Today you’ll learn about how foreign accent syndrome works, how an AI algorithm can create a physical copy of a mental picture, and how invasive marsh grass is taking over China’s coastline.
Today you’ll learn about how foreign accent syndrome works, how an AI algorithm can create a physical copy of a mental picture, and how invasive marsh grass is taking over China’s coastline.
Foreign Accent Syndrome
AI Reads Minds
Alien Marsh
Follow Curiosity Daily on your favorite podcast app to get smarter with Calli and Nate — for free! Still curious? Get exclusive science shows, nature documentaries, and more real-life entertainment on discovery+! Go to https://discoveryplus.com/curiosity to start your 7-day free trial. discovery+ is currently only available for US subscribers.
Find episode transcripts here: https://curiosity-daily-4e53644e.simplecast.com/episodes/foreign-accent-syndrome-ai-reads-minds-alien-marsh
NATE: Hi! You’re about to get smarter in just a few minutes with Curiosity Daily from Discovery. Time flies when you’re learnin’ super cool stuff. I’m Nate.
CALLI: And I’m Calli. If you’re dropping in for the first time, welcome to Curiosity, where we aim to blow your mind by helping you to grow your mind. If you’re a loyal listener, welcome back!
NATE: Today you’ll learn about how foreign accent syndrome works, how an AI algorithm can create a physical copy of a mental picture, and how invasive marsh grass is taking over China’s coastline.
CALLI: Without further ado, let’s satisfy some curiosity!
[SFX: WHOOSH]
NATE: One of the scariest parts about surgery of any kind is the fear of unexpected side effects. And one American man got a little more than he bargained for after his surgery when he permanently developed an Irish accent.
CALLI: Sorry.
NATE: What? No, don't. Don't.
CALLI: I can't actually do an Irish accent, so. Okay. On a more serious note, depending on the type of surgery you get, the most common side effects are usually things like nausea, vomiting, soreness, pain, thirst, stuff like that. But I've never heard of somebody getting a new accent after surgery.
NATE: So it's actually a little more common than you might think and is a fully diagnosable condition known as Foreign Accent Syndrome or FAS for short. It's a speech disorder that causes a very sudden change to somebody's speech patterns. And the name is sort of misleading, too, since you're not technically adopting a full foreign sounding accent. Rather, it's a sort of neurological dysfunction that alters a person's natural speech patterns in a way that can sound like a slightly different and maybe even exotic accent. Most of the time it's caused by a traumatic brain injury, but in this case, it was an effect of surgery. And there's another example from back in 2018, an American woman was suffering from severe migraines and then began speaking in what others described as a British accent. And she never stopped.
CALLI: Had she ever been to Britain?
NATE: Nope. She was a former beauty queen who had never even left the country. There have been only 110 confirmed cases of FAS since its discovery in 1907. So the migraine story was definitely interesting as it didn't come from the previously mentioned causes. The FAS as a result of surgery is also very interesting because it doesn't fit properly with those causes either.
CALLI: Huh. How so?
NATE: So this man had surgery to treat his prostate cancer, which as we know, doesn't have much to do with the brain. He is the only recorded patient with FAS that has ever had their condition linked to prostate cancer. He was also a 50 year old American who had never been to Ireland, and he didn't have anything wrong in his brain that could be linked to the brain damage that could cause FAS. He had no abnormalities in his brain, a clear psychiatric history. By all accounts, he was neurologically healthy.
CALLI: Okay. Wow. But even if he had never been to Ireland, was he, like, interested in Ireland? Like, I know there's people who put on the top of the morning infection, even if they've never even left their hometown.
NATE: So the man did live in England around 30 years ago when he was in his twenties, and he did have friends from Ireland, but it's not clear if he had any interest in Ireland. But according to his case study, he had never even spoken publicly with an Irish accent before and the accent became uncontrollable. The man would hold conversations regardless of context in an Irish brogue, and eventually it became persistent, meaning he could only speak in an Irish accent.
