Robot ships

[aramis]

Skill 2 is better tan average human; under MT (which is my default ruleset for play), Skill 2 makes up for att 4-... The TL8 is going to get you from A to B just as well as the TL13 one, but the interface is "give it coordinates"....

To list the "hundred words" - 75 here -
Commands with proper noun objects (11): Database, system, star, world, moon, port, station, Store, Execute, Delete, record.
Unary commands (6): Manual-override, Self-diagnostics, wake-up, abort, sleep, shutdown
Commands with multiple option (3): Plot (Jump, to)
relational tree words (7): Nearest, Main, secondary, of, in, per, until, for
Definition words (25): one, two three, four, five, six, seven, eight, nine, zero, degrees, minutes, seconds, meters, AU, Light-Seconds, LS, light-years, parsecs, Gee, seconds, minutes, hours, days, weeks, hertz
prowords with objects (6): waypoint, loop, Remote Control (Frequency, Enable, Disable)
Measurement prefixes (10) :nano, micro, milli, centi, kilo, mega, giga, terra, peta, un
Conditional modifiers (6): (emergency, safe, comfortable) (acceleration, rotation, roll)

This should give a workable language interface for an Limited interface computer. It would sound as stilted as siri or google glass.

 

[Carlobrand]

Skill 2 is better tan average human; under MT (which is my default ruleset for play), ...

Does MT have a robot construction set? All I see is a bit of a blurb in Referee's Companion.

 

[esampson]

Skill 2 is better tan average human; under MT (which is my default ruleset for play), Skill 2 makes up for att 4-... The TL8 is going to get you from A to B just as well as the TL13 one, but the interface is "give it coordinates"....

To list the "hundred words" - 75 here -
Commands with proper noun objects (11): Database, system, star, world, moon, port, station, Store, Execute, Delete, record.
Unary commands (6): Manual-override, Self-diagnostics, wake-up, abort, sleep, shutdown
Commands with multiple option (3): Plot (Jump, to)
relational tree words (7): Nearest, Main, secondary, of, in, per, until, for
Definition words (25): one, two three, four, five, six, seven, eight, nine, zero, degrees, minutes, seconds, meters, AU, Light-Seconds, LS, light-years, parsecs, Gee, seconds, minutes, hours, days, weeks, hertz
prowords with objects (6): waypoint, loop, Remote Control (Frequency, Enable, Disable)
Measurement prefixes (10) :nano, micro, milli, centi, kilo, mega, giga, terra, peta, un
Conditional modifiers (6): (emergency, safe, comfortable) (acceleration, rotation, roll)

This should give a workable language interface for an Limited interface computer. It would sound as stilted as siri or google glass.

Well...assuming that by 'TL8' you are referring to Low Data and by 'TL13' you mean High Autonomous, I think we may still be at a bit of disagreement here. Yes, both of them would be able to get you to the destination 'just as well' assuming it is a routine flight. This doesn't mean the Low Data robot is going to plow into the side of a mountain at the first sign of trouble but its limited explicit command set means it will be more likely to get confused or into trouble than a high autonomous robot.

Otherwise what is the purpose of the FLP? I understand it has some effect on the maximum level of Fundamental Command Program the bot can support but since Limited Basic Command gives you 100 words why not just build every single pilot robot as a High Data FLP? Why have any FLPs beyond Low Autonomous since that's the highest required for the Full Command Fundamantal Command Program?

 

[aramis]

Does MT have a robot construction set? All I see is a bit of a blurb in Referee's Companion.

Yes. The one on Drivethrurpg.com is a recent adaptation of the Bk8 material to MT. I've always just used CT Bk8, tho'. Designs are compatible, and for brains, are identical.

 

[aramis]

Well...assuming that by 'TL8' you are referring to Low Data and by 'TL13' you mean High Autonomous, I think we may still be at a bit of disagreement here. Yes, both of them would be able to get you to the destination 'just as well' assuming it is a routine flight. This doesn't mean the Low Data robot is going to plow into the side of a mountain at the first sign of trouble but its limited explicit command set means it will be more likely to get confused or into trouble than a high autonomous robot.

