Gradient News

YES

YES

YES

At least one of the six left Millennium problems is found with a counter-example

No AI will be powerful enough to directly (as a necessary condition) lead to a significant discovery in math.

At least one of the six left Millennium problems is undecidable

A unified framework for gravitation and quantum theory is proposed that is not string theory

Neural Networks are explainable and can be trained without gradient descent with better performance.

P vs. NP will be trivial (i.e., provable using simple methods taught in undergrad classes, e.g., with a counter-example)

NO

YES

NO

NO

YES

Other

MIP* = RE

The Complexity of Gradient Descent: CLS = PPAD ∩ PLS

Graph Isomorphism in Quasipolynomial Time

A Dichotomy Theorem for Nonuniform CSPs

Superpolynomial Lower Bounds Against Low-Depth Algebraic Circuits

An Average-Case Depth Hierarchy Theorem for Boolean Circuits

Learning Algorithms from Natural Proofs

Deciding Parity Games in Quasi-polynomial Time

A Proof of the CSP Dichotomy Conjecture

NO

NO