Robot ships

[Carlobrand]

CT rules require a pilot and usually a navigator, engineer and so forth.

CT Book 6 allows the construction of robots with pilot, navigator, and engineer skills, among others.

Accepting that the Imperium itself has a cultural thing about robots, and that maybe passengers prefer living beings at the controls, any particular game reason we shouldn't see robot ships elsewhere? Maybe robot freighters or couriers along the more civilized routes? A ship analogous to the Xboat, with a robot pilot and without need of a stateroom, could serve the Darrians, Hivers or maybe the Zhodani.

 

[aramis]

Carlo - Book 8 is robots, not Book 6. (book 6 is Scouts...)

 

[Carlobrand]

Carlo - Book 8 is robots, not Book 6. (book 6 is Scouts...)

Oops!

 

[MikeinBallard]

I'm familiar with CT and MT, but wasn't there a rule that there needed to be a sophont aboard for jump to work? Or was that sort of canon?

 

[Magnus von Thornwood]

More joys of ATUs.

The robophobia of the 3I is one of the silly things I never understood, so in my ATU I do have roboships.

 

[mike wightman]

I'm familiar with CT and MT, but wasn't there a rule that there needed to be a sophont aboard for jump to work? Or was that sort of canon?

Nope, not in the MT rules or any other version of Traveller. It's an idea that fanon has invented and reinvented several times to rationalise the lack of robotic ships in the OTU.

 

[McPerth]

My thoughts about it:

Starship computers must be quite reliable, as many unforeseen situations may occur and true creativity and intuition must be used when coping with them without outside help (as is in jumpspace). Until reliable synaptic processing reaches (at least) 50%, this does not appear, and that is at TL 16, according to LBB 8.

See that in LBB 8, page 20, under the TL 14 robot brains explanation, it's said:

While robot brains can now include up to 25% synaptic processing, the most reliable traditional computers (such as starship computers) still relly on deterministic massively parallel processing

I guess at TL 15, with up to 40% reliable synaptic processing, this does not change too much, while at TL 16, with up to 60% sysnaptic processing, this may begin to change. Yet, the table in page 29 (fundamental logic programs) give us TL 17 for low artificial intelligence.

So, maybe hivers (whose robot brains are told to be TL 16) might have some true robotic starship prototypes.

Yet, MT:101 vehicles, pages 1-2, give us the rules as how can robot brains be used instead of computers in vehicle design.

Time ago, by using those rules, I designed a fighter drone as a way to cheat the pilots limit in a TCS fleet (purely intellectual exercise, as no such contests were held in Barcelona). Aside from (as told) being escentially cheating (I myself would have seen it as unfair to really use them), this raised another problem if really trying to use those fighters in HG/MT combat system: how were those robot brains rated as computer for space combat modifiers?

IIRC, in Robots supplement for MT it was resolved by using the relevant skills instead of the computer number, as MT rules allow, but in CT (where skills are used in other ways) I'm not sure how could this be resolved...

As for MgT, I guess they could be used there, and in fact I used them for the barges in this thread.

 

[rancke]

IMTU robotic astrogators cause a statistically detectable increase in misjumps. No one knows why, but insurance companies have long ago put the kibosh on such vessels (except for jump torpedoes, where a 35/36 chance of summoning help is better than none).


Hans

 

[mike wightman]

CT rules require a pilot and usually a navigator, engineer and so forth.

CT Book 8 allows the construction of robots with pilot, navigator, and engineer skills, among others.

With a skill level of 4 in each skill too

Accepting that the Imperium itself has a cultural thing about robots, and that maybe passengers prefer living beings at the controls, any particular game reason we shouldn't see robot ships elsewhere? Maybe robot freighters or couriers along the more civilized routes? A ship analogous to the Xboat, with a robot pilot and without need of a stateroom, could serve the Darrians, Hivers or maybe the Zhodani.

I think a robotic x-boat makes really good sense - keep the manned ones for PR stunts but build your network around the robotic craft.

Similarly a lot of insystem traffic could be robotic freighter base, as could the LASH trading that some promote between suitable worlds.

 

[samuelvss]

I think a robotic x-boat makes really good sense - keep the manned ones for PR stunts but build your network around the robotic craft.

Similarly a lot of insystem traffic could be robotic freighter base, as could the LASH trading that some promote between suitable worlds.

