AI Led To an Increase In Radiologists, Not a Decrease
- Reference: 0180301407
- News link: https://science.slashdot.org/story/25/12/05/2217255/ai-led-to-an-increase-in-radiologists-not-a-decrease
- Source link:
> [A]lmost all of the AI tools in use by healthcare providers today are being used by radiologists, not instead of them. The tools keep getting better, and now match or outperform experienced radiologists even after factoring in false positives or negatives, but the fact that both human and AI remain fallible means it makes far more sense to pair them up than for one to replace the other. Two pairs of eyes can come to a quicker and more accurate judgment, one spotting or correcting something the other missed. And in high-stakes settings where the costs of a mistake can be astronomical, the downside risk from an error by a fully autonomous AI radiologist is huge.
"I find this a fascinating demonstration of why even if AI really can do some of the most high-value parts of someone's job, it doesn't mean displacement (even of those few tasks let alone the job as a whole) is inevitable," concludes John. "Though I also can't help noticing a parallel to driverless cars, which were simply too risky to ever go fully autonomous until they weren't."
Sarah added: "I think the story of radiologists should be a reminder to technologists not to make sweeping assertions about the future of professions they don't intimately understand. If we had indeed stopped training radiologists in 2016, we'd be in a real mess today."
[1] https://slashdot.org/story/25/09/25/1642255/ai-isnt-replacing-radiologists
[2] https://www.ft.com/content/f2e03bd9-af67-45c4-8e1e-79978b5bc48f
Ok but (Score:2)
You need lots of highly experienced radiologists to supervise ai, meaning that you have to train them without ai rather than cheat through school with it
Re: (Score:2)
It's not a LLM. There is no supervision since the software doesn't make decisions. It's probably best classed as a type of image enhancement. It just makes the job quicker and therefore cheaper.
And when something in demand gets cheaper it also get used a lot more.
LLMs do get used in healthcare, for note taking, transcribing and form filling. And that does require the doctor to review the final output. Apparently very effective at speeding up of keeping patient records - And doesn't need huge hardware re
Re: (Score:2)
Then call it what it is: machine learning, not the latest buzword.
You do not want AI examining your X-rays (Score:3)
AI is designed to take shortcuts in order to improve performance. It's already been caught more than once for example appearing to find problems on an X-ray with a very high rate of success when in actuality it had just picked up on a simple pattern where for example something is dumb as a ruler was included on the X-rays that had the problems and wasn't included on an X-ray that didn't...
Not that any of us have any say in this whatsoever. AI bullshit is going to dominate everything whether we like it or not. Just like how the price of ram has increased by 5 to 10 times and we all just have to suck it down.
We have a very small window left to the side of we are going to live in a society where around 2,000 people get to decide how we live. And we need to decide if having the girl that hands us our coffee say Merry Christmas is worth giving up everything else to those 2,000 people.
Re: (Score:2)
> It is absolutely crazy that we are all very very soon going to lose access to electricity
Calm down. Total AI power consumption (all forms of AL, both training and inference) for 2025 will be in the ballpark of 50-60TWh. Video gaming consumes about 350TWh/year, and growing. The world consumes ~25000 TWh/yr in electricity. And electricity is only 1/5th of global energy consumption.
AI datacentres are certainly a big deal to the local grid where they're located - in the same way that any major industry is a
Re: (Score:2)
You gotta log in if you think anybody is gonna read that long of a rant.
Re: (Score:2)
That "ruler study" was ancient. It's mentioned in peer review at least as early as 2018, and might be even older.
Believe it or not, people in the field are familiar with these sorts of things that you just read about.
Re: (Score:2)
Actually, I have more confidence in the result an AI would produce, than the result a lot of human doctors. Not that human doctors are bad or incompetent, it's just that they get tired, they work long shifts, they get in a hurry. AI just keeps going.
That's not to say that AI doesn't need supervision, it does. But as an assistant to, say, do initial screenings, I like the concept a lot.
Sounds like a standard medical scam. (Score:2)
They realized they could process more xrays, so instead of fewer people they ordered more, likely unnecessary, xrays. And you wonder why your insurance just keeps going up.
Re: (Score:2)
My insurance keeps going up because private insurance in America has a monopoly on access to healthcare so they can charge whatever they want until the public gets so fed up they demand a single pair of healthcare system.
If things continue the way they're going with voter suppression and right wing extremists buying up the voting machine companies I don't think it'll matter anymore and then that will be the end of that. About 10% of the country will be allowed to have health care and odds are you won't
Re: Sounds like a standard medical scam. (Score:2)
That is not true. Medical insurance margins are small, and they are legally required to spend a minimum of about 80% on care, higher for employer insurance with a high number of employees. The problem is the underlying costs, and the only way to fix it will be for doctors, drug companies, etc to take a massive haircut and earn less money.
Re: (Score:2)
What do you think drives those prices up? It's like house loans - Make it too easy to transfer the costs to the future and you immediately incentivise charging more in the present.
Re: (Score:2)
> spend a minimum of about 80% on care
Indeed, so the more the insurance cartel can charge the more profit they can suck in. That's not a hard equation. Perhaps if there were actually some minimal amount of competition allowed in the business prices might be lower, but that's just crazy talk . . .