CALLI: Okay, this might be a bit of a hot take, but that actually sounds like a complete nightmare to me. Like on top of the anxiety that comes from both cancer and surgery, you also have to deal with the reality that every time you speak from now on, you're not going to sound like yourself.
NATE: Well, not to make you even sadder, but unfortunately, the man's cancer developed into small cell neuroendocrine prostate cancer, which is very aggressive, and he passed away as a result. He continued speaking with the accent until his death. It's thought that it's very likely this particular voice change came about because of something called paraneoplastic neurological disorder or PND for short, which is a condition where damage to the brain is linked to cancer somewhere else in the body.
CALLI: That's awful. All right. So can FAS typically go away on its own?
NATE: According to the study? Yeah. The doctors concluded that it's really important for physicians moving forward to recognize that FAS is a sign that something is very wrong. Of the 110 people who have had confirmed FAS this man marks only the third whose syndrome was related to a cancer of any kind. So knowing that there is a pattern between FAS and diseases or injuries that aren't common is super helpful in understanding the relationship between the brain and the body. Case studies on people with FAS have been very helpful in identifying unexpected symptoms, rare side effects, or even innovative approaches to treatment. But more research is needed on FAS in general, especially when it comes to understanding the possible links between a FAS and how it progresses.
[SFX: WHOOSH]
CALLI: All right, Nate, you and I love playing with image generators. They are a ton of fun.
NATE: This is true. I have I have spent a decent amount of time using a mid journey, which is one of the options out there. It's wildly entertaining.
CALLI: It is so much fun. I use it for my DND campaigns. But anyways, we've talked a lot, even just earlier this week about how artificial intelligence isn't as good as human intelligence for many reasons. But there is one thing that it excels in. It can recreate the images we picture when we close our eyes.
NATE: So, for example, when I close my eyes and picture myself riding a unicorn, AI can recreate that.
CALLI: I love that that's where you went with this. But yes, there was a recent study done that demonstrates the ways in which AI can read brain scans and recreate surprisingly realistic versions of the image you're seeing in your mind. This is something researchers have been able to do for a while when trying to have people recall an image of a human face or a place. But this is the first time an AI algorithm has been used to do this.
NATE: In AI algorithm.
CALLI: Yes. Okay. So kind of like DALLE-2, which was an algorithm that went viral in 2022 due to its shockingly vivid image generation. So, for example, if you typed Nate from Curiosity daily on a unicorn, it would create an image of you on a unicorn. That algorithm functions off of being fed text prompts for whatever you want it to make before it pulls from millions of photographs from all over the Internet. This one's a little different, though, because in addition to the text and the images being pulled from, there's a third component: data from brain scans.
NATE: That seems like a pretty different dataset.
CALLI: It really is. So let's start by talking about how image perception worked in the brain. We have the occipital and temporal lobes, right? If you see an image, the temporal lobes register the information about the contents of the image, like people, objects or scenery. The occipital lobe registers information about the layout and perspective, like how big or small something is or where it's placed. These parts of the brains have changes in blood flow when you're looking at an image. So if you run a functional magnetic resonance imaging brain scan or and fMRI for short, it records those peaks in the brain activity and voila, an imitation image is printed.
NATE: Okay, so how did they input that information into the AI?
CALLI: So a group out of Osaka University in Japan took a system called stable diffusion, which is similar to the DALLE-2 system, and input some additional training by linking the images with text descriptions of what was being depicted as well as using previous brain scans as references. This allowed the AI to learn what the brain scans were depicting through both text and images.
NATE: I can't even imagine how many photos you need to have this algorithm pull from in order to manifest something from the human mind.
CALLI: Well, what's interesting about this, when compared to other brain scan A.I. algorithms is the small sample size used. The team used brain scans from only four participants who each looked at 10,000 photos.
NATE: Hmm. So basically, if I just look through my camera roll on my phone. How do these photos even look, though?
