Help - Search - Members - Calendar
Full Version: Questions about AI characters
Dumpshock Forums > Discussion > Shadowrun
Odsh
I would have some questions about creating AI characters, as described in Runner's Companion. I'm sorry if these questions have already been asked, but searching on "AI" doesn't work because it's only two characters long.

QUOTE
An AI counts as a single program toward the process limit of the node in which it is residing. It may also make connections and create icons in other nodes as a normal Matrix user, but it exists in a single node at any given time.

Question 1: if I understand this correctly, an AI (unlike an agent) can do everything without leaving its home node, as long as it has a connexion to the nodes it is trying to access of course. Is that correct?

QUOTE
An artificial intelligence has two Matrix attributes, Response and Signal, that are always the same as the Response and Signal of the node in which they are at the time. The AI also has two Matrix attributes that are independent of the node in which the AI is residing: System and Firewall. Like Rating, these are derived attributes. The System attribute is the average of the Intuition and Logic attributes, rounded up. Similarly, the Firewall attribute is the average of the Charisma and Willpower attributes.

QUOTE
The System program is limited by the base Response rating of the device it is on: if the base Response rating of the device is lower than the System rating, then the System rating is set to equal the Response rating.

Question 2: is the AI's System rating limited by the base Response rating of the node in which it is currently residing?

QUOTE
AIs may acquire, carry, and use additional programs, just like an agent’s payload. Payload programs count towards the node’s processor limit as normal.

QUOTE
AIs prefer to fight in nodes with high ratings, as this grants them an advantage. In nodes with low ratings, they will act like cold-blooded animals in a refrigerator: sluggish and vulnerable

Question 3: is the processor limit based on the AI's System rating, or the System rating of the node in which it is residing?
Neraph
1) For ease of gameplay, yes. An Agent program must go from node to node separately in order to operate. Standard players have their persona stay in their commlinks, and they create icons of their personas on the nodes they are operating in. AIs work the same way as standard PCs when operating in the Matrix.

2) Technically yes. The best way to run this is by actually having his rating be temporarily reduced, and when he upgrades his Home Node he loses whichever penalties would be applied to him from his reduced rating. Although, with an AIs Home Node getting a buff by the simple virtue of having the AI there, this is probably not even going to be an issue.

3) ... Based on the answer for Question 1, it seems they use their Home Node stats for operating in the Matrix. The sentence you quoted may be referring to AIs who's Home Nodes are higher ranked, as opposed to lower ranked.
Odsh
Thanks for the answers.

I'm still confused about question 3 though. The System rating is supposed to impose two different limitations on programs:
1) The rating of any program is limited by the System rating plus the program's Optimization rating (for a total of maximum twice the System's rating), if it has one.
2) On most nodes, the Processor Limit is equal to the System rating. For every multiple of (Processor Limit) programs running, the Response rating is reduced by 1, except for programs with the Ergonomic option.

Moreover, AIs can use two different kinds of programs:
- Inherent programs, which are part of the AI, always loaded and "do not take up any process space beyond the AI’s own". They all have the Optimization and Ergonomic options.
- Standard programs, which must be loaded and "take up processor space as normal".

For inherent programs, the processor limit is obviously not a problem, since they all have the Ergonomic option. They also have the Optimization option, although the rating is not specified. I will assume that the rating of that option is always at the maximum possible, meaning that inherent program ratings are limited by twice the System rating.

Now let's say our AI has its own System and Firewall ratings at 5. Its homenode has 7 in those ratings, and the AI operates from that node. The node is a standard commlink, i.e. not a nexus.

Question 4: are the ratings of the AI's inherent programs limited by a) 2x5=10 or b) 2x7=14?
Question 5: are the ratings of the AI's standard programs that do not have the Optimization option limited by a) 5 or b) 7?
Question 6: is the Processor Limit for the AI's standard programs a) 5 or b) 7?
Neraph
4) The ratings are limited to a) its personal System rating. Here's the quote:
QUOTE (Runner)
Additionally, AI's may improve an inherent program rating by 1 for a Karma cost equal to the new rating; the upper limit of an inherent program is equal to twice the AI's System rating.
(emphasis added)

5) The standard programs an AI would utilize would be limited by the AI's own System, as he loads them into himself, as per the last paragraph of the Codebase section of AI's, on page 89 of Runner's Companion. Also see below - I think that situation may apply as well.