I would agree, as long as there are sophants in the loop for the unexpected. I see an X-boat as analogous to a baseball, albeit a very valuable one, thrown by one intelligence through jumpspace to another; what occurs within jumpspace has a much more mechanical aspect.

I do not propose many roboships on their own, however. Type X, surely; Type-S, A, R, etc., never. The S would certainly be enhanced by having robots assisting a single crewmember; IMTU, this is an option discretely adopted by some. Robot gunners, engineers, and general utility. I have robotic janitorial staff as a norm on independent ships. The 3I's culture is such that people are used as much as is possible. Once the stateroom is bought, then often the person is cheaper than a robot; this changes if the stateroom is revenue-generating. I ran the numbers once, but forget the specifics. A basic shipboard robot suitable for a gunner could be had for under 225,000, though. A Free Trader built new under HG2, with the two turrets as one battery, then, gains a stateroom or two.

Second (and even third) engineers on Types C & M, for instance, always seemed a good deal, as did gunners on small craft.

 

[esampson]

The robophobia of the 3I is one of the silly things I never understood, so in my ATU I do have roboships.

Considering that in the very first adventure there was an attempt to make a robotic ship that decided to murder its crew I'd have to say that their robophobia isn't exactly silly.

My take is that the robot brains in the time of Traveller are such that they are capable of being quite sophisticated in what they are programmed to do but they have certain limitations that render them unsuitable for prolonged independent operation (the robotic engineer is quite capable of overhauling the engines when it is told to but trying to program one that can decide on its own that it is time to overhaul the engines is much more problematic).

As a result you always have to have a person available to give orders to the robots. You can have a small ship with a single person (such as an x-boat) with all the functions handled by robots while a human gives them orders and in larger ships robots can do an excellent job of functioning as a sort of 'force multiplier' but as you get more and more robots you are still going to need more and more people to give all those robots orders (as the problems that render them unsuitable for prolonged independent operation also preclude using one robot to order around other robots).

There is also an issue of cost. The brain and programming for a high autonomous robot with Engineering-3 (which is what I would interpret the above to be) would be nearly 300,000 Cr. While in the long run you would eventually save money over hiring someone with Engineering-3 that would take a pretty good length of time.

 

[rancke]

There is also an issue of cost. The brain and programming for a high autonomous robot with Engineering-3 (which is what I would interpret the above to be) would be nearly 300,000 Cr. While in the long run you would eventually save money over hiring someone with Engineering-3 that would take a pretty good length of time.

You can change the effective costs by assuming that robots don't last nearly as long as starships. What if a robot is only good for, say, 10 years? Or a yearly maintenance of 1% instead of 0.1% will do as well.


Hans

 

[Carlobrand]

Considering that in the very first adventure there was an attempt to make a robotic ship that decided to murder its crew I'd have to say that their robophobia isn't exactly silly.

True. I'd limit application to unarmed ships with no passengers and little or no crew. Freight-boats hauling ore from the belt to an inner planet could be roboticized, for example. A 100 dT jump courier built around a robot pilot has no need for a navigator or engineers.

(Ooh. Occurs to me that my IMTU 50dT cutter with the extensible jump field mesh that allows it to occupy 100dT for jump, with a robopilot, can serve in the capacity of the infamous "jump torpedo".)

My take is that the robot brains in the time of Traveller are such that they are capable of being quite sophisticated in what they are programmed to do but they have certain limitations that render them unsuitable for prolonged independent operation (the robotic engineer is quite capable of overhauling the engines when it is told to but trying to program one that can decide on its own that it is time to overhaul the engines is much more problematic).

As a result you always have to have a person available to give orders to the robots. You can have a small ship with a single person (such as an x-boat) with all the functions handled by robots while a human gives them orders and in larger ships robots can do an excellent job of functioning as a sort of 'force multiplier' but as you get more and more robots you are still going to need more and more people to give all those robots orders (as the problems that render them unsuitable for prolonged independent operation also preclude using one robot to order around other robots).

Umm, the robot design bookie has specific provisions for robots that order around other robots.

The real test of the robots is their ability to handle unexpected situations beyond their programmed parameters. A TL15 brain can have up to 40% synaptic processors, so has plenty of "brain" for the situations encountered by most pilots. It would be inferior in a combat or hijack situation or in a wholly novel piloting situation, but for routine piloting on a well-maintained ship along a civilized route, the robot should be more than adequate.

