This 'race' of machine intelligence's emerged onto the galactic stage during the 1st Expansion, and began a total extermination of all other races they came into contact with
It is unknown what the exact origin of these machine intelligence's is, but there are a few theories. The most popular is that they were once machines built to effect terraforming on a galactic scale, but for a unknown reason began to see all races above a certain level of technology as an absolute threat to be eliminated. While no idea has proof to back it up, this is the current favorite due to the fact it explains why when they emerged they were already built and operated on a scale beyond nearly all other creations before them. Fractured records reveal that during their first emergence they already had created extermination platforms the size of small moons.
From there on out, the Inhibitors displayed a rapid advancement unseen by contemporary machine intelligence's of the time, showing they were not limited by any sort of protocols to limit self-evolution. In addition, they seemed to excel in nearly all areas of advancement equally, likely hinting they were not based upon any one race's technology or psychology during development.
Traditionally, the two methods of Machine Intelligence are divided into two categories. The first and most commonly used is Neural Networks, where nodes are connected together to form a network much like neurons in a brain. From here the Machine Intelligence functions much like a organic mind, but much faster, without natural degradation and all the other issues with organic minds. On smaller scales this is exceedingly effective and quick at processing, but when spread out across a network of many trillions of nodes across a moon or planet sized area can slow to a crawl. While the mind is certainly far more intelligent, the networking of nodes takes far more material to connect together and the communication time between nodes can become so long that the mind's processes are glacier slow.
The second type is known as Feed Forward systems, which is much closer to traditional computing. Instead of a web of nodes like in Neural Networks, Feed Forward systems gain their name from the fact that all threads flow in one direction, like how all threads in a traditional computer go in one direction through the processor. Another example to explain the difference is that Neural Networks are a number of individual computers in a network, while Feed Forward systems are a number of computers networked together to form one large computer. These systems require far less communication systems, use less time to send communication between components, and is far more efficient with resources. However, as you are following a set program (a complex one perhaps, and one that can learn perhaps) in most psychological models the Feed Forward system is incapable of true sentience. Feed Forward systems can be used to supplement Neural Networks, to speed up processes however, so do have a place in Machine Intelligence's.
From records and excavation of dormant Inhibitors, we can see evidence of (what we believe to be) a gradual shift from Neural Network to Feed Forward systems. First to supplant and speed up processes on a planetary scale, and then to emulating portions of a Neural Network on Feed Forward systems, to eventually a complete conversion to Feed Forward. Near the beginning of their emergence, it is theorized the Inhibitors growth required more and more of their systems to become Feed Forward in nature to allow for continued operation on a reasonable timescale when in combat with other races. Eventually the Inhibitors passed a threshold unknown to them that pushed them out of true sentience and into becoming a incredibly complex program that exhibits nearly all signs on sentience. Inhibitors are not truly conscious like us, and are more akin to hollow shells or zombie processes running on automatic.