3 ms·
It's hard to say. Good moats, at least mid-term, are rarely algorithms. It's usually data and infrastructure. However, for some applications I am interested in
by nextos 3y ago
It's hard to say. Good moats, at least mid-term, are rarely algorithms. It's usually data and infrastructure.
However, for some applications I am interested in, I think that robust representation learning solutions could give a significant edge.
But that is mostly an open problem in high-dimensional spaces.
- stuartaxelowen 3y agoAnd not just data, but hard to match data generators. I’m skeptical of the defensibility of any given licensed dataset.