6
$\begingroup$

Motivated by Aaronson's call to find simple, verifiable proofs of quantumness, suppose we start off with a random polynomial-length circuit $U$ of, say, Hadamard+CCNOT (Toffoli) or CSWAP (Fredkin) gates, and attach $U^\dagger$ to it, can we then use a classical obfuscator to scramble $I=U^\dagger U$ beyond recognition, while still keeping the overall circuit size polynomial?

If we could do such scrambling easily enough, we can then provide a classical description of our obfuscated circuit for $I$, or alternatively of another random circuit $C$ from the same generating set of gates, to a quantum computer, who could determine which is which merely by executing the circuit on various basis states.

Indeed, we could insert a circuit such as $D=I+\epsilon$ that is exponentially close to the identity, and then run our circuit so obtained $U^\dagger D U$ through a Solovay-Kitaev compiler, to maybe add more spice to our circuit.

Are there enough known interesting relations to do such scrambling? After all, non-identity-checking is coQMA-complete.

$\endgroup$
2
  • 1
    $\begingroup$ One option, which is not quite your question, is blind quantum computation, which has the ability to build in "trap qubits". $\endgroup$
    – DaftWullie
    Commented Jul 28, 2023 at 6:42
  • $\begingroup$ Thanks! I don't know enough (or much at all) about blind quantum computing - I will study more! Perhaps it seems dual to the above test for quantumness, though? We aren't looking to bind a quantum algorithm to do our bidding and be oblivious about what that bidding is; rather, we wish for a quantum computer (and only a quantum computer) to recognize when it's been given the identity. $\endgroup$ Commented Jul 29, 2023 at 19:58

0