I cannot wait for "AI" to drop out of use. I hate that headlines like this both bolster LLM techbros and say absolutely nothing about the tool/technique/process. I assume this is some form of...
I cannot wait for "AI" to drop out of use. I hate that headlines like this both bolster LLM techbros and say absolutely nothing about the tool/technique/process.
I assume this is some form of machine learning that's probably pretty awesome, but "AI" is so in vogue that they know it'll get headlines and so all encompassing they don't need to actually talk about the tool at all. Just have some soundbites about how great it is.
Erm, the article does make it perfectly understandable what the tool is though? Obviously it's a machine learning algorithm which takes some data that is hard for humans to parse and finds useful...
Erm, the article does make it perfectly understandable what the tool is though?
The groundbreaking technology has been trained on millions of patients and works by extracting more information from a routine electrocardiogram (ECG) than the human eye can typically see.
The traditional ECG, which records electrical activity in the heart, including the rate and rhythm, has been a vital medical tool in diagnosing heart attacks and abnormal heart rhythms for a century.
But it cannot detect heart disease. That requires an echocardiogram, a type of ultrasound scan, which patients often have to wait months for.
Now a team have developed an AI tool that can spot signs of heart failure and heart valve disease – two of the most common forms of heart disease – from ECG results in “the blink of an eye”.
Obviously it's a machine learning algorithm which takes some data that is hard for humans to parse and finds useful patterns in it. Or in common parlance - AI. What term are they supposed to use when this is the one that is used by all and understood by (almost) all?
Besides, they're literally quoting
Dr Ahmed El-Medany, a BHF clinical research fellow, who led the Imperial College London analysis, described the tool as a “superhuman AI” and said the next challenge would be to design handheld AI-led ECG readers for healthcare professionals to use.
"AI" has just become about as devalued as "algorithm" or "program". I'd prefer Machine Learning since that's at least a bit more specific. Those sentences are just repeating the same thing a few...
"AI" has just become about as devalued as "algorithm" or "program". I'd prefer Machine Learning since that's at least a bit more specific.
Those sentences are just repeating the same thing a few times while juggling adjectives. 'Routine', 'traditional', and 'type of ultrasound' (Which... I guess? Kind of? It's receiving signals from inside your body? An ultrasound is a fundamentally different tech/process/method though. ), they're just padding the article defining an ECG again, and again, and again. And then saying an ECG only detects acute problems, but using an ECG they can detect heart disease. The article does seem like it was written by an LLM.
This team is using ECGs as the raw data to fed into whatever ML process they're using to detect chronic cardio problems.
That is very cool. I'm currently wearing a heart monitor because the result of the ECG was, "well nothing's happening right now." Which is a very common result because it's a snapshot of the minute or so that you're wearing the probes.
But in addition to being bottom-tier science reporting saying "a team did something" in 20 paragraphs instead of one, they're feeding into the hype machine of making it sound like someone just got around to asking Claude et. al. about a problem and "AI" gave an answer. Instead of however long it took these researchers to build and train this ML model and get actionable data back.
Why? It looks just like any other reporting on any subject whatsoever. I had to take an advanced English class last year and part of it was journal articles. This is written pretty much in line...
The article does seem like it was written by an LLM.
Why? It looks just like any other reporting on any subject whatsoever. I had to take an advanced English class last year and part of it was journal articles. This is written pretty much in line with the articles we studied and the principles of writing such articles.
Those sentences are just repeating the same thing a few times while juggling adjectives. 'Routine', 'traditional', and 'type of ultrasound' (Which... I guess? Kind of? It's receiving signals from inside your body? An ultrasound is a fundamentally different tech/process/method though. ), they're just padding the article defining an ECG again, and again, and again.
Am I misunderstanding something or is your takeaway from those sentences that the article is claiming ECG (electrocardiogram) is a "type of ultrasound"? Because it's not. They're saying echocardiogram is a type of ultrasound, which is just factually correct, it's an ultrasound of the heart. The whole article and especially the part I quoted is essentially just dry collection of facts. I apologize if I misunderstood what you were saying, but if that is what you were saying, then you just fundamentally didn't understand what was written. Frankly at least an LLM wouldn't have made such a basic error.
making it sound like someone just got around to asking Claude et. al. about a problem and "AI" gave an answer
There is absolutely nothing in the article to suggest any of this. If you associate the term "AI" with "Claude et al" that's entirely on you.
Thank you. I was wanting to respond but honestly I was just in the comments to see if I would be interested in reading the article, so I wasn't going to actually check the article to make sure my...
Thank you. I was wanting to respond but honestly I was just in the comments to see if I would be interested in reading the article, so I wasn't going to actually check the article to make sure my response was accurate.
I feel like your feelings against the term "AI" might be biasing you a bit here to look at this article in the worst lens possible. It didn't ping me as AI-written or poor reporting, it's fairly...
