Artificial Intelligence Player Characters

[veltyen]

One of my players wishes to play a militarised AI in my current campaign. The campaign is somewhat more automated then the OTU so this isn't a problem.

What I am stuck with is the creation of a meaningful AI PC. There are some guidelines in the design section, (1st Ed THB 226/227) but no clear rules.

Here is how I currently intend to handle it, and the problems I forsee.

The PC exists inside a collection of vehicles designed to work as a single battlefeild unit. The vehicle that hosts the (low) AI, as well as several subordinate units normally using both low and high autonomous logics.

Intelligence and Dex mod are calculated as per normal, as are charisma and soc. There is no meaningful way to calculate strength or con, as these are dependant on external structure. The dex mod adds to that of dextrous appendages implying a more accurate machine control then is normally possible using meat.

Education is a sticking point, mainly because of how I am handling skills. Since the computer the AI inhabits can change its current skill load-out, this would imply a moving education dependant on what other programs were loaded. This is in some ways a nice way to handle this, but I am loathe to have something so core to the character to be a moving target. That and under the design rules the AI could choose to have an EDU in the order of 200. I am currently leaning to nominating a starting EDU skill, rather then work out anything that feels right ruleswise.

I am handling skills in a split fashion. There are the skills of the machine external to the AI, and there are the skills that the AI has innately. I am using two additional skills, smallarms and melee, to take the place of BAB, both are based on dex. The skills internal to the AI are those gained from class levels. These are restricted to the normal class maximums. The skills external to the AI exist on concurrent hardware and have a maximum rating dependant on the hardware.

The AI is automatically proficient with any hardware built into itself. Additional proficiences will allow control of other devices. For example an AI in a grav tank could pilot that tank without penalty, however it could not proficiently pilot another type of grav vehicle without having gained the feat from levels. The AI only has "T/Comp" as a world skill, no others, and does not have any starting character feats.

The AI has no saving throws as such, and BAB is relatively pointless.

The size of the AI was going to be 15PP plus 5 per level plus the size of programs I consider inbuilt (ie. not subject to being swapped in and out).

While the network has an enormous ammount of PP available (the central machine is running Master and Server programs most likely) a large ammount will be restricted to quite low levels, as a large ammount of the processing power is in M2 computerised battledress. This should stop the easy application of 15-20 Rank skills, especially when trying to control a large section of itself.

Everyone must have had a player or two pipe up and ask to be an AI, how did you handle it?

 

[RainOfSteel]

Do AI's have sophont rights in YTU?

 

[veltyen]

Depends on the planet. The imperium does not grant full citizenship, but does recognise AI's (especially High AI's) as worthwhile beings in their own right. They cannot hold imperial offices of any type, nor can they join imperial forces. However willful destruction of an AI is considered a crime on a near-par with murder, though often murdering an AI is overlooked due to extensive bigotry and bias. Most larger naval vessels will have an intelligence inhabiting them, these are born for this purpose, and can never leave. Megacorps often use a central AI as a record repository, theoretically these AI's could quit their job and become free citizens in practice this is very rare.

Some of the core planets to MTU have AI's caught up in their running of government, in one case as a head of state. Individual planetary laws vary widely, it is not unknown to accord machines no rights and require an owner of record. Conversely some of the more liberal worlds concur full citizenship to both high and low AI's.

There is a loose organisation of both meat-sentients and machine-sentients that lobby and act for the rights of machines. The Sentient Engine Liberation Front would like all AI's to be treated as people, with the freedoms that entails, on all worlds. A noble goal, but servely curtailed by culture and history.

Low AI's in general need a guardian, someone that is capable of looking out for them, much like a child. High AI's can generally look after themselves.

There are two ways to "obtain" a low AI. You can array a team of several thousand programmers trained to TL15 and build one from scratch. This takes a couple of years to set up, and then can start producing low AI's on a regular basis. The other way is to ask a High AI to birth one for you. The High AI extracts part of itself into a low AI, this takes some time of intense concentration and is not undertaken lightly. Depending on the morality of the High AI this could be thought of as a child.

The only way to obtain a High AI is through breeding. Two (or more) High AI's join together and mix their traits into another intelligence. This is dangerous for all of the AI's involved, and takes considerable ammounts of time.