6) It would be both actually. a) the AI, because he loads the programs into himself, and b) the Home Node, because the programs the AI are running are still active in that node.

Again, most of the time the AI's System will be lower than his Home Node, so more often than not situation 5 and 6 will not be an issue - you will only have to worry about the lowest System.
Odsh
I see. So the ratings of all Programs are limited by the AI's System rating, but the number of running programs is limited by the lowest of the AI's System rating and the node's System rating. Honestly, to me this additional System rating is confusing and not very useful. And it is rather a shame that the AI can't even benefit from its homenode's improved System rating.

But thanks for clarifying all this to me.
Odsh
This is really bugging me. I see two possibilities:

1) If the AI's System rating is lower than the node's System rating, everything depends on the AI's System rating.
2) If the AI's System rating is higher than the node's System rating, everything except the Processor Limit depends on the AI's System rating.

However, since the AI's and the node's System ratings are limited by the node's Response rating, the second possibility would only occur when the node's System rating is lower than its Response rating. Which, in practice, never happens. So here we have it: keep it simple and always use the AI's own System rating for everything, just don't forget to lower it to the node's Response rating when necessary.

Did I miss something?
The Jopp
You can really twink out your AI if you really want to optimize it.

Take the pilot origin so that you can use Autosofts and jump into drones.
Basically all autosofts can cover all your skills including hacking/computer skills.

You will be limited to a skill base of 4 due to skillsoft limitations but with optimization options from unwired you can run them from basically any commlink.

Also, since programs are being run on the node the AI is inhabiting it is that Response you calculate limitations from in regarding to running programs.

First of all you have a Codebase of 5 programs on a rating 5 AI
Then you have 8 Ergonomic programs (Rating 5 commlink+AI Rating/2) that doesnt count towards your running programs. After that you can still run about 32 ADDITIONAL programs and autosofts with a -4 to your response.

This would be equivalent of lowering your reaction score of 8 to 4 on a regular human.
A rating 5 AI with Born Rich quality and Piloting Origin can sit in a node with Firewall of 9 and run LOADS of software and still have a main drone body and several well equipped drones.

Add Homeground autosoft since you can use it and gain +3D6 to Matrix perception in your node.
Add an Agent with Replicate and homeground autosoft and Analyze. This Agent can copy itself into all your drones and constantly check for infiltrating matrix threats.
Odsh
Well it's not that I have problems twinking my AI (it's not even my AI, I'm just the GM), I'm just trying to figure out when to use the AI's ratings or the node's ratings.
Saint Sithney
Basically the only time the AI is going to hop out of its home node (and become vulnerable) is when it's in a drone, which, by the way, is a pretty good means for bypassing wifi-negation techniques.
The Jopp
QUOTE (Saint Sithney @ Feb 25 2010, 10:22 AM) *
Basically the only time the AI is going to hop out of its home node (and become vulnerable) is when it's in a drone, which, by the way, is a pretty good means for bypassing wifi-negation techniques.


Unless the drone IS the home node.
Neraph
QUOTE (The Jopp @ Feb 24 2010, 06:20 AM) *
Then you have 8 Ergonomic programs (Rating 5 commlink+AI Rating/2) that doesnt count towards your running programs. After that you can still run about 32 ADDITIONAL programs and autosofts with a -4 to your response.

Kind of, but mostly no. And you're thinking small.

The Codebase of an AI all counts as Ergonomic and prevents AI's from accepting any more Ergonomic programs, since the programs they use have to be loaded into themselves (Runner's Companion, page 89, Codebase, 3rd and 4th paragraphs).

Also, I would like to respectfully show you Software Suites from page 128, Unwired. And I would like to remind you that Autosofts are programs. If you shell out 82,000 nuyen.gif you can get nearly every autosoft packaged into 6 themed Suites, nearly all of which you can be run at the same time, as shown below. The ones that you can't run don't need to be run except under certain circumstances (IE: the Chauffer does not need to be run at all times; neither does the Repairbot or Medic).

[ Spoiler ]


QUOTE (Saint Sithney @ Feb 25 2010, 03:22 AM) *
Basically the only time the AI is going to hop out of its home node (and become vulnerable) is when it's in a drone, which, by the way, is a pretty good means for bypassing wifi-negation techniques.

