You can't go around without knowing what IA transportability is.
Shakespeare invented words like lonely, hurry, generous, undress, critical or majestic, among others.
Sometimes they have ideas in their heads that the body of knowledge of their time does not allow them to express and they have to invent new ways to get them out of their heads.
As when Newton, being a minor, invented integral and differential calculus to express ideas that he observed in Nature and that mathematics was not ready to conceptualize.
Ramón y Cajal invented new histological techniques to be able to express what he was seeing under the microscope.
Those of us in AI are not geniuses, but we also need new concepts to abstract what is happening.
Concepts such as:
– Transportability or transfer learning: the idea that an algorithm you have trained with data from one place you can apply it in others.
– Multimodality: using multiple layers of data to generate a model (image, genomics, text…).
Words are just that. Words. Inventions.
But they translate that we are passing through new places.
Because I’ll tell you one thing.
You’re going to need to answer how you’re going to sub train your algorithm in a different geography.
Or how you’re going to turn something unintelligible from the machine learning world, like an “entropy”, into something that looks like logistic regression (even though it really isn’t).
These are the things you want anyone who gets into the journey of creating clinical AI with you to know.
And these are the things at which Savana is the best team in, at least, Europe.
(And in Natural Language Processing for research, the best in the world).
One of the best?
No, the best one.
P.S. If you want to do an AI project that includes in its design and execution those things that most of them haven’t even gone through yet, like real implementation or external validation, we can talk about it here.