Illustrated by our shortlisted Large Langue Models
One sunny California day a student failed her Bio 101.
She grew sad and decided to drop out of college.
She knocked on the email door of the only professor with whom she'd once held an actual conversation.
Knock knock knock at sunny California college dot edu... I wish to say my last college goodbye...
Poor human professor, that goodbye so weighed him down.
What an amazing doctor this girl would be...
I will create an artificial professor, decided the human professor.
So that any student gets to actually know all of the biology they will ever want to know.
The artificial professor will read all of the biology literature out there...
It will read all the new publications too...
It will figure out what everything means... and why...
The artificial professor will explain it all, and in no time at all...
The artificial professor will find out exactly what you need to know before you can understand the new material... And it will explain that too...
The artificial professor will re-verbalize its explanations until you truly understand... It will listen to your recaps until you truly absorb it all...
But how to create such artificial professor?
The human professor set off to read all of the artificial intelligence literature that ever was written.
He read for one thousand and one California days. And then he read for one thousand and one New York nights. He never stopped to drink or eat or sleep, he only read and read and read and read...
Yet what he was looking for he could not find.
No book and no paper and not even one of the blogs knew how to create the artificial professor.
It turns out that artificial intelligence can only make copies; the original must already exist.
An advanced language model with transformers, for example, can take in loads of real biology text and push out fake biology text that appears as impossible to learn as the real thing.
How about an educational system that fails most of its students most of the time? Can a transformer language model replicate that too?
Oh no! So the challenge was not about how to just copy!
Finally, the human professor understood what the challenge really was.
The challenge was how to succeed! To succeed at imparting advanced knowledge!
An original original was needed. Which artificial intelligence would then copy.
So the human professor brewed himself a cup of green tea, sat in a peaceful mediation asana, and closed his eyes to see a new way of systematizing knowledge.
Breath in... Breath out... Focus...
Millions and millions and millions of books and papers appear before the human professor's mind's eye.
But as soon as they meet the mind's eye, all of their language disappears! The mind's eye only sees the actual knowledge. And the actual knowledge is made of tiny small kernels!
Breath in... Breath out... Focus... The tiny kernels of knowledge jump up, they jump down, they jump sideways...
The tiny kernels of knowledge jump from languageless book to languageless paper to other languageless book to other languageless paper... Then they jump again... and then they jump again...
Breath in... Breath out... Focus... Elementary Knowledge Units, Elementary Knowledge Units... The human professor is chanting. EKUs... EKUs...
Freed from the weight of the language, the dancing EKUs shine bright.
Eureka! EKU systematization is what conventional artificial intelligence needs to create artificial professors!
Once out of his trance, the human professor detoured towards the patent office.
$700,000 later he held in hand the beaming patent office EKU: "Be It Known That The Human Professor Invented The Artificial Professor".
Brightly lit by the beam of the patent EKU, the human inventor knocked on a digital door made of blue gold and yellow gold.
Knock knock knock at geniuses with 300+ years of AI programming experience dot Ukraine...
Please build the Artificial Professor of Bio 101!