I feel like your feelings against the term "AI" might be biasing you a bit here to look at this article in the worst lens possible. It didn't ping me as AI-written or poor reporting, it's fairly normal for articles to stretch out simple stories way longer than they should be. Even then, I think this article was just fine since it conveyed the necessary information for a layperson to understand the importance.
(Also, the last paragraph has a typo ("hep" instead of "help"), which I think is a big point against it being written with AI.)
Besides all that, I think you're looking at this particular article from the wrong angle. Maybe some other articles are more interested in the technological side of how it works, but this one's purpose seems to be explaining to the average person why this tool is so important.
It opens up explaining what ECGs and echocardiograms are, how ECGs can't detect heart disease, and how there is usually a months-long wait for ultrasounds. That gives context for how valuable it is that this tool can use ECG results to identify people with potential higher risk for specific issues.
All the quotes build on the point that this will allow for faster diagnosis of heart disease (at least for heart failure and heart valve disease). It can't outright diagnose people, but the results can be used to prioritize getting further scans and bypass the standard waiting lists. They emphasize that this can be run on all ECGs run at a hospital, so even people getting a test for something unrelated can get flagged and diagnosed BEFORE issues become evident.
Remove the term "AI", and this article's overall message and value to the public remains the same. I'm fairly certain this article would have been published even without AI hype, because the point is to tell the average person about a new tool in healthcare that can potentially save countless lives.
AI has meant so many different things over the years. I prefer the term “automated statistics”, but I think it’s valid to say AI any time a computer performs statistics and acts on the result....
AI has meant so many different things over the years. I prefer the term “automated statistics”, but I think it’s valid to say AI any time a computer performs statistics and acts on the result. Linear regression, random forests, neural networks, transformers - all AI.
"AI" used to refer to optical character recognition, natural language processing, and computer vision. The modern interpretation started around the 1950's. Historically it's been continuously...
"AI" used to refer to optical character recognition, natural language processing, and computer vision. The modern interpretation started around the 1950's. Historically it's been continuously repurposed for the hottest new research domains. With the rising negative sentiment, I could see Big Tech ditching it and rebranding though.
I have seen several attempts at something like this (diagnosing covid from the sound of coughing, diagnosing whatever from x-rays etc.) and as far as I know none of them made it to practice...
I have seen several attempts at something like this (diagnosing covid from the sound of coughing, diagnosing whatever from x-rays etc.) and as far as I know none of them made it to practice because it was too difficult to separate the actual signs of disease from unrelated variables.
For example, x-rays from a specialized pulmonary clinic look slightly different than from a standard hospital, and x-rays of patients so sick that they cannot stand up and a mobile x-ray unit is used also look different. The model learned to recognize these and correctly rate patients from a pulmonary clinic as more likely to be sick, and patients x-rayed using a mobile unit as highly likely to be sick. After adjusting for these, it was not accurate enough to be useful.
I know that the capabilities can be there theoretically, but I'll believe it when I see it in practice.
For what it's worth, the article emphasizes that this isn't meant to definitively diagnose people but identify signs that someone's at higher risk of certain heart issues. The intent is that...
For what it's worth, the article emphasizes that this isn't meant to definitively diagnose people but identify signs that someone's at higher risk of certain heart issues. The intent is that doctors can use it to flag potential at-risk patients and prioritize getting them ultrasounds for proper diagnosis, since the standard waiting list for those can be months. The article text and quotes place all the emphasis on this speeding up the diagnosis process rather than replacing it.
I don't know how truly accurate this tool is, but this one sounds promising to me just as a means to expedite treatment. Even if it has false flags, the ultrasounds and other scans would ultimately give the doctors more information to work with.
IIRC that's what the older studies aimed to do as well. I'm completely fine with that, it's just that the article doesn't exactly address the biggest potential issues found in these methods so...
IIRC that's what the older studies aimed to do as well. I'm completely fine with that, it's just that the article doesn't exactly address the biggest potential issues found in these methods so far. It is quite possible that ECG readings are generic enough that this is not a problem, but I'm not educated enough to know this.
I cannot wait for "AI" to drop out of use. I hate that headlines like this both bolster LLM techbros and say absolutely nothing about the tool/technique/process.
I assume this is some form of machine learning that's probably pretty awesome, but "AI" is so in vogue that they know it'll get headlines and so all encompassing they don't need to actually talk about the tool at all. Just have some soundbites about how great it is.
Erm, the article does make it perfectly understandable what the tool is though?
Obviously it's a machine learning algorithm which takes some data that is hard for humans to parse and finds useful patterns in it. Or in common parlance - AI. What term are they supposed to use when this is the one that is used by all and understood by (almost) all?
Besides, they're literally quoting
"AI" has just become about as devalued as "algorithm" or "program". I'd prefer Machine Learning since that's at least a bit more specific.
Those sentences are just repeating the same thing a few times while juggling adjectives. 'Routine', 'traditional', and 'type of ultrasound' (Which... I guess? Kind of? It's receiving signals from inside your body? An ultrasound is a fundamentally different tech/process/method though. ), they're just padding the article defining an ECG again, and again, and again. And then saying an ECG only detects acute problems, but using an ECG they can detect heart disease. The article does seem like it was written by an LLM.
