klhugo

Hi there,

I had an interesting idea on how to perform useful compute with proof of work. We could wrap a problem using Homomorphic Encryption techniques. Each node gets a different problem to solve. If the solution matches a certain pattern, they can write the next block (similar to already existing methods).

The key difference is that the research institution interested in submitting a job would pay a fee, and that fee is split among all miners that worked in solving the puzzle - creating a second, STEADY source of revenue to GPU miners!

This would put a massive number of GPUs out there into service of science, while increasing the revenue of miners. Efficiency would improve overtime as FHE algorithms evolve. ASIC miners would be very welcome.

Please criticize.

Thought experiment.

Let's assume that a number of services only work well/are viable when the app collects a lot of information from its users.

What if we make all app's collected data available publicly in a database - with user identity anonymized.

This would make me a lot more comfortable to use the service. I would have less of that feeling "they are profiting over me". And it would incite competition since all companies would have access to the data as well. Something like open banking [1], but for user data.

The challenge is to ensure the data is published in a format where you can not trace back to the users, while still allowing ML apps to use it [2].

I'm curious to hear community thoughts on this.

[1] https://en.wikipedia.org/wiki/Open_banking [2] https://en.wikipedia.org/wiki/Differential_privacy

Crazy idea here :)

The better I get at writing clean code, the less comments I need to use. What if there was a "compiler mode" where comments were forbidden. Or maybe you are entitled to a limited number of comment words per file.

Could that spur programmers to write better code?