Microsoft's Purposeful Marketing Deception About AI's Ability To Beat Physicians In Diagnosis
Looking At The Reality Of MAI-DxO, What Was "Actually" Tested, And The Junk Marketing Research Design In
Hey Seekers!
Welcome to Today’s Edition of the Seeking Sageship Newsletter!
Your Daily Guidance to Go Beyond Leadership!
All Subscribers have Access to the Free Section…
1%+ Compounded Improvement in 1% of Your Time Maximum!
Plus, a Preview of the “Psychophysiological Freedom” Paid Section Is Available!
Let’s Dive In…
Microsoft’s Purposeful Marketing Deception About AI’s Ability To Beat Physicians In Diagnosis
Looking At The Reality Of MAI-DxO, What Was “Actually” Tested, And The Junk Marketing Research Design Involved
When you look at the world of scientific research...
If you are not careful...
You will be fooled into believing things that are not actually supported by the data.
Very often, statistics and study designs are manipulated in ways...
To make you believe something...
That is clearly not true and does not represent reality in any way.
This is often referred to as “Marketing Research” or “Junk Research”...
And it, sadly, is overwhelming in our world today.
Let me give you an example...
Coming from the mixed world of Microsoft and AI.
Microsoft created an AI...
Called “Microsoft AI Diagnostic Orchestrator” or MAI-DxO...
That they pitted against real doctors to see who could do better...
And the results “sound” astounding.
In the study...
The MAI-DxO outperformed human doctors at a high rate.
The doctors scored an average of 20% diagnostic accuracy.
But the MAI-DxO?
Had an 85.5% diagnostic accuracy.
That makes AI sound vastly more capable than human doctors!
So, should we just hand over our medical diagnosis processes to AI?
My answer...
Is a STRONG NO.
Which is not a joke...
You should understand that I am very wary about placing my trust in today’s AI for any medical purposes...
And this “Study” only adds to my suspicions.
Why?
Well, let’s start with the design.
Who was this MAI-DxO placed against?
21 “experienced” physicians...
But if you look closely...
This is also just clever marketing.
Some of the physicians only had 5 years of experience...
And “none” of them had over 20 years of experience.
If you actually wanted to test “experienced physicians”...
You would want to look at physicians who had 40 or more years of experience.
That is “truly” experienced.
On top of this...
21 physicians is “not” a strong data set.
If you wanted to test a strong population of physicians...
You would want at “least” 100...
And their average should be “well” above 20 years to be considered “experienced”.
So, we can already see that there is purposeful manipulation of study design for the goal of getting a specific result.
But further than this...
These physicians were “all” general practitioners.
Now, if you do not know the medical world, there are different types of physicians that exist.
The absolute least specialized, often requiring the least amount of technical knowledge, are your general practitioners.
Now, do not get me wrong...
These guys are crucial in the healthcare system.
They are the first person you are going to see the vast majority of the time (outside of major emergencies)...
And they handle the vast majority of common cases...
Which allow other doctors to be extremely specialized and only handle very specific cases that fall into their very specific perfews.
To give you an understanding...
There are several “hundred” forms of specialty and subspecialty doctors that exist.
No single person can claim to have all of the knowledge that exists in the field of medicine...
So physicians specialize in “literal” hundreds of categories...
And multiple levels of deeper and deeper specializations.
Your general practitioners sit at the top, usually never getting into deep specializations...
And often too busy handling the majority of patients to ever “learn” all of what is being discovered at the deepest levels of these specialties...
And if there are any cases that exist outside of what they know, it is their responsibility to “refer out” to the right specialist...
Not try to figure out the problem themselves.
So, not only are we not looking at truly experienced physicians...
But we are also looking at those least likely to be able to diagnose tricky cases.
Which brings us to the next point...
What did the study test?
300 case studies, aka “published” material (which will become important later)...
That comes from the New England Journal of Medicine’s weekly series of case studies...
That are, by design, meant to showcase the absolute most difficult case studies to stump physicians.
How?
Most physicians are either too generalized...
Or are not specialized in the specific areas for these different studies...
Or the case studies are based on research that is so new that the doctors involved in those case studies are essentially “discovering” new things in medicine.
So...
The truth is that the vast majority of physicians “would” fail to diagnose these specific cases...
That is the whole design of having these case studies in this journal.
So, it really is not surprising that the “general physicians” scored 20% on diagnosing correctly...
Because most of what they would be looking at is vastly outside of their wheelhouse...
And in many ways, it is extremely impressive that they got 20%...
Especially when you consider another aspect of the study design...
Which does not represent what happens in reality in the slightest.
The general physicians who were being tested...
On some of the most difficult case studies that exist in medicine today...
Were restricted in the following ways to make their diagnosis...
No internet access...
No collaboration with any other physicians...
And of course...
No use of AI tools.
But on the other hand...
MAI-DxO has...
