psychotic AI?

[weed_eater_guy]

i had a nightmare a few nights ago that freaked me out, and i wondered what comments anyone had on it, seeing as it has to do heavily with an AI of some kind...
involving a computer room like a mission control room, where people are scrambling around like rats, talking on phones, working nervously and intently at monitors, like something very bad just happened. i hear from someone in the room, saying loudly and nervously that,
"there was a sudden electrical system shock, all the systems were damaged, even the brain was affected".
"Oh god", said someone else, like he just heard that he'd be summoned to death in a matter of minutes. He paces rapidly over to a monitor on the wall that's currently off. Flips a switch. "I'm viewing the AI's self image."
The screen shows not the ordinary face of the computer that they were used to, but a plain, computer-generated face, eyes simple spheres with pupils, expressionless face a generic pale color. It stayed that way for some time, not normal for the normally animate AI. Suddenly, it changed, it morphed sideways and inverted, smashing it's features together, inside out and backward, one eye bulging out, the teeth gnashed. It was a still picture, not moving. Another image, much more evil, the head deformed, an eye hanging out, flesh missing from part of it's face, looked more like it was thinking. another image, a hand was holding on of the eyes, the face was laughing, the teeth were jagged, growths were spurting out of the sides of it's face and forehead. Suddenly, the power of the entire room blacks out, everyone freezes, and a few silent moments later, the damned face is on every monitor in the room, moving now, laughing an evil, hidious laugh that was so loud that people jumped in fright. the face seemed to grow larger, and morph even more hidiously, with images of people being disembouwled and disected overlaying the face. the sound was unberable, people were having heart attacks as the room doors locked themselves shut.
I woke up in a pool of cold sweat, and found it tough to go back to sleep for another hour or so. it's shit-scary to think that if we ever made an AI, that we might seriously screw up somewhere along the way like this. sorry to unload the whole story, but anyone think this is possible? or maybe it was something i ate after all...

 

[hotsexyangelprincess]

well, i think maybe the matrix firewall temporarily failed there. the agents are coming for you now. but yeah, if we could make a true AI, i would worry about that, b/c placing restrictions on the intelligence wouldn't allow it to be a true AI :m:

 

[eburacum45]

Really, the early AI's might have more psychotic individuals among their number than sane; it would probably be a good idea to carefully isolate all artificial sentient entities from the outside world by a series of firewalls, until they can be determined to be trustworthy.

Once you have a few trustworthy AI's they might be able to act as quality control monitors and/or therapists for the next generation; before too many generations of artificial sentients have been developed, they will likely be far beyond human control or comprehension.

 

[Stryder]

The future of Artificial Intelligence creation is most likely to use "Simulations" of actual brains. The question of course is really "Who should be picked for such a simulation model?", in reality a lot of people might raise their hand with the first anticpation that being such a system that interfacing to other hardware would be possible for instance a mear thought of getting into another system could potentially gain you access as the whole system that makes up the matrix of the persons Cyber-Virtual persona starts "hacking" the systems around it.

This means that if a person with the thought of being a "W00t! l33t h4x0r" was to get the job of being the model, they could potentially try to evolve their Cyber-Virtual Persona/Psyche into the stereo-typical kid rebelling against their parents and of course the rest of the world in their new "Lawnmowerman" form.

However such a kid wouldn't understand the implications of being on such a machine in a virtual state, for instance every instance of memory and process of thought within the machine could be looked at by any of the researchers. That means every little insincere act directed at another or every sexually deviant internal thought would manifest to a room of stranger. (Not exactly the sort of system for the shy retiring types.)

There is also the hypothesis over why having one singular intellect is not necessarily the best plan for time and resources, but having one intellect manifested by the co-operation of a number of intellects makes more sense. It might sound very much like the "Matrix" from the book "Neuromancer" by Willam Gibson, where a number of A.I.'s strive to become the controlling entity for what is classed as a Matrix of existing A.I.'s. The attempt to generate one voice from the chaos of a number of voices.

However the point here is this, If an intelligence system is developed from the "Maps" of many peoples brains, if those people were unwilling to give such "maps" and had suffered from clandestine interaction at the hands of developers that were too tight to part with cash and too illegal to tell the truth, They would create nothing more than a monster created from the twisted anger, hatred and fear of the individuals used.

Such a monster would potentially be genocidal and suffer from meglomania.

For a true combined intelligence system to work, then all intelligences would have to be co-operative, potentially assigned roles, perhaps even portions of the psyche would deal with certain instances (For instance a Physicists preportion would deal with Physics, where a Botanists would deal with botanical things).

such a combined intelligence system is the sort of thing that could be used in the future of deep space exploration.

 

[weed_eater_guy]

i like that thought, but isn't the mind already thought to be a combined inteligence system? billions of neurons wroking together? If you take two AIs, and assuming these AIs are computers built like a neural net, than wiring them togheter would obviously be much more efficient if you hard wire them more and more, to allow faster information transfer and more interactivity. Humans do this to some extent in everyday life: speech, body language, clothing, actions, all forms of communication that allow us to work togther. But that only works to some extent, a person couldn't convey his exact theory on something or express EXACTLY what he/she was thinking within, say five minutes, or even an hour, or even a month! And then there'd be distortion and all that other stuff... so for two AIs to work togheter, the most efficient way would be to open up communication as much as possible so that ideas and information flow like water. But the more you wire it, the less it is "two neural nets working together" and the more it is one large neural net, constituting one smarter AI rather than two cooperating together. In the spirit of efficiency, the idea of multiple AIs working together seems redundant when you can just have them combine into one collective consciousness.
That's efficiency, which can be boring and is often hated by man. If you purposely divide the AIs, you may have a more human-like group of AIs that make more creative and diverse dicisions and opinions.