CALLI: Surprisingly realistic, but it wasn't at first. All of the images start off as noise, kind of like TV static. And as the image evolves, each piece of static is replaced by a different feature as the brain activity patterns are compared against the training dataset. The imagery creates the contents of the picture. So let's go back to your unicorn example. It recreates the contents, i.e. you on a unicorn. The layout of the image or where you and the unicorn are in the frame of the vision and the perspective or what angle. I'm looking at you from where I am when I see you. That being said, there were some issues with recreating objects. One of their examples was a clock tower. When they ran it through the algorithm, it created more of an abstract figure.
NATE: I mean, a clock tower is probably one of the most iconic shapes in kind of whatever image it's in. So it sounds to me like this experiment's a bit of a wash.
CALLI: Okay. I wouldn't be so sure. It seems like the more information you input into the program, the more accurate it becomes. So what the researchers did is they decided to link keywords from the image captions on the Minnesota fMRI data set so that it matches the image and kind of locks in what the brain activity looks like when you see that image. So, for example, if one of the training photos had a clock tower, the brain activity pattern would be associated with that object with the keyword clock tower. If the system saw the same brain pattern again, it would put the keyword into stable diffusion's normal text image generator and a clock tower would be added to the recreated image, resulting in a convincing imitation of the real photo.
NATE: That is really interesting and it is also a little scary.
CALLI: Okay. It's important to remember that these keywords are only associated with things the participants have already seen. So for example, I can't just type Nate riding a unicorn and it will replace a clock tower with you writing a unicorn because nobody's ever seen that. So the algorithm can only understand that a certain object is present if the brain pattern matches one you've already seen in the training data. It can't recreate anything it hasn't seen before. They ran the test through a bunch of other pictures of objects, including a toy bear, airplane, clock and train. And in each case, a convincing imitation was made.
NATE: I mean, that is awesome. But honestly, it seems pretty limited, especially since it relies on such a limited data set.
CALLI: For now, it kind of is, and it'll be a little while before this becomes accessible tech. The system was tested on only four people after all, so expanding it to anyone else would take retraining the system to match their brain scans. Still, though, these models are unprecedented. They're able to create realistic images at a scale that no other algorithm presently can handle. With further refining, this technology could be used to do things like intercept thoughts and dreams, or even understand how other animals perceive reality through their eyes. Making For this rare time on our show where we get to show that A.I. has something that the human mind may never have.
[SFX: WHOOSH]
NATE: Good news. China has gone green. Bad news? It's because of an invasive weed that's growing faster than the country can keep up with. More good news, though. There is a plan to get rid of it. And it's being called the largest action plan for wetland invasive species control in China and maybe even the world.
CALLI: Okay. Invasive species. It does bother me when I find like a beautiful plant and then figure out it's actually invasive and is destroying other things that I love.
NATE: Yeah, I kind of hate knowing that, like, English ivy is invasive. Oh, yeah. But anyway, did you know that China has over 11,000 miles or 18,000 kilometers of coastline? Well, the problem is that Spartina Alterniflora, also known as smooth cord grass, has grown over practically every bit of coastline in the country. It's tall, green, thick, and it's putting migrating birds at risk because they can't access the marshes easily to eat, while also clogging all the shipping channels and ruining China's clam farms. Other species of shellfish, as well as a ton of waterbirds, are affected by the invasion.
CALLI: Back up a bit. Spartina alterniflora. Sounds familiar. Isn't that like the really tall reed like grass that's all over the East coast of the United States?
NATE: It is. But back in 1979, it was brought to China to help make some of their tidal mudflats a bit more stable so they could be turned into something like farmland or land that could be developed for commercial properties. This Spartina project was a success, but in a very. Be careful what you wish for twist. It was too successful because the Spartina kept spreading and now outnumbers all the native plants that provide food for any native species in the area. To put this in perspective, Spartina now covers over 262 square miles of land in China. That's almost as big as New York City.
CALLI: I can't even fathom that many weeds in one place.
NATE: Thankfully, China has a plan to stop the pesky Spartina problem once and for all. A plan that will be so effective it could eliminate as much as 90% of it.
