Reading What Machines "Think": A Challenge for Nanotechnology

Determining what machines "think" can be considered a well-set problem. Computational systems have recently become so complex to motivate a parallelism with the human brain. Such a parallelism may represent a test-bed for brain-imaging as brain interpretative models can be tested on a much simpler case. We performed a virtual observation of a computational machine: the machine activity has been observed using a software program that snapshots the machine memory. Images of the memory activation states have been produced with bit resolutions. Building on these results, we are interested in what can be physically reading the machines' "thoughts" and what can be its technological implications. This is a peculiar challenging task of nanotechnology, the elementary information unit 1/0 (the bit) nowadays corresponding on chip to a physical elementary unit of nanometric dimensions. To capture activation states in physical memories we need devices that do not interfere with the chip both from a mechanical and an electro-magnetic point of view and have imaging resolutions comparable with the minimum line separation typical of the modern processors. The present work explores a new scientific field that can foster advances in neurosciences and, secondarily, in computer diagnostics.