Hi, Shrewd!        Login  
Shrewd'm.com 
A merry & shrewd investing community
Best Of BRK.A | Best Of | Favourites & Replies | All Boards | Post of the Week!
Search BRK.A
Shrewd'm.com Merry shrewd investors
Best Of BRK.A | Best Of | Favourites & Replies | All Boards | Post of the Week!
Search BRK.A


Stocks A to Z / Stocks B / Berkshire Hathaway (BRK.A)
Unthreaded | Threaded | Whole Thread (17) |
Author: ajm101 🐝🐝🐝  😊 😞
Number: of 12539 
Subject: Re: OT: The Future and AI
Date: 09/30/2024 11:02 PM
Post New | Post Reply | Report Post | Recommend It!
No. of Recommendations: 6
> "I heard this almost verbatim from a successful radiologist - "Radiologists that use AI effectively will replace 10, or perhaps even 100, radiologists that don't"."

I worked for a software medical device company that utilized NNs and other algorithms to extract features from physiological sensor data. It was in a technical role but not directly in research. I wouldn't bet on a widespread replacement of physicians with AI models happening any time soon.

The capability exists today to send you an inexpensive wearable device capable of sending a clinical quality ECG signal to the cloud in close to real time. I suspect anyone looking at the raw data in a time x voltage graph might be able to notice atrial fibrillation, or at least be aware something is wrong. They might not know what a p wave is, but if they compared a normal cardiac reading it wouldn't look the same, and if they saw a reference a-fib reading they might be able to correctly identify it by sight without any knowledge of the underlying physiological cause. And computationally, for automated detection, you don't need AI for this (eg, discrete wavelet transform). There are certainly products that augment or replace this with ML models which might be even more accurate, but it isn't required.

Another more practical example might be automated external defibrillators. In order to not kill an otherwise healthy person who has fainted, AEDs do ventricular fibrillation (v-fib) detection, in the device, prior to administering the defibrillation shock. This isn't AI, either, it's closer to signal processing in embedded systems (to my indirect understanding).

The point is that - unlike x-rays or MRIs in radiology - tests for automated diagnosis of serious cardiac conditions can be done without AI, without excessive computational cost, without expensive and scarce equipment, with low patient risk, and without a trained technician. And this is *not being done*. So why not? And why would AI change this?

I don't believe AI will change this in the near future for three reasons. First are problems around medical record access and portability, and resistance to that is well deserved. Given the chance, insurers have not earned a lot of trust for secure or ethical use of data. Second would be liability, because automated (mis)diagnosis is a fear to overcome and challenge to insure. Third is what I'd call the "so what?" problem and it applies equally to radiology as cardiology. Whether a doctor, a classical algorithm, or AI accurately diagnosed you with a condition, somebody would still have to treat it and pay the treatment, and the US has a problem delivering and allocating medical treatment. AI diagnosing a metastatic tumor is not going to materialize the money for chemotherapy or a nearby hospital where it can be administered.

On one hand, I'm still excited by the technical possibilities of AI in medicine. An assistant with a perfect memory of every test result in your charts offering a quick nudge to your human physician about several sub-acute trends in your lab results could catch problems earlier and with better outcomes. On the other hand, I'm worried AI without legal safeguards will be used for anything other than making fewer rich people richer and more poor people poorer, faster.

At any rate I'm sure BRK is already widely using AI at all levels like others suggested. Their insurance and investing competition are have been doing so for a long time.


> "But it was fascinating to see how AI allows really rapid, scaled, inexpensive AB testing to find out what works and doesn't on social media to rouse us a group and create an us vs. them narrative"

We've had such a long run of computer technology improving life (arguably) that it can feel unnatural to consider it as a net negative for humanity. It's worth considering how destabilizing AI and programmatic manipulation of social media has been in the past decade, and how much worse it can get. It might be better to consider the threat of disruption it poses to BRK.
Post New | Post Reply | Report Post | Recommend It!
Print the post
Unthreaded | Threaded | Whole Thread (17) |


Announcements
Berkshire Hathaway FAQ
Contact Shrewd'm
Contact the developer of these message boards.

Best Of BRK.A | Best Of | Favourites & Replies | All Boards | Followed Shrewds