The main goal of algebraic complexity theory is to show explicit polynomials whose computation requires algebraic circuits of superpolynomial size. It was a longstanding open problem to prove superpolynomial lower bounds against algebraic circuits of even depth 3, whereas constant-depth boolean circuit lower bounds were long known. In a recent breakthrough result of Limaye, Srinivasan, Tavenas, the first-ever superpolynomial lower bounds were shown against constant-depth algebraic circuits. We improve this lower bound by using a refinement of their techniques. We also show that our lower bound cannot be improved significantly using these same techniques.
This is joint work with C.S. Bhargav and Nitin Saxena.