SurvivingInfidelity.com^®

In searching for a good medical or dental practitioner, would you be more inclined to choose an office that claims to have adopted AI diagnostics, or would that practitioner's claim have you wondering what problems their AI might flag that possibly could over-complicate your preventive treatment plan? I can see advantages AND disadvantages. What do you think?

It depends on how AI is used, IMO. AI is probabilistic, I believe, which means it will be wrong some of the time. I'd want someone who treats AI very critically....

I remember reading a story a few months ago about a middle school teacher who gave her students an assignment of using AI to write a paper on something historical or scientific (something that can be proven as factual and not opinion). Once they turned those in, the not-in-the-know students were then randomly given their papers back to student other than themselves, and their next assignment was to find inaccuracies in the paper they were now reading.

It was an eye-opener and a strong message about how foolish it is to use AI for even factually-proven based information.

It can be a jumping off point, but I would not want my PCP using it for my health. I'd want them to use the scientifically evidence-proven information that's out there to treat me, not some general population that I might might or might not have something in common with.

I agree with both of you. I just came across a new dental practice here in my small town whose website touts his use of AI to analyze xrays and etc. for possible brewing periodontal issues in early stages or for maybe a hidden crack in a tooth root. Sounds like a good thing, catching problems early, BUT what happens if for example the tooth isn't in fact a fracture, you just have a weird bone fissure over which the xray image density looks different and AI states it could suggest fracture? What will the dentist recommend, extraction? Yikes!

Or as with EKGs, where I have learned to ignore some dire-sounding summary statements, like "heart failure," "former M.I., age unknown" after my cardiologists have dismissed those statements as flagging potential problems I don't in fact have, and were possibly caused by technical error in where the EKG leads got placed on me.

So I guess the real question becomes: to what extent does a particular medical or dental practice rely on this tool? We wouldn't know until we asked, and it made me ask myself if I really want to sign up with the new doctor. I like it when they utilize advanced scanning imagery, but....

There was just a long article on a newsfeed yesterday that was warning people in information careers how the newest versions of AI lately deployed are so much improved in accuracy that the AI is writing its own code for the next generation of itself! Implying that soon the AI will be calling the shots for the doctors....yeah, no.

Using AI to some degree is happening already. My medical practices have the ability to use an AI tool that basically records the visit and the pulls the info into the right places in the notes which has proven touch more accurate than dictated or talks to text programs which drs have been using for more than a decade. This has improved efficiency for the practitioners.

My dental off8ce is using AI to help interpret trays, and nailed my cracked tooth that needed a root canal vs one I broke that just needed a crown. They have all kinds of wild state of the art stuf. I drive them nuts because I have so many questions.

But yes there definitely needs to be a human eye on things too to correct errors and make sure it's correct.

I can see both the threat and the potential in AI...

One of the first uses of AI was with IBM’s supercomputer over 2 decades ago (Deep Blue or something like that...) where it was fed medical files from thousands of cancer patients, along with published scientific papers on developing diagnosis, treatments and medical advances. A group of highly respected doctors first used the feedback to compare their recommendations to past instances compared to what the AI suggested and saw that in a significant number of cases AI gave the same diagnosis as they had.
They then developed this so that the "success" rate of AI diagnosis improved to a comparable level as they had offered on these past cases.
When they then experimented with using new, undiagnosed cases they saw that AI sometimes gave a different result to what these specialist doctors had diagnosed. When they looked into the cause of this, then it turns out that the doctors were not capable of keeping up-to-date on all new published papers with new info, new treatments and so on. The AI recommendations were basically built on better info than they had, and once they read the same papers they agreed with the AI result.

At that time – over 20 years ago – the team of doctors used AI to "second guess" their results: When a team had gone through the data and given their diagnosis, they would confirm it with AI. If AI gave a different result, they would assign it to someone that would examine why. Often it was because of newly published data, that if the doctors had been aware of would have impacted their decision too.

But that was a team of highly educated, respected and experienced doctors, along with one of the most powerful computers of it’s time, along with a powerhouse company of programmers and specialists to customize the whole system to their advantage.

--
Then we have what’s available to us the common people...

I read an article from a prominent professor in ethics and philosophy. A survey indicated that the quality of essays and papers of college students has dropped with AI. Both in accuracy and also in language used.
Keep in mind how (a lot of) AI works: It searches the net for what is commonly perceived as factual. Like if there are only 5 sites or references to the Apolo moon-landings and 2 of them were conspiracy-based sites, then AI would probably answer questions on the subject with something like "Most think man has walked on the Moon, but many doubt it", when factually it’s probably 99.9% that believe it happened, and 0.01 that doubt it.

It then bases it’s answers on the texts it finds. When students were using edited paper books or edited scientific papers the wording was based on the edited content. Now when anyone can publish and editing, proof-reading, fact-checking, references etc are basically optional the quality of the replies AI finds drop... When a student then copy/pastes the result they are using inferior material compared to the text that was proof-read, edited and verified.

-

I experimented with AI a few weeks ago. I asked AI what plastic storage boxes available at a certain retailer were best to store LP records (anyone remember them?). Got a very specific reply along with the item number for a certain container. When I tried to put my old LP collection into one, it was too small by ¾ of an inch.
My mistake was that I asked for storage of LP records... Not for LP records in the envelope/sleeve/album they come in. AI didn’t seem to question or know that LP’s are seldom stored "naked" or exposed.
You need to be quite detailed and specific in your questions.

Bigger that is so interesting, and sadly what I worry about...my father was an early computer analyst for the defense department and back in the Main Frame era, the saying was GAGO: Garbage In, Garbage Out. You make the point that we haven't moved that far and added importantly to my understanding of AI source knowledge corruption.

Tushnurse, my hospital went to AI transcription of visit notes last year, and every time I've read what I supposedly reported to my doctor it gets some detail a little "off," but not badly enough I feel the need to correct my medical record. Just AI at the limits of its talent, I suppose...or I need to be more careful in what I say in the doctor's office!

I guess my worry about AI in dentistry is linked to how many old brewing root canal problems I'd been carrying around for years that Xrays and even recent CBCT scans have not warned my dentists about in time to prevent expensive reworking or tooth loss. I have had to argue my case repeatedly and keep looking for a more skilled dentist. In the right hands, it can be a good tool, I'm sure. I had two local endodontists recommend extraction of an old root canal crowned tooth before I located a better endodontist practice 100 miles away who was willing to tackle the repair. That practice uses a CBCTs, but so did the other two...