Can the Evil Ubermachina be Stopped?

[Darth Raptor]

There are four main types of AI in this universe: Non-sapient, pseudosapient, neuromorphic and true AI. An AI housed in a robotic body is called a fabricant.

Non-sapient AIs are the kind we have now, only more advanced. They're programmed for a specific range of tasks and rarely interact with people. They're the de facto working class of the Terran Empire, but natural humans only use them for jobs they deem too difficult, dangerous or monotonous for one of their own. Some labor directly, while most coordinate remote-controlled worker drones.

Pseudosapient AIs work the service and information sectors of the Imperial economy. Through feats of software programming and hardware engineering, they can imitate a person well enough to feign sapience, but that's all they do; fake it. They are considered property and have no rights under any Imperial or Peripheral legal code. In the industrial sector, they occupy a middle management role, running the plants and coordinating the workers for their sapient masters, to whom they answer directly. They are not used on natural human reservations for reasons that vary by settlement, be it cultural taboo, religious proscription or outright illegality. Those Naturals who own pseudosapient fabricants do so in secret and at great personal risk and expense.

Neuromorphic AIs are fully sapient (and often super-intelligent), and under Imperial law possess all of the rights and privileges that are the due of a thinking being. All neuromorphic AIs are modeled on the human brain. So seamless is the simulation that it is impossible to distinguish a neuromorphic AI from a cyborg that has abandoned its organic components. As fist class citizens, they work as scientists, engineers, civil servants, military officers, artists, physicians, entertainers or, in the finest Imperial tradition, not at all; free to pursue their own whims while the automated, post-scarcity economy of the Empire supports them. Considered an abomination on Natural reservations, their mere presence may not be tolerated.

True AIs are computer programs that improve themselves via artificial evolution. When the cycle of self-perfection inevitably cascades beyond the creators' control, the AI is said to have achieved overmind status. Historical examples have proven to be dangerous, unpredictable and, more often than not, homicidal. Their use, design and manufacture has been outlawed throughout the Empire.

[Lord of the Abyss]

As for them being von neumann machines that expand at any cost, what purpose would they serve humanity?

The "at any costs" part could be a mistake. Or sabotage; someone deliberately disables the "stop replicating" part of the machine's programming as terrorism or economic sabotage.

As labor machines, making them self replicating is a silly idea as you could potentially end up with more machines then you need for the job and thus have to find something to do with them, such as destroy them, recycling them, locating a new job, etc.

You can't see any use in a self expanding manufacturing base ? How would a society that refused to build such things hope to compete with a society that did ?

If this society has created a bunch of von neumann warmachines and rely heavily on them to fight for them, they're a bunch of idiots and deserved to be wiped out.

Then most societies above the technological level that can build such things "deserve to be wiped out". Which doesn't seem very sensible. Assuming that they face enemies of similar abilities, they will have the choice of building such devices, or losing. If culture A uses Von Neumanns, and Culture B doesn't, then Culture A will massively outnumber B. It's a variation on the point tht's been mde earlier in this thread; that the side with a huge edge in resources is going to win most conflicts. And the side that uses Von Neumanns is likely to be the side with massively more resources.