CALLI: Oh, very nice. Okay. How do you even begin to get rid of that many weeds?
NATE: It's not an easy plan to define. And so to understand how it works, we got to look at how they've dealt with Spartina in the past. Back in 2001, Spartina was planted at the Chongming Dongtan National Nature Reserve in China, and almost immediately it spread quickly, ruining the habitat for a lot of fish and birds. So to remove the weed, engineers built a seawall and flooded the wetland to drown the grass. This project was successful in getting rid of nearly 95% of the wild Spartina across nearly nine square miles or 2400 hectares, which led to plant and bird populations coming back but at a huge price since the whole project cost roughly $150 million. So it's not going to be easy and it's not going to be cheap.
CALLI:Yeah. Seriously, is there any way they could have done something cheaper? Like could you dig it up or use a herbicide?
NATE: There's pros and cons to each of those methods, especially considering how much of it there is. So, for example, one possible solution is releasing insects that eat weeds, which is a tried and true method known as bio control. The problem here is that no known insects can eat Spartina fast enough. We could also try digging up the Spartina, but that would disturb the habitats of any creatures dwelling there. And also that would only be a short term solution since Spartina grows back very fast. Herbicides are very effective but have to be used year after year every year and even flooding the method they successfully used that Chongming can starve the soil of oxygen, which would kill all the worms and other creatures living in it.
CALLI: Gotcha. So nothing will work?
NATE: Well, actually everything works, which is why the official suggestion is for the Chinese government to combine methods to figure out the best way forward through each individual situation. Experts think this could be a feasible plan, but the country needs to be focused first and foremost on preventing regrowth by targeting surviving Spartina hangers-on saying you have to hunt them down one at a time.
CALLI: He kind of makes it sound like hunting clickers on an episode of The Last of US.
NATE: It's no joke, though. Even one remaining Spartina plant could compromise the recovery process quickly to the point that New Zealand dealt with their Spartina problem by using drones and trained dogs to find patches or even lone plants, which is insanely impressive. Obviously, the Chinese coastline is much bigger than that in New Zealand. But what I'm trying to say is there are a lot of options available.
CALLI: Crazy. Okay, so what is the big deal here? Like, I don't want to sound rude, but why should we care about this?
NATE: Because as you mentioned at the beginning, Spartina can be found all over the east coast of the U.S. as well as many other areas. It's an invasive species. And if China can get their Spartina problem under control, so can other areas dealing with Spartina or other invasive species. As the researchers on this problem, same conclusion. Maybe we'll look again at all the other problematic invasives and say, well, if it can be done with Spartina, why not?
[SFX: WHOOSH]
NATE: Let’s recap what we learned today to wrap up.
NATE: If your friend starts speaking in an accent that isn’t their own, and they claim to have no control over it - they might not be messing with you, and something might be seriously wrong. Foreign Accent Syndrome is a rare diagnosis for this very phenomenon, and happened recently to an American man who began speaking uncontrollably in an Irish accent. It turned out that his brain activity was being affected by a cancer that began in his prostate, making him the first man to have FAS linked to prostate cancer.
CALLI: Have you ever struggled to recreate an image you see in your mind? A new AI algorithm might have the answer for that! An altered version of the Stable Diffusion AI model has been used to recreate images people have seen - from their mind to the page. And the results are shockingly realistic. Right now, the sample data comes from a small population, so more work will need to be done to make this accessible on a large scale - but one day, the tech could be responsible for everything from realizing your dreams to even understanding animals’ thoughts.
NATE: We’ve always dreamt of a greener planet, but in China’s case, unfortunately, that greenness comes at a cost. Spartina alterniflora is an alien species that has ravaged China’s coastlines over the past few decades, but thankfully, a new plan has been set in motion to eradicate the weed. It will cost hundreds of millions of dollars, but through a combination of herbicides, digging, flooding, and biocontrol, China will soon eliminate as much as 90% of their little green problem!