They don't leave their Home Node to jump in to a drone, they just connect like it like anyone else and jump in. AI's have icons that they are allowed to use in other nodes, just like standard players (Runner's Companion, page 88, Welcome to My Mind, second paragraph).
The Jopp
QUOTE (Neraph @ Feb 26 2010, 08:51 AM) *
Kind of, but mostly no. And you're thinking small.

The Codebase of an AI all counts as Ergonomic and prevents AI's from accepting any more Ergonomic programs, since the programs they use have to be loaded into themselves (Runner's Companion, page 89, Codebase, 3rd and 4th paragraphs).


Actually, no.

The ergonomic programs for the codebase are inherent in the rating of the AI. The AI counts as a running program on the node itself, very much like an agent on a commlink.

This means that the Codebase does NOT count towards the ergonomic rules as they are not running software per se, they are the codebase, the very essence of the AI. They dont even count as separate running programs, they are inherent into the running Agent/pilot/AI sitting in the home node.

First Paragraph, Codebase page 89:
"They are always loaded, but do not take up any process space beyond the AI’s own" which mean that the AI is ONE program running by itself.

Ergonomic programs running APART from the codebase count towards the MODIFIED response/system rating on the node it is running on.

This means that the running programs becomes:
AI = Codebase
Ergonomic programs [running on MODIFIED node's Response, limited by AI System rating]
Other Software

This is very much how agents function today - they use programs loaded on the node they are running.

Hmm, I'll have to check on software suites...
The Jopp
QUOTE (Neraph @ Feb 26 2010, 08:51 AM) *
Also, I would like to respectfully show you Software Suites from page 128, Unwired. And I would like to remind you that Autosofts are programs.


Interesting...

I assume I can also add program options to the suites, but do i do that to the entire suite or per individual program?
Saint Sithney
QUOTE (Neraph @ Feb 25 2010, 11:51 PM) *
They don't leave their Home Node to jump in to a drone, they just connect like it like anyone else and jump in. AI's have icons that they are allowed to use in other nodes, just like standard players (Runner's Companion, page 88, Welcome to My Mind, second paragraph).


No, they don't have to, but it's the only way for an AI to get through a Faraday cage without laying down cable. It's basically the AI equivalent of leaving the van.
Neraph
QUOTE (The Jopp @ Feb 26 2010, 03:02 AM) *
Interesting...

I assume I can also add program options to the suites, but do i do that to the entire suite or per individual program?

Each one individually.

QUOTE (The Jopp Posted Today, 02:41 AM )
First Paragraph, Codebase page 89:
"They are always loaded, but do not take up any process space beyond the AI’s own" which mean that the AI is ONE program running by itself.

Ergonomic programs running APART from the codebase count towards the MODIFIED response/system rating on the node it is running on.

This means that the running programs becomes:
AI = Codebase
Ergonomic programs [running on MODIFIED node's Response, limited by AI System rating]
Other Software

This is very much how agents function today - they use programs loaded on the node they are running.

Wrong on at least two points. Let's start with Agents:

QUOTE (SR4, page 228, Payload)
Agents can be loaded up with copies of your programs so that the agent may employ them on its own.

Agents do not run programs on a node, they run programs that they themselves are loaded with, and these programs are further limited by the Agent's Rating as detailed further in that paragraph.

This is similar but different to how AI PCs work. AI PCs must also load programs into themselves in order to use them:
QUOTE (Runner)
These programs follow the normal rules for programs. The AI may load them into itself, carrying them as would an agent.

Now technically you can argue that "may load them into itself" does not equal "must load them into itself", in which case AI PCs may run programs simply loaded into a node. If, however, you are trying to load Ergonomic software into the AI itself, you can't.
QUOTE (Runner)
Inherent programs automatically have the Ergonomic and Optomize (pp. 114-115, Unwired), but may not have any other options.

QUOTE (Unwired, page 114, Ergonomic)
Note, however, that there is a limit on how many ergonomic programs you can have loaded equal to the processor limit or payload.

As AI's Codebase are simply programs with Ergonomic, Optomize, and cannot be cracked, copied, traded, infected, or run by anyone else, this means that unless you've raised an AI PC in-game, you will start with the maximum amount of Ergonomic programs loaded that you are capable of running.
The Jopp
I still have to disagree to your POV.