This team is using ECGs as the raw data to fed into whatever ML process they're using to detect chronic cardio problems.
That is very cool. I'm currently wearing a heart monitor because the result of the ECG was, "well nothing's happening right now." Which is a very common result because it's a snapshot of the minute or so that you're wearing the probes.
But in addition to being bottom-tier science reporting saying "a team did something" in 20 paragraphs instead of one, they're feeding into the hype machine of making it sound like someone just got around to asking Claude et. al. about a problem and "AI" gave an answer. Instead of however long it took these researchers to build and train this ML model and get actionable data back.
Why? It looks just like any other reporting on any subject whatsoever. I had to take an advanced English class last year and part of it was journal articles. This is written pretty much in line with the articles we studied and the principles of writing such articles.
Am I misunderstanding something or is your takeaway from those sentences that the article is claiming ECG (electrocardiogram) is a "type of ultrasound"? Because it's not. They're saying echocardiogram is a type of ultrasound, which is just factually correct, it's an ultrasound of the heart. The whole article and especially the part I quoted is essentially just dry collection of facts. I apologize if I misunderstood what you were saying, but if that is what you were saying, then you just fundamentally didn't understand what was written. Frankly at least an LLM wouldn't have made such a basic error.
There is absolutely nothing in the article to suggest any of this. If you associate the term "AI" with "Claude et al" that's entirely on you.
Thank you. I was wanting to respond but honestly I was just in the comments to see if I would be interested in reading the article, so I wasn't going to actually check the article to make sure my response was accurate.
I feel like your feelings against the term "AI" might be biasing you a bit here to look at this article in the worst lens possible. It didn't ping me as AI-written or poor reporting, it's fairly normal for articles to stretch out simple stories way longer than they should be. Even then, I think this article was just fine since it conveyed the necessary information for a layperson to understand the importance.
(Also, the last paragraph has a typo ("hep" instead of "help"), which I think is a big point against it being written with AI.)
Besides all that, I think you're looking at this particular article from the wrong angle. Maybe some other articles are more interested in the technological side of how it works, but this one's purpose seems to be explaining to the average person why this tool is so important.
It opens up explaining what ECGs and echocardiograms are, how ECGs can't detect heart disease, and how there is usually a months-long wait for ultrasounds. That gives context for how valuable it is that this tool can use ECG results to identify people with potential higher risk for specific issues.
All the quotes build on the point that this will allow for faster diagnosis of heart disease (at least for heart failure and heart valve disease). It can't outright diagnose people, but the results can be used to prioritize getting further scans and bypass the standard waiting lists. They emphasize that this can be run on all ECGs run at a hospital, so even people getting a test for something unrelated can get flagged and diagnosed BEFORE issues become evident.
Remove the term "AI", and this article's overall message and value to the public remains the same. I'm fairly certain this article would have been published even without AI hype, because the point is to tell the average person about a new tool in healthcare that can potentially save countless lives.
AI has meant so many different things over the years. I prefer the term “automated statistics”, but I think it’s valid to say AI any time a computer performs statistics and acts on the result. Linear regression, random forests, neural networks, transformers - all AI.
I agree. ML is usually a far more precise term with more credibility.
"AI" used to refer to optical character recognition, natural language processing, and computer vision. The modern interpretation started around the 1950's. Historically it's been continuously repurposed for the hottest new research domains. With the rising negative sentiment, I could see Big Tech ditching it and rebranding though.
I have seen several attempts at something like this (diagnosing covid from the sound of coughing, diagnosing whatever from x-rays etc.) and as far as I know none of them made it to practice because it was too difficult to separate the actual signs of disease from unrelated variables.
For example, x-rays from a specialized pulmonary clinic look slightly different than from a standard hospital, and x-rays of patients so sick that they cannot stand up and a mobile x-ray unit is used also look different. The model learned to recognize these and correctly rate patients from a pulmonary clinic as more likely to be sick, and patients x-rayed using a mobile unit as highly likely to be sick. After adjusting for these, it was not accurate enough to be useful.
I know that the capabilities can be there theoretically, but I'll believe it when I see it in practice.
For what it's worth, the article emphasizes that this isn't meant to definitively diagnose people but identify signs that someone's at higher risk of certain heart issues. The intent is that doctors can use it to flag potential at-risk patients and prioritize getting them ultrasounds for proper diagnosis, since the standard waiting list for those can be months. The article text and quotes place all the emphasis on this speeding up the diagnosis process rather than replacing it.
I don't know how truly accurate this tool is, but this one sounds promising to me just as a means to expedite treatment. Even if it has false flags, the ultrasounds and other scans would ultimately give the doctors more information to work with.
IIRC that's what the older studies aimed to do as well. I'm completely fine with that, it's just that the article doesn't exactly address the biggest potential issues found in these methods so far. It is quite possible that ECG readings are generic enough that this is not a problem, but I'm not educated enough to know this.