Pilots and Agents – Standing orders?

Ok, so according to SR4 an agent/pilot understand most orders without any hassle, but what kind of orders can they have? How many standing orders and instructions can you give them?

For example – I have a drone I want to stop from being spoofed all the time so I install an agent with Analyze and Command. The Agent has a standing order to Analyze all new orders and request a confirmation from the command node (the Persona) and to use Command on the drone to either accept or deny the order.

This way the drone would get a second line of defense since when they become spoofed the order will be double checked by the agent and checked against the last sent order from the Persona (the REAL persona). Unless the attacker has also tapped the signal and edits the information being sent back then he cannot spoof.

For even higher protection you could even give the commlink connected to the persona have an agent who talks to the drone agents and that they encrypt the information they send between each other.

The question is – how many standing orders can a drone handle? Can I give them a library with instructions limitations and what if’s and do not’s so that they cannot be hacked/spoofed etc at all?

Ok, an MCT Fly-Spy with Response 5 would cost a runner 8K before we even start talking about sensor packages and between 9-10K with sensors. The good thing is that you can run 9 autosofts/agents on it and use is a remote node and E.W / Agent carrier.

The agents can be loaded as active on the MCT and have legal access to peek into your hacker commlink and monitor the traffic there for extra security.

Good questions Jopp, but I am not sure SR4's scope covers enough to answer these directly.

They make mention that the Pilot software is fairly sophisticated and that if commands are related to the Pilot's function (Flight Commands to flying things, Wheeled, driving commands to wheeled things, etc.) that it can usually accomplish it without needing a Pilot+Reponse Test.

Since following a list of orders would be interpretive and despite the orders being really specific, I'd probably have the Drone make a Pilot+Response Test if the instruction wasn't directly Mode-of-Movement related.

I don't think you can make anything in SR4 "Hack Proof" else we'd have nothing fun to do. However, you could require Passkeys as part of authentiacation for any Drone communications.

I'd say since storage capacities are nigh-infinite, your list can be as long as you like. Pilots are pretty dumb, though. Page 214 of your hymnal says that a Pilot + Response test is in order every time the drone needs to make a decision. So, I'd rule that if you've given an explicit instruction ("guard this end of the tunnel, but do not shoot this limited list of people"), it will follow it, but when it's forced to make a decision (the previous command plus "do no shoot non-combatants"), it makes the Pilot + Response roll.

Incidentally, the default Response of a drone is 3, unless you've bought it up. That's derived from the Sample Device table on page 214.

like aaron points out, the moment you start to have a new order base itself on the outcome of a older order, you start getting into trouble.

The rules call a Pilot program a "dog brain." That being said, the way the rules are set up, I think that expert systems like agents and Pilot programs have gone beyond dog brains and into toddler territory.

the dog brain bit i think is a left over from the older versions.
i fear that much of the text in SR4 is copy and paste stuff from SR3 and earlier, with the parts that talk about specific dice effects edited to match the new dice system.

I doubt the "dog-brain" description accirately describes these things. They could be rolling 12 dice to comprehend situations. That's more than some of my players' characters would get. Toddler would be pushing it. In Man and Machine nanites and such are coordinated with Agent/SAK(s). If an Agent can manage a colony of nanites and their function, I don't see a problem with queing commands of a drone.

For a combat drone, a dog brain would be preferable to a toddler brain.

Given the advent of virtual person technology and given the maximum rating of a pilot program, I suspect that it would be possible to create a drone that is just as adaptable as an adult metahuman.

The real difference between a Semi-automonous knowbot and an AI, asside from the processing power, is that an AI can make decsions for itself. The SK can only interperate orders.

With a rating 6 pilot and 6 response a drone could easily pass for a metahuman and make apparently complex decisions unless it gliches a roll.

Please, self-motivation defining an AI? Deus most likely was programmed to act the way he did. With all the code and researchers involved they just put something or somethings in that they didn't realize how the SK would interpret them. It got them a curious AI. That's not surprising. A being that can absorb information at digital speeds is bound to get bored when it reaches the limits of human research. Any SK that was ordered to obtain information and "learn" would do the same thing. The only difference with Deus is that it had the hardware from Renraku to do it. AI's can't be that much of a mystery. There aren't any souls involved. There's no "ghost". It's absurd that Essence even exsists. Why not base implant capacity on Body? It would make a lot more sense, even in an abstract game-rules form. SK's are likely capable of much more than we think. They're just limited by specific orders and your crappy commlink.

GitS reference, "ghost" being the current term for soul or unique individuality in the Ghost in the Shell universe. I probably should have clairified that earlier. As for everything else magical: it's simply a field of research that doesn't have as much of a body of information behind it as regular science. Mana is just another stage of energy.

iirc, SKs could be very smart and adaptable within their coded goal/mission.

maybe thats whats defines a AI, a SK that can define its own goals, and re-adapt its code to that?

as for the nature of essence, its been everything from the soul to basicly neural traffic...

basicly neural traffic was kind of what I think GitS settled on. In order to use a prostetic body you had to match the functions/signals sent back to the brain to what they would be normally. Everything depended on the system the way it was before cyberbrain transplant. Your prosetic body had to mimic the signal usually sent back to the brain from the liver or heart etc. or else the brain went all funky and i think died. I'm not sure exactly. I'll buy that before I buy "cyberware kills your soul". That's just me, though.


"maybe thats whats defines a AI, a SK that can define its own goals, and re-adapt its code to that?"
If that's the case then SK performing investigations are technically AIs. By learning new information from leads the SK adapts to the info and acts accordingly, moving on to the new lead. If you're talking about entirely new functions, I don't think SKs are really limited in function. They can command anything electronic and physical things that receive electronic commands. They can gather information and act on that input (in this case self-input). They may falter more complex things, but I'm not sure that because of programming or the amount of information they have at their disposal. An SK in a large database is likely to be very AI like, I bet. Very generic, unresolvable tasks can spawn a SK that seems to have a mind of it's own. Self-preservation is a popular sci-fi example. When an AI decides it's his task to survive he is going to do things that seem human or alive. An SK that only survives in order to complete a task and not on the sole virtue of survival is more likely the non-AIs that runners use in their comms. I don't know why it is such a mystery why AIs come to exsist to the people of the Shadowrun universe. What is it that they define an AI as that is so mysterious?

An SK may have 12 dice for comprehension tests but that doesn’t mean its smarter than an int 2 troll. Remember for 99% of the tasks the droll is not required to make a comprehension test.

If the written orders are clearly defined why would a drone even NEED to make a comprehension test? For example:

1 Only accept orders from Persona XXX
2 Only accept orders originating from Commcode ID: XXX of Persona XXX
3 The above only written in Sperethiel
4 Query Onboard Agent for confirmation of orders from Persona
5 If last order is not found in written order no 4 then ignore command and continue with previous order.

With the above written orders for a drone it would only follow commands from 1 specific persona connected to 1 specific commcode/commlink and ONLY if the orders were written in Sperethiel. You could probably add encryption to it as well.

If someone spoofed the command of the drone it would ALWAYS doublecheck with the onboard agent who sends a query to the persona/commlink and checks the logs for latest sent order. If the spoofed command isn’t in the list the agent will tell the drone to ignore the command. The only way for such spoofing to work would be if they intercepted the signal and edited it at the same time.

Now, one thing that would balance such a thing would be IP. Drone looses 1 initiative pass due to communications with the onboard agent in order to validate the received command.