Develop AI systems to efficiently learn new skills and solve open-ended problems, rather than depend exclusively on AI systems trained with extensive datasets
Denoising, chaos to order, sharpening?
In this one, our reasoning points to the idea that we shouldn't really worry about trying to transform all the isolated particles (squares) into the expected form, but rather learn to see the expected form underneath the noise.
Filter out the noise and you're good to go, no transformations needed.
It seems through the 6 examples given in each input, we can reason about what the expected form is supposed to be. Maybe the mode value for each position?
Discover other posts like this one