Intel is adding new x86 instructions ideal for machine-learning applications to its upcoming Intel Xeon and Xeon Phi processors.
The instructions are part of the AVX-512 family: they are described by Chipzilla as "vector instructions for deep learning" with enhanced word-variable precision and single precision floating-point operations.
The actual instructions aren't yet documented – that's to come later, apparently. Right now, code has been submitted by Intel staffers to the Linux kernel development process that detects CPUs that support the new instructions. The changes are also mentioned in this Intel change log PDF.
Unless you're into assembly-level programming, you shouldn't really have to worry too much about this: your compilers and toolchains will eventually be updated to produce the new instructions from higher-level languages. If you like to dig deep to optimize your AI software, well, keep your eyes peeled for an update to Chipzilla's reference manuals. ®