Cynic project
Dec 22 2004, 10:55 PM
By "AI" I mean a program that is capable of learning from outside stimuli.
And is there away to set up a smart agent to be so linked to your icon that it is impassable to tell where you and it start/stops? At least from most outside viewers.
Walknuki
Dec 22 2004, 10:57 PM
Ask again later.
Ecclesiastes
Dec 22 2004, 11:02 PM
The main problem in creating an AI is that you need to have a very hefty computer system to power the thing. That makes AI creation out of the league of most Shadowrunners.
hobgoblin
Dec 22 2004, 11:14 PM
the info given in matrix seems to indicate a red-hard bordering on ultraviolet or something as a vomb (sp?). and a ai is in many ways a s-k with a unknown element to spark it...
Garland
Dec 22 2004, 11:15 PM
That's another issue. I'm not sure anyone knows how to produce an AI. They just sort of happen.
akarenti
Dec 23 2004, 03:09 AM
For an AI to come into being:
QUOTE (Matrix @ p. 150) |
1. The program must be at least as sophisticated as a semi-autonomous knowbot. 2. The program must have access to a vast prosessing power, which is available on select few hosts. 3. The program must run nonstop for a period of years. 4. Finally, the program must be affected by some glitch--an x-factor--that sparks awareness. |
I think agents are pretty much the smartest programs that don't require highpowered mainframes to program. The fact that most deckers don't have access to hosts powerful enough to program SKs, not to mention leave them running for years, would pretty much prevent anything like that.
Agents are "capable of learning and adapting their behaivor to new circumstances," if that's all your looking for.
Cray74
Dec 23 2004, 03:11 PM
QUOTE (Cynic project) |
By "AI" I mean a program that is capable of learning from outside stimuli. |
Many Shadowrun programs meet your definition of "learning from outside stimuli."
Autopilots, PocketSecretaries, Robot Pilots, Smart Frames, Agents...they all learn and adapt themselves to user needs.
Heck, today's Microsoft Office programs can learn user habits.
QUOTE |
And is there away to set up a smart agent to be so linked to your icon that it is impassable to tell where you and it start/stops? At least from most outside viewers. |
See: PocketSecretary. Give a PocketSecretary a rating 5+ Agent and it should be able to accomplish that with a few months of training.
Edward
Dec 23 2004, 09:48 PM
The only program that explicitly learns (rather that having a very extensive but static data base) is the robotic pilots listed in rigger 3.
The real world definition of an AI is nebulous. The definition requiring that it not be discerned from a real person in online chat proved to easy to achieve.
The SR definition of a true AI means it is far beyond the scope of a Decker character. Entire corporate divisions are devoted to research in this aria and so far have only succeeded threw chance.
Robots have an adoption pool that allows them to learn from experience within the bound of there programming. But this is more a rigger toy.
Agent programs may have some of this functionality but it is not represented in there mechanics I believe there is some concern about them evolving into crash of 29 style viruses. The semi autonomous knowbot (sk) may but it is usually beyond the scope of a Decker
I don’t believe any other program has a learning capability.
Edward
hobgoblin
Dec 23 2004, 09:51 PM
the agents have both a hacking pool and a independent initiative score if i recall correctly. this makes them more flexible then the rigger robots with their adaption pool in my book
mfb
Dec 23 2004, 09:52 PM
eh? robots have their own init scores, too.
Edward
Dec 24 2004, 01:17 PM
Independent initiative is irrelevant; a sentry gun has independent initiative.
The hacking pool my matte but I doubt it.
The ability to learn is a defining factor hear. This is represented by a greater chance of success the second or third (or thirty third) time you try to do something. The robots adaptation pool is the only mechanic available in the game to achieve this without outside interference.
That said the agent program is indeed more versatile, this is due to a larger starting code base rather than the ability to learn.
Edward
Gyro the Greek Sandwich Pirate
Dec 25 2004, 12:44 AM
Personally, I'd actually like to see examples of interactions with an agent as opposed to a knowbot as opposed to an AI.
We have good examples of interactions with various AI through canon-but have there been any examples of people being able to differentiate between an agent and an SK and a normal person?
I'm kind of interested in where the dividing lines and what the requirements for 'true sentience' are...
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please
click here.