I can settle for either a compromise or duel at dawn. grinbig.gif

A few snippets that caught my attention.

Since an AI is neither an Agent or Sprite and more of a person they should have less of the restrictions - and the rules can certainly be read in multiple ways.

QUOTE
Welcome to my mind: P88"
"AI counts as a single program toward the process limit of the node in which it is residing"


Ok, the AI is a single running program

QUOTE
Codebase: P89
"AIs have programs that are part of its code."


The wording here is important. The RAW gives an explanation to why the codebase does not count towards running programs, but what it says in the description is that it is part of the AI’s code – which I interpret as part of it’s singular running process and not additional running programs.

QUOTE
The Body Electric: Page 89
"An artificial intelligence has two Matrix attributes, Response
and Signal, that are always the same as the Response and Signal
of the node in which they are at the time."


Here’s my compromise part. If we would say that the ergonomic programs running in it’s codebase and thus prevents further programs running as ergonomic then we should at least allow it to use the Node’s modified Response attribute – which more often that not, is higher.

Now we come to a contradiction – which rule applies below.

QUOTE
Unwired: P114 Ergonomic Programs
Programs equipped with the Ergonomic option do not count
towards a node’s processor limit (see p. 48) or an agent’s payload
(p. 228, SR4). Note, however, that there is a limit on how many
ergonomic programs you can have loaded equal to the processor
limit or payload.


An AI is similar to an Agent, but it [i]is[/] not an agent. So which apply?

If we count the AI as an Agent it would be worse than a hacker - And why the hell would I pay 110BP when its Response bonus for one wouldn't give it a bonus? Having a Response bonus is kinda moot if I'm forced to use the AI's own Response for ALL running programs and not the full power of it's home node.

The AI should be similar to a hacker but BETTER, with the above interpretation they are actually worse if we count them more like an Agent.
The Jopp
QUOTE (Neraph @ Feb 26 2010, 04:35 PM) *
Wrong on at least two points. Let's start with Agents:

Agents do not run programs on a node, they run programs that they themselves are loaded with, and these programs are further limited by the Agent's Rating as detailed further in that paragraph.


Oh, I thought they changed that? Or was it only that Agents payload counts against the node response? They seem to have removed that - unless it is clarified in SR4A.
Neraph
QUOTE (The Jopp @ Mar 1 2010, 04:00 AM) *
The wording here is important. The RAW gives an explanation to why the codebase does not count towards running programs, but what it says in the description is that it is part of the AI’s code – which I interpret as part of it’s singular running process and not additional running programs.

Welcome to the age-old debate of "Crunch vs. Fluff." What you quoted was a fluff section - it had no bearing on the rules at all; there was nothing in that sentence that actually built on the rules framework we have. The actual rules for their codebase is that they are programs with the Ergonomic and Optomize options. Since that is what they are, they count as Ergonomic programs against the maximum Ergonomic programs allowed to be running.
The Jopp
QUOTE (Neraph @ Mar 1 2010, 09:12 PM) *
Welcome to the age-old debate of "Crunch vs. Fluff."


Allright, I give in and agree with you.

The interesting part of this is then, WHY have a codebase apart from always having a few running programs?

If we skip the codebase for a rating 5 AI in a rating 5 node then he has Response 8 and System 5.

With a codebase he can have 5 ergonomic individual codebase programs.

If we make him an E-Ghost he can run 5 ergonomic suites at rating 5 (or 8 suites if we go by the home node's response.)

Sure, the suites have the problem that they are easier to crash and suchlike but that’s a small price to pay…
Neraph
QUOTE (The Jopp @ Mar 2 2010, 02:08 AM) *
Allright, I give in and agree with you.

The interesting part of this is then, WHY have a codebase apart from always having a few running programs?

If we skip the codebase for a rating 5 AI in a rating 5 node then he has Response 8 and System 5.

With a codebase he can have 5 ergonomic individual codebase programs.

If we make him an E-Ghost he can run 5 ergonomic suites at rating 5 (or 8 suites if we go by the home node's response.)

Sure, the suites have the problem that they are easier to crash and suchlike but that’s a small price to pay…

The reason why to have a Codebase is you're effectively getting thousands of nuyen worth of programs for free.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Dumpshock Forums © 2001-2012