Otherwise what is the purpose of the FLP? I understand it has some effect on the maximum level of Fundamental Command Program the bot can support but since Limited Basic Command gives you 100 words why not just build every single pilot robot as a High Data FLP? Why have any FLPs beyond Low Autonomous since that's the highest required for the Full Command Fundamantal Command Program?

There really is little reason to do so, and it may be a violation of the Shudaasham accords to do so.

The 3I forbids autonomous combat capable units outside the Imperial Military. A ship with reactionless drives is a major weapon all by itself. "Ram 3° 17' 3.45" 375km from beacon ID 4453, starport Regni Downs, System Regina Database Regina Subsector, 6G, safety override" would be a a rather impressive "BOOM" from the far side of Assinoboia.

Anything past the TL13 High Autonomous brain is likely unlawful for a starship.

 

[rancke]

The 3I forbids autonomous combat capable units outside the Imperial Military.

It does? That seems to contradict the statement about how the Imperium does not enforce the Shudusham Concords, leaving the for the member worlds to do (or refrain from doing).


Hans

 

[samuelvss]

The high autonomous TL13 is probably passable as a crewmember. As in, able to replace your pilot full time, under the supervision of a human who probably doesn't need to do much more than give directions. And it's good enough to function as both pilot and navigator at once.

I agree. You state it as I always as I have envisioned it; you want a person to be around to give direction when something happens that isn't covered in the robot's skill set. E.g. the TL13 brain, doubled up to about 500kCr, has Commo-2, as well, and has as part of its instructions to follow commands from the starport authority. If the instructions sound spoofed to the human, he can intervene. I think in most situations, the human is not even going to be giving many directions. You could, at TL13, have a Captain, a Chief Engineer, and maybe a Gunnery Officer if there are enough turrets. The rest of the crew can be robots.

I think the concerns about safety are misplaced. Human error, from disregarding established protocols, is a leading cause of accidents of all types. Even tactical mistakes are often that: ignoring protocols. So we have a robot who doesn't get lazy and decide to cut a few corners, and a human to look for the situations where the protocols do not apply. ("That sounds fishy. I think the comms are getting spoofed.")

Leadership, broker, streetwise, bribery... forget it. Pilot, better than humans, on average.

 

[rancke]

I think the concerns about safety are misplaced. Human error, from disregarding established protocols, is a leading cause of accidents of all types. Even tactical mistakes are often that: ignoring protocols. So we have a robot who doesn't get lazy and decide to cut a few corners, and a human to look for the situations where the protocols do not apply. ("That sounds fishy. I think the comms are getting spoofed.")

But humans make mistakes that a sophisticated autopilot can help avoid (note: TL 7 autopilots is not what I think of when I say sophisticated).

Leadership, broker, streetwise, bribery... forget it. Pilot, better than humans, on average.

A human pilot on his own may well be more error-prone than a TL 13 pilot robot on its own. But a human pilot with an autopilot covers two diffent sorts of errors; a robot pilot and an autopilot covers the same sort of errors.


Hans

 

[Carlobrand]

...A human pilot on his own may well be more error-prone than a TL 13 pilot robot on its own. But a human pilot with an autopilot covers two diffent sorts of errors; a robot pilot and an autopilot covers the same sort of errors. ...

Interesting point. Humans alone are subject to errors from fatigue and inattention. Automated systems are not vulnerable to that, but the weakness of the automated system is precisely what the human brings to the mix: the ability to handle something unexpected or unusual.

 

[McPerth]

Depending upon interfaces the two 'robots' might even share a single body. They just have separate brain interfaces (or even share a brain interface) to the ship's controls.

Well, the robot might be the ship itself, if enough autonomy for the brains is achieved. That means the robot brain will be directly interface with the ship's computer or substitute it.

