Measurement of entanglement remains an important problem for quantum information. We present the design and simulation of an experimental method for an entanglement indicator for a general multiqubit state. The system can be in a pure or a mixed state, and it need not be "close" to any particular state. The system contains information about its own entanglement; we use dynamic learning methods to map this information onto a single experimental measurement which is our entanglement indicator. Our method does not require prior state reconstruction or lengthy optimization. An entanglement witness emerges from the learning process, beginning with two-qubit systems, and extrapolating this to three, four, and five qubit systems where the entanglement is not well understood. Our independently learned measures for three-qubit systems compare favorably with known entanglement measures. As the size of the system grows the amount of additional training necessary diminishes, raising hopes for applicability to large computational systems.