roake

The mind grows numb...

As a physician, working at a hospital system that utilizes the “Cadillac” of the electronic medical records (EMR) software, I can tell you that the machine learning and “automation” has a long, long way to go.

The “automation” part of the machine learning basically looks for key words in my documentation as well as looking at lab results and diagnoses that have been entered by humans to try to match patterns and suggest additional medical issues.  The suggestions are wrong probably 70 percent of the time, and have been already documented 25 percent of the time (but the software doesn’t recognize this).

The few times that the suggestions are “relevant,” I’m getting asked to clarify something like which bacteria is responsible for causing a pneumonia.  How the f*%k should I know?  Most of the time respiratory cultures don’t grow the culprit bacteria; if it DOES grow, it usually takes a few days before there is enough to identify, and we document that anyway.

I’ve found that the AI portion of this 100’s of millions of dollars EMR is more of a distraction than anything else.  I don’t know any physician who doesn’t simply ignore it.

On a related topic, I would point out that the documentation that a physician dictates must meet an EXTREMELY complex set of criteria, covering tons of nonsense bullet points that change based on the patient and various diagnoses.  Dictating an entire note as a single recording into a watch would bypass all the advantages that the crazy expensive EMR software grants you, such as automating certain parts of the note (insertion of lab values, test results, etc.).  The article also doesn’t say whether it’s (A) SIRI transcribing (which works extremely poorly when medical terminology is involved), (B) Custom dictation software such as Dragon Medical is involved, or (C) A human transcription service types this up.

Lastly, to those of you that think computers are “good” at reading studies such as XRays, EKG’s, etc, I can assure that while this may be true in science fiction movies, these things suck in real life.  The automated EKG interpretation algorithms have been progressing for decades, but still suck.  They get some things right, but get just as much wrong.  We aren’t going to see these things nearly as good as humans anytime in the foreseeable future.

gatorguy said:

AppleExposed said:

BxBorn said:

AppleExposed said:

Surprised!!

Apple please hurry up and take Home seriously! I'm at the point of thinking Apple needs to acquire some home security companies and make these spyware devices irrelevant. Like Siri, iPhone and podcasts, this whole industry was Apple's idea and they're laying back allowing everyone in.

And cry me a river about the Siri thing. All information was anonymous and untraceable. Apple could release these recordings to the public and it would do diddly-squat!
Unlike spyware companies like Facebook and Google who know who your children are, where you live and what you purchase.

Why do you think Apple will be any different outside of the fact that in light of this Apple may more explicitly tell you that humans will periodically review footage to ensure the AI engine is correctly identifying what it it was programmed to identify? The human involvement won't change.

Unlike Facebook/Google/Amazon, Apple doesn't care what you do they only care about improving their services.

Privacy is a big deal to Apple.

gatorguy said:

AppleExposed said:

Surprised!!

Apple please hurry up and take Home seriously! I'm at the point of thinking Apple needs to acquire some home security companies and make these spyware devices irrelevant. Like Siri, iPhone and podcasts, this whole industry was Apple's idea and they're laying back allowing everyone in.

And cry me a river about the Siri thing. All information was anonymous and untraceable. Apple could release these recordings to the public and it would do diddly-squat!
Unlike spyware companies like Facebook and Google who know who your children are, where you live and what you purchase.

Anonymized and untraceable are not one and the same.

Google anonymized their voice recordings too before sending them out for 3rd party transcription. It still did not prevent a dishonest contracted person from taking a few containing personal information such as an address or person's name to a newspaper reporter who dedicated themselves to identifying the person behind the voice. successfully I might add allowing them to leap to the inference Google must be sharing personal information.

But Apple and Google were each doing the same thing, relegating a user to a number instead and otherwise attempting to separate the recordings from an identifiable person.  So yeah it's not all "untraceable".

Side note: The same source claimed Apple also assigned certain demographics to the supplied voice recordings such as general location, male or female, what app the recording came from, that type of thing. So Apple does apparently keep track of a little more than you understand them to.

The source claimed the recording were completely anonymous and untraceable.

Go back and read the original AI article again, as well as the original news report. I think you either never have or now choosing to just make something up. 
"The whistleblower said: “There have been countless instances of recordings featuring private discussions between doctors and patients, business deals, seemingly criminal dealings, sexual encounters and so on. These recordings are accompanied by user data showing location, contact details, and app data.”

Any recording that includes an address or name, and especially if combined with approximate location, is traceable. You knew that. 
Apple's voice recordings were no less (or more) traceable than Google's. Gosh is that me throwing Google under the bus too? 

No.  That’s you as a Google apologist trying to say Google is just as dedicated as Apple about privacy.  You are, as usual, wrong.

Don't think I want to wear the Unpatriotic Apple Watch.

These are some of the finest first-world problems that I have ever encountered.