Deciding if an NP-complete problem can be solved in polynomial
time is the most fundamental unsolved problem in theoretical computer
science. We propose a new algebraic model of this problem, replacing the dense
input by a new algebraic input. This makes it possible to make significant
advances. In this expository lecture, we shall explain this new model and present
the relevant known results as well as conjectures.