In fact, in MT 101 vehicles (pages 2-3) is told for vehicles (and I guess extrapolable to starcrafts and, latter, starships) that a robot brain may substitute the computer and a crewmemeber. So a single seat fighter might well be robot brain controled (as the ones I told before).

That's one of the reasons I've never bought the price issue to bear, it is quite cheaper than any computer (and you need 3 computers).

Even tactical mistakes are often that: ignoring protocols.

And so are the brightest tactical (and other) achievements.

That's one of the keys. A computer will be programed to follow protocols. In my work (and I guess in most of yours) we have protocols for many things, but I must admit I often broke them (as I guess most of you) because I think situation requires it, and more often than not I'm right. This is what discriminates true intelligence from good programing (apparent AI), the capacity to decide when a protocol has to be broken and how to act then.

Will this have not been achieved at TL 15 (or even lower)? I don't know, but in traveller is quite clear that it doesn't until TL 17.

 

[mike wightman]

Not true ai, but give it a directive, and the TL 13+ are almost people... but not innovative, and don't learn new levels, just new methods within the same levels of skill.

By the rules as written they do learn new levels - see page 35:

TL 9 high data - ...can use the data to learn and gain "experience". The robot can improve the skill level of its application programs on its own.
TL 11 low autonomous - ...can use the data to learn and gain "experience". The robot can improve the skill level of its application programs on its own.

 

[esampson]

Well, the robot might be the ship itself, if enough autonomy for the brains is achieved. That means the robot brain will be directly interface with the ship's computer or substitute it. . .

The only reason I can see that not happening is one that you still need a human pilot (to my reasoning) so building the robot directly into the ship isn't going to really save you anything. You still have to build the controls for the human in case of a problem. You still need a bridge and a stateroom.

On the other hand if the robot is built directly into the ship and you want to replace it with a different robot (maybe the old one broke, maybe the new one is better programmed) or move the robot to a different ship for some reason that wouldn't be possible if the robot is really 'the ship itself'.

What would be more likely, IMO, would be for the robot to be installed somewhat similarly to a modern car stereo. Somewhere on the ship is a niche of a somewhat standardized shape and size with the brain interface connections for various robots (like a server rack). You slide in the new robot, bolt it down, and then run a cable to connect it to the ship. Easy-peasy.

Some people might be viewing the robot as being 'the ship itself' at that point, or at the very least part of the ship, but I think it is a little (tiny bit) more middle ground than that, being so easy to install and uninstall. Heck, I suppose for the people willing to spend a few extra credits (you're already talking about a 300,000 Cr robot, and this would be something like 5,000 Cr) the robot could have wheel and limbs that fold into the body and install and uninstall itself into the cabinet (holding itself in place with special grippers rather than actually bolting itself in)

Which come to think of it is kind of how R2-D2 worked.

 

[esampson]

By the rules as written they do learn new levels - see page 35:

low autonomous - ...can use the data to learn and gain "experience". The robot can improve the skill level of its application programs on its own.

Yes, but its actual judgement for when to use those skills or application programs never really improves. The robots already have a high enough skill that they are better than human pilots when things are running fine. It's their decision making ability (for decisions outside the immediate 'I need to stop over-rotating') that is the issue with them being fully autonomous.

Even high data robots can learn new levels. That's the difference between high and low data.

 

[mike wightman]

Yes, but its actual judgement for when to use those skills or application programs never really improves. The robots already have a high enough skill that they are better than human pilots when things are running fine. It's their decision making ability (for decisions outside the immediate 'I need to stop over-rotating') that is the issue with them being fully autonomous.

The "judgement" comes from the "experience" which is represented by the skill level.
A robot only has to roll if it has to do something outside of its application skill set - page 45.

There are no rules for robot "judgement to apply a skill to a task" - it is assumed that a robot with pilot can pilot, a robot with navigate can navigate.

