Next Steps for Machine Learning

What next?

Well, naturally we are going to have to go a helluva lot further.

HEHEEH i speak authoritatively on subjects I know little about!

Non Euclidean Geometries.

Currently we are usign tensors which are euclidean by design. And yes! we get a lot of good results. BUT REMEMBER, euclid's first 4 postulates still hold in non euclidean geometries. So this is an avenue especially as we figure out how to actually model these in different ways...

Though, then again, i mean kinda fuck it tho, like our world was confirmed to be euclidean, or at least where we set ourselves to be, so... idek maybe the curve is just really hard to see... anyways, idt we will ever know. We must imagine Sisyphus as Happy

Minimizing Uncertainty over Loss

Fuck a loss function. There is no right or wrong, or correct or incorrect classification of data. We need to move clearly into the ability to draw easy decision boundaries between data. As we understand it at least. Nothing is the exact same, and nothing will have the same embedding, but hmm idk this one has a lot of other implications I need to consider.