A quadratic map is a function of the form f (x) = ax2 + bx + c, or more generally,
.The only simpler polynomials are those of degree 1, omitting the terms aijxixj, and the constants f(x) = c. This delightful short note proves that quadratic maps, simple as they may seem, are hard to sample. That means there is no constant-depth circuit of size polynomial in n and complexity AC0 that outputs the probability distribution (x,f (x)), where x is uniformly distributed. This is in contrast to degree 1 maps, which can be sampled by an AC0 circuit by a theorem of Babai. A quadratic map that cannot be sampled is explicitly constructed, where x ranges over the field 0,1 with two elements. Namely, the counterexample is the composition of the inner product quadratic map (x1, … xn) ↦ x1x2 + x2x3 + … + xn - 1 xn with a random linear transformation. This is remarkable because the inner product quadratic map on its own can be sampled with an AC0 circuit. This example advances the understanding of which maps can be samples in AC0.