Skip to content

Comment on Ask HN: What is a 2024-2030 moat for AI

Comments

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.

And not just data, but hard to match data generators. I’m skeptical of the defensibility of any given licensed dataset.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.