Hacker News new | ask | show | jobs
by JulianWasTaken 2298 days ago
I am way out of my depth so maybe this is pure nonsense, but presumably the goals aren't so much to evolve better performing models directly for some dataset specifically, but to see what kinds of model families it evolves and whether we've thought of all of them?

Once we've got one we can then presumably train specific models in a more targeted way.