Access to “vast” amounts of research...
Plus...
Multiple types of AI that are designed to act as “collaborative physicians” to get to better answers...
Such as...
“Dr. Hypothesis”...
“Dr. Test-Chooser”...
“Dr. Challenger”...
Which works as a team of specialized “AI Physicians” to try to come up with the best possible solutions.
So, really, what we are seeing is a team of generalized physicians...
Who are not truly deeply experienced...
Who are blindfolded with both hands behind their back...
Trying to solve some of the most difficult cases requiring teams of specialists to solve normally...
Against a team of specialized AI with incredible amounts of information at their disposal.
This is clearly not a fair test or demonstration of what happens in reality.
But here is what I find perhaps most interesting in all of this...
There is no record of what MAI-DxO was trained on.
So, it is “absolutely” possible that the AI System “had” already been trained on the case studies that were utilized...
Meaning it “should” have been able to diagnose 100% of the research, but still “failed” nearly 15% of the time.
But...
If we “assume” that this was not done in good faith...
Which we already know was clearly not the intention with the design of the research...
The AI system still had vastly more resources at its disposal than the physicians it was pitted against...
Showing that this “clearly” does “not” reflect what would happen in reality.
Now, should AI be able to look at vast amounts of information and find patterns that may not be immediately visible to humans?
Yes.
That is a large part of their design.
Could AI become a great tool as a “first step” in helping physicians diagnose?
Yes.
It likely will as systems are improved...
Though there is a significant debate on whether LLMs are actually capable of any more improvement - It may turn out that the AI used in the future will “not” be LLMs.
But “this” specific research does not truly represent the truth of what would happen in the real world.
It is Junk Marketing Research that is purposefully and deceptively designed to make AI “sound” vastly more impressive than what the study actually reveals.
If you gave human physicians the same types of resources as the AI system...
There is a high likelihood that even inexperienced generalized physicians could outperform the AI system.
In fact...
When we look at most technological advances...
We see that it is “always” humans “using” technology that beats technology alone.
This is likely what we will see happen in the future...
Not because of AI’s capabilities...
But because of what humans are capable of with the right resources at their disposal.
What becomes incredibly important when we look at this specific research...
Is the importance of understanding how to read research.
When you hear that AI was 85.5% effective at diagnosis, while “experienced physicians” were only 20% effective...
It paints a very specific picture in our minds...
One that is, as we have seen, a clear deception of what the research shows.
If we do not understand what the research actually shows...
It becomes easy to manipulate us.
If “I” wanted to manipulate you even further with this research...
I could have also used the exact same statistics to say that the AI outperformed humans in diagnosing complex cases by 427.5%.
That is using the “exact” same information...
Because 85.5% is 427.5% of of 20%.
But 427.5% is “sounds” vastly more impressive...
And that is likely why the creators of the research did not use this number...
It sounds “too good”...
Because, as we have seen, it “is” too good.
See...
This is another tactic used in Junk Marketing Research...
You want things to “sound” amazing...
But if it begins to sound “too” amazing...
People begin to look more closely at the data to see how it is purposefully manipulated.
Those who create Junk Marketing Research have to find the “goldilock statistics”...
They sound incredible “enough” to get published and make a wave...
But not “too” incredible as to get flagged as purposeful manipulation.
If you do not want to become manipulated...
You need to begin to look at what is actually being tested...
How it is being tested...
Whether the tests are fair...
And whether there is information that is being left out or minimized.
These types of tactics are “everywhere” in scientific research today...
And it is a large part of the reason why I personally mistrust the medical establishment so much...
As well as the supplement and health-tech industries.
All of these industries use a significant amount of Junk Marketing Research to make things appear significantly better than they are...
While also largely hiding the problems caused and long-term results of different “solutions” in today’s world.
If you do not know how to read the research...
You will be manipulated into believing things that are not true...
While also potentially endangering yourself to what is purposefully hidden from you.
Do not allow yourself to become victimized.
Those Who Seek Sageship Also Cultivate Others!
Share this Article Today With Someone Who Needs It and Earn Exclusive Rewards!
Are You Ready To Become A Sage?
Paid Subscribers will Gain ‘Exclusive’ Access to…
Read the Entire “Psychophysiology Freedom” Deep Dive Section for Exponential Results
Access to the “Full” Paid Archive (Check Out the Cultivation Center for More)
Connect Further in the “Subscriber Only” Chat & Comments Section
“Directly” Impact Regenerative Projects around the Globe
With “More” Coming In The Future…
It’s Time to Change The World!
Psychophysiological Freedom
For Paid Subscribers
So if you do not want to be misled by research…
There are a few things to pay attention to…
Which will help you to understand when research is really valid…
And when you are looking at Junk Marketing Research that is deceptively designed to manipulate you into believing something that is not really true.
I have 7 things to pay attention to so that you are not deceived.
Let’s Dive In…
