what they determined is those new onset early colorectal cancers, about 20 percent or a little less than 20 percent are due to these genetic variants. So it's really important to know your family history. And What I mean by that is family history, first degree relatives is very important, especially if they're diagnosed under the age of 50, or if you have multiple first degree relatives, meaning parents, siblings, offspring. And then based on your family history of cancer, as GI professionals oftentimes were able to determine if further testing is needing, such as genetic testing, or what cancers you

might be at risk for. The other thing to know is if your parents or your siblings have colon polyps, especially those advanced colon polyps. The general recommendation is to get a colonoscopy 10 years prior to the onset of their colon polyps, especially if they're advanced. To be clear, an advanced colon polyp is not a cancerous polyp, but one that is pre-cancerous and has a higher risk of turning into a colon cancer if left untouched in the colon. There's not as much of a known impact with colon polyps as there is with colon cancers. So it is still important to understand the role of genetics. And of course, it's really complicated.

You are active on social media at gut gals with a Z on TikTok and on Instagram. Why was it so important for you to get on social media and start spreading awareness about colorectal cancer and gut health? Well, I realized I am one individual who's only able to see so many people in a day. and this just gave me a broader reach to people. I do it in my free time after work. I don't get paid for it. You know, it is just a way to kind of reach a wider audience and not only talk about colorectal cancer,

but talk about other GI diseases, you know, digestive health in general, to kind of help prepare people to be an advocate for themselves. It's tough to walk into a doctor's appointment and try to advocate for yourself, especially when you don't know the right questions to ask. So this is just kind of a tool that I would hope that people are able to utilize when they go into their doctor's office.

That's why I want people to have an open conversation if they're experiencing these symptoms. And again, I mentioned in my video, if people are finding that they're having a hard time talking to their provider, sometimes advocating for yourself is really important. And having these tools of information and knowledge is also really important to utilize that in your self-advocacy. That's all the time that we have for now. I'd like to thank my guest, Jennifer Fior, Gastroenterology Nurse Practitioner, at Virginia Mason-Franciscan Health based in Seattle. Thank you so much for joining me.

Thank you so much. I appreciate it. That's all the time that we have for today. A lot of folks helped make the show happen, including Nehima Ahmed, Emma Gomez, Sandy Roberts, Robin Kasmur, and many more. Tomorrow we'll talk to two journalists who dove into the sounds of space and came away with a new understanding of the universe. But for now, I'm SciFRI producer Kathleen Davis. Thanks for listening.