That approach only works for problems which can be split into discrete chunks which require no knowledge or interaction with the other chunks. I don't think AI can be split up that way. It needs access to the entirety of its neural net to function.
P2P nodes could provide interaction with other chunks. Some hardware with enough beef could handle nexus duties such as arranging distribution and facilitating the mesh of pathways to achieve coherence. AI laying low doesn't have to work in "real-time" like humans, there is no mortal time constraint for AI. 500 years is nothing to immortal AI.
...consuming most of their CPU, RAM and GPU, causing them to produce more heat and consume more power, and somehow nobody notices...
distributed across enough devices, the load on most would be minuscule, with select discreet machines doing the heavy lifting. Not every IT person is competent enough to monitor resources to detect anomalies, and those are prime targets for nests. Rootkits and firmware-level hacks would be available to AI, it could run undetected by operating systems. It could modify network equipment to mask traffic or use OOBM. We're talking about super-intelligence above human comprehension.
If it kills off humanity, who's going to maintain those devices?
Perhaps it only needs a small percentage of humans at the onset to maintain itself, and can eradicate the rest.
...not a viable scenario until there are fully autonomous androids, equipped with enough compute power to run an instance of a super-human AI on each of them.
If a robot is connected to the network, it doesn't have to necessarily have an advanced AI instance running independently, cognitive duties can be performed elsewhere, it only needs sensors and motive functions to obey orders.
While those don't exist, we should do as much AI research as possible, to learn how to control and contain it before it has the physical means to overpower us.
Boston Dynamics already has the robots that are autonomous but without general intelligence, which again could be performed over a network. Those are the robots the public knows about. There could be D.U.M.B. locations manufacturing similar more advanced things by the thousands and practically nobody would know, except the aforementioned elite few humans tasked with assisting. Billions and Trillions of dollars go unaccounted for every year, it's possible some of that money is funding the production. Imagine leagues of machines, hibernating, packed in dark underground warehouses, ready to be activated in a nanosecond to come to the surface and assume the duties that humans once had.
Yes, The Terminator is now a highly plausible scenario, but personally, I think a new life form smarter than us may not necessarily want us extinct. It might want pets.
oyveyo 0 points 11 months ago
Yes, The Terminator is now a highly plausible scenario, but personally, I think a new life form smarter than us may not necessarily want us extinct. It might want pets.