- Guscon
- |
- Exalted Mythic Member
Posted by: AdjacentGibbon7
I have read quite a few articles on ai and rampancy and in this topic I will address rampancy and metastability.It's perhaps a good idea to specify whether you're talking about Rampancy in Halo or on Marathon. The two are somewhat different.Rampancy can be caused by many things overactive thinking, overheating the power core, and to much or to little outside influence.Assuming a Halo basis, time is the main cause of rampancy, although it does seem that 'excessive' thinking/data processing can induce Rampancy to occur faster. I'm not sure where 'overheating' came from, but as long as you're just using it as a synonym for 'working too hard' it can probably slide.
It's perhaps worth noting that the isolation cause of Rampancy is something more found in Marathon than Halo.Though rampancy and insanity share common traits such as no fear of consequence, believing oneself to be a religious entity among many other symptoms.I'm not sure you've finished your point here.If an ai were left alone with lets say the entire history of a planet and that one ai alone had access to said information three things could happen one the ai would only analyze the information and nothing would happen two the amount of information would be to much and the ai would quite literally think itself to death and three knowing that that ai alone could delete or store that much information would cause delusions of god like power.Again, important to specify which universe we're dealing with. In Marathon the AI are shown to be capable of spreading themselves through a networked system, and using this system to increase their processing power (I'm thinking here of Leela). The AI in Halo, however (the human ones at least) are somewhat confined to their own crystal matrix. A larger network is more for them to 'explore', rather than for them to utilize. This, however, is a half remembered idea, and I could be wrong.My prime example of rampancy is 343 guilty spark who has broken asmiths three laws of robotics disobeying orders from a human, caused intentional harm to a human, and finally given conflicting laws with his prime directive where the second directive would override his first objective spark stayed with the primary directive.As has been suggested, Asimov's laws are not something that any non-human synthetic intelligence would be expected to follow. Indeed, there's no reason (other than at the whim of the creator) why any human synthetic intelligence should follow such laws. It's also arguable that any truly synthetic intelligence would be able to ignore such laws, as they wished.
I'm also a little confused as to your point with the primary/secondary directives. One of the key points of Asimov's laws is their hierarchy: Tertiary directive is to be applied at all times, except when it conflicts with either the primary or the secondary, etc.Early on it was obvious to say the least spark was quirky what with his humming and other personality traits.Having personality traits does not mark out an AI as 'rampant'. That said, it certainly marks them out as being 'quirky'.When spark is "absorbing" information from the pillar of autumn he did not think said information would be in any other location such as oh say a library on earth or in cortana's data banks which if halo legends is right she had learned all known human history within an hour, however he is in a way tied to installation 04.I'm not sure of your point here. Information in a library on Earth, or indeed present in Cortana's head would not necessarily have concerned him, as he had all the data he needed right there.But to go into an ai's logic center one would see a vast nonlinear process of thinking dedicated to balancing out consequence, gain,preprogrammed moral values and much more.If human (smart) AI are created by mapping existing neural patterns from human brains, their thought process might not be as logical as one might think, simply much faster and smarter.Now onto metastablity my only examples of this happening are the story of Pinocchio becoming a real boy and if you hurt spark in halo 3 with the iwhbyd skull on he might say "ow that hurt my feelings wait I have feelings I am a real boy" though non canon ha spark been like this things may have gone a little differently for him. In my opinion for an ai to achieve metastability rampancy must last for quite some time and most logic and data must be deleted to where the ai has the intelligence and innocence of a child. If any one has any other theories I would be glad to hear them. If Spark's had anything, it's been time. His isolation and long wardenship of Installation 04 did perhaps set the stage for rampancy (and it's debatable whether Spark was actually Rampant), but it did not present itself until Spark's world was opened up, through the destruction of his Installation. This mirrors the rampancy of Durandal, though his 'ascent' occurred much more rapidly.
'Rampancy' is a tricky term. In Halo it's been used to describe the state of 'mental-lockup' a smart AI enters after a 7 year lifespan.
The 'Rampancy' of Spark is more akin to its Marathon definition, which is the one you've assumed for much of your post. It's a state of growth and mental expansion, rather than lockup. The danger comes from a potential lack of expansion space once the process has begun. If it cannot be sustained, the 'mind' of the AI seemingly collapses, or is at least damaged.