Re: (Score:2)
"a single pair of healthcare system"?
I am actually for AI reviews (Score:3)
My mother was just diagnosed with HER2 breast cancer. She actually felt the lump and used her own echo machine to scan it before going to her GP. The problem though is that in June she had received a CT scan for measuring an aneurysm in one of her arteries. This CT scan also included the chest. After re-reviewing the CT scan it was plain as day that there was a mass in her right breast. The radiologist missed this incidental finding when reviewing the CT and she went MONTHS without starting treatment. If there was a secondary layer of AI review then I believe it would have been found in June and addressed earlier. Now she's fighting the healthcare system to accelerate her treatment because of this miss. I am as skeptical as any other of LLMs and claims of the next coming of jesus christ, but even with a high false positive rate these kinds of AI re-reviews of CTs, PETs , Xrays or echos could be a valuable tool in evidence based medicine.
Re: (Score:2)
Yep, me too. Doctors work long hours, they get tired, they get in a hurry, they miss unusual signs. AI doesn't suffer from these shortcomings.
Way too early, way too primitive (Score:2)
The current "AI" is a predictive search engine. It's not AI at all. It looks at something and analyzes what it thinks the result should be. The next gen AI we have, maybe 5-10 years from now, will be the one that starts replacing highly-trained jobs. Right now, you still need a person because of hallucinations, malfunctions, bugs, etc. In a decade, there will be AI machines running on quantum processors that examine everything and give correct answers 99.99% of the time. Once that happens, then you will see
Re: (Score:2)
> The current "AI" is a predictive engine.
And *you* are a predictive engine as well; prediction is where the error metric for learning comes from. (I removed the word "search" from both because neither work by "search". Neither you nor LLMs are databases)
> It looks at something and analyzes what it thinks the result should be.
And that's not AI why?
AI is, and has always been, the field of tasks that are traditionally hard for computers but easy for humans. There is no question that these are a massive leap fo
Re: (Score:2)
Don't mistake ChatGTP as representative of all types of AI, because it's not. And you shouldn't mistake "highly-trained jobs" as being only ones requiring extensive education, it takes years of training to run the largest heavy equipment but China already has fully automated open pit mining. AI is automating drafting of building plans, logistical planning, and purchasing, all fields where 'hallucinations' would be catastrophic and yet it doesn't occur. The lack of education about this exceedingly importa
Re: (Score:2)
More importantly, these uses are not using massive datacentres to perform their function. It runs fine on local hardware.
... and (Score:2)
Jevons Intensifies
What Everyone Is Getting Wrong About AI And Jobs (Score:2)
[1]https://www.youtube.com/watch?... [youtube.com]
Jevon's Paradox
From radiology to software engineering, the pattern repeats: as technology makes tasks cheaper and faster, demand for human creativity and judgment grows.
YC's Garry Tan explores what history, economics, and real companies show us— that technology doesn't replace people, it redefines what we can do.
[1] https://www.youtube.com/watch?v=IqwSb2hO1jE
Re: (Score:2)
And what about the half of the population which has an IQ lower than 100? Dumb people need to eat too, and if they can't they pick up pitchforks and torches (metaphorically). When robots run by AIs are washing dishes and picking strawberries the people who previously did those jobs are not going to be doing things requiring "human creativity and judgement".
Sure (Score:5, Informative)
This is the good kind of AI. This isn't a LLM.
Re: (Score:2)
Thanks for pointing that out, far too many people think that ChatGPT is representative of all AI. There are a LOT of exceedingly important uses AIs of various types have been put to, this being one one of them. Robotics for example relies on it completely, the days of writing thousands of lines of code just to get your robotic quadruped to walk in a straight line on a level floor are gone (thank goodness). Atlas can do a backflip and Picklebot can unload a truck of random boxes because they are controlle
Re: (Score:3)
Most of these new AI tools have gained their new levels of performance by incorporating Transformers in some form or another, in part or in whole. Transformers is the backend of LLMs.
Even in cases where Transformers isn't used these days, often it's imitated. For example, the top leaderboards in vision models are a mix of ViTs (Vision Transformers) and hybrids (CNN + transformers), but there are still some "pure CNNs" that are high up. But the best performing "pure CNNs" these days use techniques modeled
Re: (Score:2)
I don't see any reason not to trust the Autobots, lets just put it charge of everything then!
Re: (Score:2)
Nonsense. It takes years to train up a radiologist. It's supply and demand. Nothing to do with AI.
Re: Sure (Score:1)
Can AI help me to do my own radiology at home, and interpret it without having to bother real humans with my insignificant problems?
Re: (Score:2)
Yep. There are ultrasound wands that can plug into your phone, and AIs that can analyze the scan, for under $3000.
[1]https://spectrum.ieee.org/mems... [ieee.org]
[1] https://spectrum.ieee.org/mems-ultrasound-history
Re: Sure (Score:1)
So those who crave human interaction can continue to visit real doctors and be social, and those of us who become more suicidal with more human interaction can still get some kind of medical care and if it kills us, who will miss us?
Re: (Score:2)
Does that really matter? A radiology AI, a graphical AI, a video AI, an LLM--they all work on the same underlying principles.