But what exactly does it mean to “measure” a qubit? As described at the beginning, quantum mechanical systems are in all possible states until they are observed or measured. Until a measurement is performed, a quantum system is, so to speak, a “black box” (a closed system where the insides are unknown). Even if the state of the quantum system is not known, it can still be influenced from the outside without losing destroying the superposition. While there is no classical analogy for superposition, it is not difficult to find an example of a black box in that regard: Even if one does not know the contents of a cooking pot, one can still heat it on a stove or add spices to influence the contents. If one also has the information about the insides of the pot at the beginning, one can imagine (with sufficient prior culinary knowledge) what happened while cooking it. By looking into the pot and tasting its contents, certainty is achieved.
This brings us to one of the most important statements of this post: A computation on a quantum computer is nothing more than the controlled modification of probabilities. As a simplified picture, one can imagine that an operation on a quantum computer shifts the points on the Bloch spheres of several qubits, thus changing the probabilities of measuring 0 or 1 for each qubit. Like this, one generates different probabilities for different states of the register, namely, a probability distribution. A single measurement, whose technical implementation depends strongly on the chosen qubit architecture, then resembles a random experiment. Because of the importance of this term in connection with quantum computing, it shall be considered again in more detail by means of a small example:
When rolling two normal dice, a random experiment is performed. In this case, the most probable result is a 7, since most value combinations of the dice result in this value (1 and 6, 2 and 5, 3 and 4). Rolling a 2, on the other hand, is very unlikely, since both dice must show a 1. It can be calculated that the probability of rolling a 7 with two dice is about 17%, while the probability of rolling a 2 is about 3%. So when rolling two dice 100 times and writing down the results, one expects a 7 in about 17 cases, whereas a 2 should occur in about 3 rolls. Similar to a quantum computer, rolling the dice (measuring) once does not give any information about the probability distribution of the numbers on the dice (states). However, if one performs the same random experiment more often, the frequency of the results behaves according to the probability distribution.