There is also an issue of cost. The brain and programming for a high autonomous robot with Engineering-3 (which is what I would interpret the above to be) would be nearly 300,000 Cr. While in the long run you would eventually save money over hiring someone with Engineering-3 that would take a pretty good length of time.

Considering that a ship may operate for more than 40 years, and a robot has a service life ranging in decades, an up-front investment that represents 4 or 5 years of a human salary is a significant long-term savings. That's not even considering savings associated with not needing a stateroom or life support. If your robot's under half-a-million credits, the ability to build a ship with one less stateroom pays for the robot all by itself, not to mention giving you 4 dT for other needs.

Consider an X-boat designed for a robot pilot. Lack of need for a crew stateroom pays for the robot. Pilot costs Cr72 thousand a year. Over the 20-year career of a pilot, that's MCr1.4 saved from the budget.

 

[McPerth]

The real test of the robots is their ability to handle unexpected situations beyond their programmed parameters. A TL15 brain can have up to 40% synaptic processors, so has plenty of "brain" for the situations encountered by most pilots. It would be inferior in a combat or hijack situation or in a wholly novel piloting situation, but for routine piloting on a well-maintained ship along a civilized route, the robot should be more than adequate.

But most profesionals are precisely needed because all too often "routine" operations/situations are not so. One learns to cope with the unexpected, that is what expertise (and so skill levels) give to a human, and probably the most difficult thing to program in a computer or robot brain.

Will you thrust a multimillion credit worth ship (not to talk about valuable cargoes and passengers) to a brain that cannot react to unexpected situations?

As told before, even while at TL 16 robot brains may be up to 60%, true low artificial intelligence is TL 17 programing, hinting that it needs more than this 60%, and that is quite more than the 40% allowed at TL15.

 

[esampson]

. . .Umm, the robot design bookie has specific provisions for robots that order around other robots. . .

I assume you are talking about the slave modules? Yes, that allows a robot to directly control another but that doesn't mean that the issue with a robot making all decisions on its own is resolved.

The real test of the robots is their ability to handle unexpected situations beyond their programmed parameters. A TL15 brain can have up to 40% synaptic processors, so has plenty of "brain" for the situations encountered by most pilots. It would be inferior in a combat or hijack situation or in a wholly novel piloting situation, but for routine piloting on a well-maintained ship along a civilized route, the robot should be more than adequate.

Actually the real test is what book 8 calls the Fundamental Logic Program. At TL15 the best you can get is High Autonomous. High Autonomous is described as "The robot has the abilities conferred by Low Autonomous and
is able to understand most inferences. Commands can be vague and the robot can still "figure out what you meant". Requires at least the "Basic Command" command program." To be able to run without human intervention I think you need Low AI, which is TL17.

Now you probably could get by with robot pilots under some of the circumstances you've given above. As an example a robotic x-boat pilot would probably be a possibility since the handlers could tell it when and where to go and when it arrives the new handlers could guide it to where it needs to go. You would still be running some risks because if something unexpected happens such as an engine malfunction the robot would have a hard time dealing with it without someone giving it orders.

Considering that a ship may operate for more than 40 years, and a robot has a service life ranging in decades, an up-front investment that represents 4 or 5 years of a human salary is a significant long-term savings. That's not even considering savings associated with not needing a stateroom or life support. If your robot's under half-a-million credits, the ability to build a ship with one less stateroom pays for the robot all by itself, not to mention giving you 4 dT for other needs.

Consider an X-boat designed for a robot pilot. Lack of need for a crew stateroom pays for the robot. Pilot costs Cr72 thousand a year. Over the 20-year career of a pilot, that's MCr1.4 saved from the budget.

The problem with that is that you are no longer using a standard design, so you don't get the standard design discount. If you do make so many of these ships that you are now able to get a standard design discount you now run a real danger that if there's a problem (shortage of robots, cylon revolution, etc.) you have lots of ships that are completely useless because they can't be flown by a person. Mostly even ships where the intent is that they will be completely manned by robots would have a stateroom as a backup.

 

[atpollard]

But most profesionals are precisely needed because all too often "routine" operations/situations are not so. One learns to cope with the unexpected, that is what expertise (and so skill levels) give to a human, and probably the most difficult thing to program in a computer or robot brain.

