Part 6. Takeaways

Today, FHE is an underutilized technology. Its implications are so paradigm-shifting that most people don't yet know how to think about it. People in positions to use FHE haven't heard of it. People who have heard of it still think it's too slow to do anything useful. And many people building FHE are thinking about it wrong: its real power isn't in letting a client outsource a single computation to a server — it's in enabling arbitrary multiparty computation with $O(1)$ communication.

It's tempting to compare FHE's performance to plaintext and conclude it's not viable. But that misses the point. FHE's viability isn't determined by raw speed on the hardest computations — it's determined by what becomes possible at all. We sent people to the moon on computers with kilobytes of memory. Imagine what's possible with the “primitive” computations FHE enables today.

FHE will be a revolutionary technology. The world just isn't ready for it. It's 0xPARC's mission to change that.