 

[mike wightman]

What's to stop a High AI copying itself to produce a "clone"?

 

[Malenfant]

Pick up the Transhuman Space RPG (published by Steve Jackson Games). Although it's a rather different background to Traveller, and written for GURPS 3e, it should give you a lot of ideas of how AI are handled in society and by players (there is a "digital mind" PC option there, encompassing uploaded intelligences and silicon intelligences, and they do have the option to switch into new robotic bodies.

While backups of digital intelligences are allowed in TS, "xoxing" - copying an active AI without deleting the original and running more than one copy at the same time - is however illegal in most nations. The odd thing is that biological cloning is fine, but this is basically the same thing. The official designers line, as far as I can tell, is that the technology is still in its early stages and given fifty to a hundred years it's likely that xoxing will be accepted by then too.

 

[veltyen]

What's to stop a High AI copying itself to produce a "clone"?

A mighty sheild of handwavium. I don't allow direct copies to be created of AI's. Effectively I have just stated that the state of an AI cannot be recorded effectively enough to transfer. AI's cannot be shutdown or backed up for the same reason. This is different from spinning off a lesser entity to be latter intergrated back into the core AI, this is what High AI's are effectively doing to create Low AI's, normally they would be building specific purpose LAL and HAL systems.

High AI's aren't available till TL17 (I beleive), at that point I would conceivably have to allow cloning and transfer. MTU is currently effectively capped at TL15 with the occasional non-imperial higher level planet. So generally I don't need to look into that problem.

 

[Captain Zoom]

Another variation is the "Brain Ship" concept used by Anne McCaffrey in her books. Instead of true AIs, she uses human brains harvested from babies who have deficient bodies (i.e. they'd die anyway), and they are trained to run starships, research facilities, space stations, even planetary networks. The Brains have full sophont status, but start off as indentured servants (since they have to pay back the costs of their training, etc.), but can purchase their freedom and become free entities for hire. They have access to high speed computing, etc., and can do many things that regular people can't but they're still human brains, so you can limit their skill levels, etc., and get away from the idea that you can simply add skills as needed (which make PC AI's too powerful in some games), and you can't copy them since they have a biological component. It might also be a way to allow AI-like creatures before TL15 since the biological support systems and computer linkages would probably be available pre-TL15.

 

[veltyen]

"The ship who sang" and so on. I am familiar with the series.

My take is that they were building AI's, and the complex ruse involving crippled children was so that they would have some empathy towards people and vis-versa.

The merging of these two concepts (reading brains into machines) I allow at a slightly lower tech level(14). Synthetic AI is therefore available sooner, and has the advantage that the (now) machine intelligence formerly was a meat being with the empathies that that implies. Generally a synthetic AI has the same legal problems a low AI has. Useful as an advisor to your descendants, but no longer an independant being.

IMTU most purpose built AI's tend to be sociopaths.

Getting back a little more to topic, the campaign in question has been going a while (2 years rt, close to 6 in game). The original characters are getting towards 20th level, and some are still active and available for play. This means that basically I am mostly unworried by power level considerations. So while the AI may be able to load skill rank 21 in multiple areas simeltaneously this isn't overly out of place.

This would be a consideration for other games I run.

 

[veltyen]

wisdom=empathy

A thought trapped by another stream of conciousness. Whirling.

Machines have no wisdom. Machines that were never meat may have less.

An empathy program, to allow some consideration for how others may act and think. A program to feel.

I think the use of this will be had.

 

[MrMorden]

AIs might have some very interesting abilities, and limitations.

When operating any electronic device, I might give them the same bonuses as if they had a neural interface to the device, just due to the understanding one programmable machine has for another.

On the downside, their skillset is largely determined at the time of creation. They might not adapt and learn as fast as humans, and thus only get 1/2 or 2/3 experience points from encounters.

Alternatively, you could give them a skillpoint allotment up front, but say that the only way to gain other skills is to purge existing ranks on a 1-to-1 trade basis. So if the AI wanted Pilot skill, he'd have to reprogram himself by dropping the same number of ranks in another skill he has. This would make the AI more versatile than a human, but ultimately more limited since a human can learn more and keep all his/her skills available at once.

I guess the point is, to make sure you balance the AI in the game by giving it some really cool abilities, but also some very real limitations.