The Imperium deliberately keeps robots dumb, other cultures embrace the possibilities

Even high data robots can learn new levels. That's the difference between high and low data.

Yup, I wasn't finished editing my post

 

[rancke]

There are no rules for robot "judgement to apply a skill to a task"

Of course not. Evidently the rules assume that there will either be someone around to tell the robot what to do and when to do it or the robot will have non-discretionary orders to do it if certain anticipated conditions apply..

- it is assumed that a robot with pilot can pilot, a robot with navigate can navigate.

Yep, that's what assumed all right. Just like humans with skills. Note that humans don't roll their skill to decide whether to use their skills either.


Hans

 

[McPerth]

The only reason I can see that not happening is one that you still need a human pilot (to my reasoning) so building the robot directly into the ship isn't going to really save you anything. You still have to build the controls for the human in case of a problem. You still need a bridge and a stateroom.

On the other hand if the robot is built directly into the ship and you want to replace it with a different robot (maybe the old one broke, maybe the new one is better programmed) or move the robot to a different ship for some reason that wouldn't be possible if the robot is really 'the ship itself'.

What would be more likely, IMO, would be for the robot to be installed somewhat similarly to a modern car stereo. Somewhere on the ship is a niche of a somewhat standardized shape and size with the brain interface connections for various robots (like a server rack). You slide in the new robot, bolt it down, and then run a cable to connect it to the ship. Easy-peasy.

Some people might be viewing the robot as being 'the ship itself' at that point, or at the very least part of the ship, but I think it is a little (tiny bit) more middle ground than that, being so easy to install and uninstall. Heck, I suppose for the people willing to spend a few extra credits (you're already talking about a 300,000 Cr robot, and this would be something like 5,000 Cr) the robot could have wheel and limbs that fold into the body and install and uninstall itself into the cabinet (holding itself in place with special grippers rather than actually bolting itself in)

Which come to think of it is kind of how R2-D2 worked.

If I read you well, here you're advocating to put a robot as crew of a ship, while I was talking about making the ship the robot itself (that I guess was the initial idea for the OP). It would be like current drones, that have not even a place for the pilot (while the idea of the robot brain being such that it could be replaced seems quite interesting).

The idea in MT:101 vehicles is also to use the robot brain instead of both computer and a crewmwmber, as shown in the vehicle 23.

Even though, I've stated many times in this same thread I don't believe those true robots exist in the 3I, as to be really independent in action they would need at least low AI (TL 17).

So, how to explain those contradictory thoughts?

I believe those drones (either space fighter drones or as vehicle 23 in MT 101 vehicles exist in 3I, but, being just high autonomous FLP, they still need to be controled by a sophont with true intelligence. Even so, they are a little less reliable than sophont crew should anything unexpected occur, but, as only material losses result of their losses, that is deemed as acceptable.

Even so, their use is limited to contact range with the controling sophont for reaction to unexpected, being drones, not true independent robots.

That line of thought, of course, would rule out robotic starships, as they would be out of reach from the controlling sophont for at least 2 weeks after jumping, while would allow for robotic spaceships that can be guided by a sophont, as the fighter drones I told about or the robotic barges for my own designed jump frame and barges also told about before.

This I would apply even for the Hiver drone armies, whose HQ would be hiver and in contact with them, while the whole army could be formed by those drones and the HQ remain safely in orbit (for planetary assaults) of as far as posible (for space combat).

 

[mike wightman]

I wish I'd thought of this when we were thrashing out Spica - Hiver ship robots could be centuries old with emotion simulation programs. The old timers could be envious of their smarter more modern cousins.

What about upgrades as TLs advance? A TL16 brain that has been around centuries and was first switched on at TL12.

What about copying the data to new machines?

Think I may dust off my notes an start plotting a Hive federation/solomani game again.

 

[atpollard]

The "judgement" comes from the "experience" which is represented by the skill level ... it is assumed that a robot with pilot can pilot, a robot with navigate can navigate.

