True, but significantly faster global optimization of multimode functions by any means turns out to be pretty useful for exactly the types of ML problems that a

True, but significantly faster global optimization of multimode functions by any means turns out to be pretty useful for exactly the types of ML problems that are stretching the limits of current hardware. This seems like a promising first step towards technology that could be really useful if it comes anywhere close to doing what it claims.