Will you thrust a multimillion credit worth ship (not to talk about valuable cargoes and passengers) to a brain that cannot react to unexpected situations?

As told before, even while at TL 16 robot brains may be up to 60%, true low artificial intelligence is TL 17 programing, hinting that it needs more than this 60%, and that is quite more than the 40% allowed at TL15.

I think that this is the overlooked key.

Anything that allows a robotic starship at TL 15 (or lower), no matter what the books describe as 'low or high autonomous', will spit in the eye of the core concept that AI is a hallmark of TL 17.

Let's take the 40% (TL 15) and 60% (TL 16) values at face value ... so a robot pilot will successfully perform any pilot task 40% of the time at TL 15 and 60% of the time at TL 16.

[sarcasm] Who wouldn't want to invest in a robot pilot that will only crash the ship 60% of the time at TL 15 and 40% of the time at TL 16? [/sarcasm].

I think robots ships are a great idea, and at TL 17, when they can actually do the job, I suspect that they will be very common. Until then, a large ship may have a human 'chief engineer' and a human 'pilot/captain' and 8 robot crewmen to handle all of the routine stuff ... but don't expect a bank loan without those last few human crew members.

 

[esampson]

. . .[sarcasm] Who wouldn't want to invest in a robot pilot that will only crash the ship 60% of the time at TL 15 and 40% of the time at TL 16? [/sarcasm]. . .

The percentages are the amount of CPU that can be synaptic instead of parallel or linear.

The reason the maximum is important is because to make a robot with a high int using linear units takes 40x the volume and 20x the weight compared to making it with synaptic units (though it is considerably cheaper). Assumedly a robot with Pilot-3 wouldn't crash any more often than a human with Pilot-3 (not entirely true, however).

 

[Carlobrand]

I'm having a wee bit of difficulty with the concept that the design book assigns skills to the device, provides absolutely no penalties for their use by the device, but we nonetheless speculate that we are courting disaster when the device uses those skills. It sounds more like something some of us want to see happen to promote our own view of a biocentric milieu than something canon says is actually happening.

I don't mind a bit of discussion of IMTU preferences, but my intent was to explore how canon handles it. Book-8 tells us of TL15 robot brains: "Computers become 'alive', thanks to sophisticated programming and remarkable visual imaging devices. 'Pseudo-reality' machines can artificially produce the illusion of existence of whatever they are programmed to produce, with the help of holographic (3-D) displays. For example, the personality of a dead individual can be programmed into a computer, allowing one to 'converse' with the dead individual (via the computer) as if that individual were still alive."

Book-8 further tells us, "For the most part, robots are not creative ... However, robots are expert at creating 'decision trees' in their memories by quickly generatlng a complete list of alternatives to a situation. Sometimes this ordinary feat can seem like creativity, but looking inside the robot's brain reveals how this actually works. It is just thoroughness, combined with the ability to weigh alternatives and then choose the best one, that yields 'creative' thlnking in robots."

This tells me that if I have a robot with, say, Pilot-3, I can expect it to do what any human with Pilot-3 would do. What I can not expect it to do is something a human with Pilot-3 has never encountered (or at least the specific set of humans with Pilot-3 that were drawn on to help develop the program), or something outside of the scope of pilot training. Now, perhaps I'm wrong, but absent any rules instructing that a robot with Pilot-3 has some chance of meeting disaster during routine operations that a human with Pilot-3 does not have, I'd have to say that the odds of running into "something a human with Pilot-3 has never encountered" are low enough that a robot pilot can be reasonably used in routine operations - if those events weren't rare, then they'd be encountered by human pilots and by extension be part of the Pilot-3 programming.

Put another way: if the game offered us rules for making a hammer with no accompanying rules for breakage, it would be a curious thing to argue that we should introduce new rules allowing a chance for the tool to break if you were using it to drive ordinary nails in ordinary wood. It might indeed break, but the event is rare enough that few if any of us have personally experienced it and emulating it statistically would be silly. Slam the hammer against an anvil - yeah, time for a roll. Use a tool for the purpose that the tool was designed to perform, in a game where the rules don't offer a chance of failure when the tool is used for its designed purpose - that's more in the gamemaster's-will/act-of-god category than something that can be stated as a percentage chance.

