I got excited when I saw IBM’s announcement of their Q System One – a commercial product for scientific and commercial use. Pretty slick-looking package, big partner program, etc. Serious commitment from IBM, no doubt of that.
Closer examination showed we are still a long way from useful computation. Quantum computers today are physically much like the analog computers that preceded today’s stored program digital computers (so-called von Neumann machines). The basic element of a quantum computer is the qubit, or quantum bit. A qubit, like a classical bit, has two possible states that can be seen by an external observer or readout, 0 and 1. However, the internal state of the qubit can be a superposition of both these states, which can be visualized as a vector in a two-dimensional complex space (Bloch Sphere). Rather than a stored program, the algorithm is expressed as a set of circuits that interconnect qubits and logic gates. These gates can operate on the qubits’ internal state, allowing complex matrix calculations to be performed much more compactly and rapidly than a von Neumann machine. The problem is that the inputs and output of this computation must be classical bits, using only the basis states, 0 and 1, of the qubits. We humans can only perceive cats as alive or dead. So a 16-qubit machine can only, at best, give a 16-bit result. Sure, this result can be fed back as a starting value but that gives up the benefit of the quantum computation.
The bottom line is that today’s quantum computers are toys. There is no obvious way of getting around the I/O limitation, as the rules about that are a fundamental part of quantum physics as it is presently understood. Can the toys be scaled up? I hope so, but there is a long way to go.
Financial Reality 