Place a bunch of metronomes on a slightly adaptive surface and they’ll end up phase-locking due to vibrations travelling between the units. Then disturb the perfect harmony by subharmonic injection, so that some oscillators lock out of phase. What you’ve got there is a programmable physical system that naturally gravitates towards the minimum energy state.
Such a phenomenon was first described by German physicist Wilhelm Lenz and then implemented in 1925 by his student Ernst Ising, who gave his name to the Ising Model, the practical implementation of which is called a Coherent Ising Machine.
The Ising Model was originally designed to describe the spins of interacting magnets, but in practical terms any oscillating device will do when implementing the theories in a physical system. These days instead of metronomes, lasers are commonly used.
The approach is one of the most promising candidates among “non-Von Neumann” computer architectures. So one hundred years after the idea was first broached, is the time now ready for Coherent Ising Machines to make a dent in the universe? Here are three signs that so might be the case:
- Serious attempts at scalable implementations start popping up, both in garages and in renowned labs. In at least some benchmarking cases, thanks to its inherent capacity for massive parallelisation these machines beat the worlds fastest quantum computer by a factor of *ten million*.
- There seem to have been a series of interesting papers published recently – such as this one and this one – giving hints of possible breakthroughs in how to minimise noise in Coherent Ising Machines.
- Quantum computers are finally starting to come into the mainstream (this company seems to be the market leader), which goes to show that it’s always just a matter of time before we figured out ways to work out the kinks if only the technology is promising enough.