So, unless the ship gets boarded by Ethereal Jump Space Zombies, or meets the Flying Dutchman, or is suddenly infested with never-before-seen superdense-eating hull-lice, the robot should perform as well as a human at piloting. Not so good at dealing with hijackers, but good at piloting.

 

[rancke]

I'm having a wee bit of difficulty with the concept that the design book assigns skills to the device, provides absolutely no penalties for their use by the device, but we nonetheless speculate that we are courting disaster when the device uses those skills. It sounds more like something some of us want to see happen to promote our own view of a biocentric milieu than something canon says is actually happening.

The setting itself tells us that something must be happening. It is a biocentric milieu; if there were no downside to using robots, it wouldn't be. The existing rules happen not to explain what it is that's the fly in the oinment. Is that because robots work perfectly and last for decades? Or is it because PCs usually don't need to worry about their robots lasting for decades and any risk of failure is subsumed in such rules as a natural roll of 2 is an automatic failure?

Put another way: if the game offered us rules for making a hammer with no accompanying rules for breakage, it would be a curious thing to argue that we should introduce new rules allowing a chance for the tool to break if you were using it to drive ordinary nails in ordinary wood. It might indeed break, but the event is rare enough that few if any of us have personally experienced it and emulating it statistically would be silly.

Yes, but would you use the non-existence of rules for breaking and losing hammers to prove that the organization of your choice never buy new hammers, because they bought all they needed many years ago and hammers never break or are oost; the rules say so?


Hans

 

[esampson]

. . .This tells me that if I have a robot with, say, Pilot-3, I can expect it to do what any human with Pilot-3 would do. . .

That's the core of the problem. You aren't making any separation at all for the Fundamental Logic Program. By your reasoning a robot with the Low Data FLP and Pilot-4 will fly just as well as a human with Pilot-4. It will handle every single skill it has as well as a human with the equivalent skill. Given that, why would you give a robot anything beyond the Low Data FLP?

Well, the Low Data FLP says "The robot remembers all data taken in by its sensors. It cannot analyze or learn anything from the data. Commands must be explicit." If you were to give such a robot Pilot-3 then what I would assume is that it would be capable of following an explicit order just as well as a human with Pilot-3 could do. If you say "Fly this heading" it will fly that heading and if something occurs that might prevent it from fulfilling that command (such as one of the engines malfunctioning which causes thrust that tries to turn the ship) it will counter that will all the skill of a human with Pilot-3. It is pretty much not a whole lot more than an autopilot.

High Data is basically the same thing with one exception. It has limited learning capabilities and over time it will get better at Pilot. Now it can fly the heading it was given with a Pilot-4 or 5 or whatever it has developed to. It is still going to follow its orders just as the Low Data would. It is just able to follow those orders better.

So what about Low Autonomous? "Low Autonomous: The robot can take independent action without direct commands and is able to understand simple inferences. Commands no longer need to be as explicit and the robot may be able to "figure out what you meant". It can analyze data and arrive at some very simple obvious conclusions. However, robots with this program are not truly creative they cannot originate ideas on their own. This is not yet artificial intelligence. The robot remembers all data taken in by its sensors and can use the data to learn and gain "experience". The robot can improve the skill level of its application programs on its own. Requires at least the "Basic Command" command program."

This is probably about the first level where you could probably turn over control of a ship for anything more than 'I want to take my hands off the controls'. With the ability to take independent action it would be capable of deciding 'well, he said to fly this way but I'm guessing he didn't know about the asteroid on a collision course' and take some kind of appropriate reaction with its Pilot-3 skill. Before this the robot would have happily flown into the asteroid (with all the skill of a human with Pilot-3) because you didn't explicitly tell it to look out.

High Autonomous robots are going to be better at making spontaneous "decisions" than Low Autonomous robots, but what does that mean? Well, unfortunately the rules don't really tell us and we're going to have to use some of our own capabilities to decide that. What I would assume would mean is that the Low Autonomous units probably use much simpler solutions to the problems, such as simply stopping until they get new orders. High autonomous units would be more likely to go around the problem. They would probably also be better able to handle multiple problems at the same time, such as traffic up ahead while trying to deal with turbulence during a landing. However they still lack the full decision making capabilities that something like a TL17 Low AI would have.