So what is the difference between a Robot Pilot with 'Data', 'Autonomy' and 'AI' logic? All three have 'pilot skill'.

And then this previous question resurfaces:

Who tells a robot pilot when and where to fly?
So let's say it has safely landed at the right place on the starport. How long does the ship stay there? When is it time to take off?

These are all normal decisions made by a human pilot that have nothing to do with the skill to actually operate a craft.

 

[esampson]

If I read you well, here you're advocating to put a robot as crew of a ship, while I was talking about making the ship the robot itself (that I guess was the initial idea for the OP). It would be like current drones, that have not even a place for the pilot (while the idea of the robot brain being such that it could be replaced seems quite interesting). . .

Yes, I was referring to using the robot as 'crew'. An awful lot of the thread is a debate on how well a robot can function with complete autonomy. One side feels that because of the rule that robots don't need to make judgement rolls when they are working within their skillset any robot with Pilot-4 could fly a ship and would be practically immune to crashing, using Pilot-4 to avoid making poor decisions such as a course correction that might lead to complications at a later time.

The other side (which is the side I agree with) feel that the intent of that rule isn't meant to be that broad. It is more like 'if windshear is pushing the ship off course the robot doesn't need to make a decision to counter it', but more complex decisions that may still be pilot related (is it better to fly around the mountain range or over it?) are related to the Fundamental Logic Program, so Low and High Data robots would use fairly simplistic decision making to determine whether or not to fly over the mountains while Low and High Autonomous would use more complicated decision making routines.

As for the 'drone ship' model, in my view that is only workable as long as there is someone around to help out the robot if it gets into one of those 'tough decisions' spots, probably via radio. The bigger issue I see is that if you build a ship this way and there's a reason you can't use a robot pilot (maybe it's just broken) the entire ship is grounded. If you build the ship so that the robot is a 'plug in' then if the robot isn't working a person can still fly the ship. Additionally you wouldn't be likely to save any money on such a ship because it wouldn't get the standardized ship design discount, unless you ordered the drone ships in massive quantities. Better to order a regular ship and then plug the robot in, even if it gets all its command decisions through radio (or doesn't need someone outside at all).

 

[McPerth]

Yes, I was referring to using the robot as 'crew'. An awful lot of the thread is a debate on how well a robot can function with complete autonomy. One side feels that because of the rule that robots don't need to make judgement rolls when they are working within their skillset any robot with Pilot-4 could fly a ship and would be practically immune to crashing, using Pilot-4 to avoid making poor decisions such as a course correction that might lead to complications at a later time.

The other side (which is the side I agree with) feel that the intent of that rule isn't meant to be that broad. It is more like 'if windshear is pushing the ship off course the robot doesn't need to make a decision to counter it', but more complex decisions that may still be pilot related (is it better to fly around the mountain range or over it?) are related to the Fundamental Logic Program, so Low and High Data robots would use fairly simplistic decision making to determine whether or not to fly over the mountains while Low and High Autonomous would use more complicated decision making routines.

True, and I guess I've already shown in which side I am...

As for the 'drone ship' model, in my view that is only workable as long as there is someone around to help out the robot if it gets into one of those 'tough decisions' spots, probably via radio. The bigger issue I see is that if you build a ship this way and there's a reason you can't use a robot pilot (maybe it's just broken) the entire ship is grounded. If you build the ship so that the robot is a 'plug in' then if the robot isn't working a person can still fly the ship. Additionally you wouldn't be likely to save any money on such a ship because it wouldn't get the standardized ship design discount, unless you ordered the drone ships in massive quantities. Better to order a regular ship and then plug the robot in, even if it gets all its command decisions through radio (or doesn't need someone outside at all).

Sure, that's why I still see it necessary to have a sophont controller, as told several times.

About the price, as I am talking about fighters, I guess the standarized design could well apply, as it's unlikely only one or two are buit, but large quatities of them. And even if it does not, as the use of the robot brain saves you equiping it with a computer, and that's a major saving...