if you peek outside of math/physics it's pretty much landscape of relative truths; imho ideal thruth/fake news detecting machine would simply require axiomatic/weighted input, ie. "I trust MIT with public key 0x..., youtube jesus from la with public key 0x... and my childhood mate 0x..." – based on that is "X true, false or undefined"? (weighted output). Because I trust MIT with weight ie. 500% and MIT trusts Caltech, my trust graph will favour Caltech view of the world. Yes you can throw blockchain and AI into it and it actually makes sense.