What technical projects do you think should exist in the world? What hard problems exist in your field? Where do you think the frontiers of science/technology lie?<p>I'm interested in learning about hard technical problems you think are worth solving.
Creating a "Wikipedia of Arguments", one can refer to in public discussions.<p>It turns out that this is a way harder problem than I thought. I'm in this space for some time now and am happy to talk about it.<p>Relevant links:
<a href="https://github.com/canonical-debate-lab/paper/blob/master/README.mediawiki" rel="nofollow">https://github.com/canonical-debate-lab/paper/blob/master/RE...</a><p><a href="https://www.societylibrary.org/" rel="nofollow">https://www.societylibrary.org/</a>
Way outside my field, but a platform for global climate management that links current actions to future outcomes and accountability to nations and companies would be very cool, assuming it is taken seriously by parties involved (which is an even harder challenge). This is technically hard, but harder still from a people perspective.
A Laser Mosquito Turret for homes, that uses a directional microphone to home in on the target.I don’t believe that it would be difficult to build the laser part, but the difficulty would be in making it safe for home use.
An alternative to the html model for building websites. The html model is antiquated for building web apps, and inconsistent across various browsers x platforms. The direction of the spec imo is becoming an overengineered mess that is increasingly controlled by the Google (see portals and web packages).<p>With the arrival of webgpu, I predict we'll begin to see end to end frameworks for building web apps that own the entire pipeline of application state management to drawing pixels on the screen.
I have two regarding embedded software stack,<p>1. Memory Safety
C-language literally dominates embedded software stack, while every on knows it is not memory safe and terrible memory bugs will be found but not yet discovered.<p>2. Compile time
The whole GNU toolchain take minutes if not hours to make, which makes CI/CD painfully long.
No one seems to trust vpns. I wonder if a VPN company almost needs to be law technology. All of our attempts at privacy is technically sound but lacks the backing of solid lawyers ready to hold it down. That’s got to be worth something.
Making large transformer based models like Jukebox [1] efficient enough so that they can be trained on a single 8x GPU machine to the same level of quality as the original model, in the same amount of time.<p>[1] <a href="https://openai.com/blog/jukebox/" rel="nofollow">https://openai.com/blog/jukebox/</a>