Computation through photons, rather than electrons, is not new as an idea but in the past 30 years silicon has proved so effective from a manufacturing point of view and photonics applied to computation so elusive that in practice we only have few photonic chips for very specific applications and not a generalized optical computation chip comparing to silicon ones.
Technology, however, has progressed and photonics is the way to go in transmission (optical fibers) and some sort of optical computation in terms of optical flow management (like in optical add drop multiplexers) is widespread.
Last year the application of photonics for neuromorphic chips (particularly suited to image recognition) made a significant step forward.
A quite different approach has been described in an article on Nature Photonics where photons result from the emission of electrons moving from a higher to a lower energy state. To push an electron to a higher energy state the researchers at the University of Regensburg used extremely short electromagnetic pulses (in the order of THz, each pulse has a duration of 100 femtoseconds, one quadrillionth of a second). Being so short the pulse is sufficient to move the electron to a higher energy level but does not displace it, hence the electron will not bump into another electron (which is the case when electrons move around in a computer chip leading to energy dissipation in form of heat).
With this approach, called by the researchers "lightwave electronics", it is possibile to generate light pulses of the order of a few femtoseconds (hundred times shorter than the pulses used to hit the electron). These latter can be used as signals to perform computations. The result? The possibility (theoretical!) to create computers that are up to 100,000 faster than today's computer (that use signals having a duration in the order of nanoseconds).
Notice that we are in the realm of pure research, and it is quite different to be able to control one single femtosecond "signal" from controlling billions of them, as it is happening in a computer chip.
Nevertheless, the researchers managed to demonstrate that it is feasible to generate and control signals with such incredibly short duration. As Richard Feynman noticed: "there's plenty of room at the bottom"!