Large, complex neural networks are typically run on graphics processor units (GPU), special chips with powerful parallel processing capabilities. The many processors in a GPU each simulate thousands of neurons. GPUs have revolutionized the capabilities of neural networks, greatly expanding what can be done with machine learning. GPUs are rapidly becoming more powerful. However, the computer that simulates your brain clone may utilize even larger neural networks fabricated directly in silicon. Neuromorphic chips contain artificial neurons instead of processors. Human neurons can be as small as 4000 nanometers (nm) across. If we could create artitificial neurons that are about 40 nm across, roughly the size a 14 nm process transistor, then neuromorphic chips would start to approach the densities needed to rival the brain.