Artificial Intelligence in the Traveller Universe

[Adam Dray]

When Traveller was originally developed, AI was one of those science-fiction constructs. Marvin Minsky had some ideas about computer intelligence and expert systems that were mostly academic. AI in commercially available games -- arguably where the front edge of AI was for a long time -- didn't really start to get interesting until the 90s.

AI in general

Today, AI has successfully defeated our best chess players and best Go players. It's driving cars. It's eliminating a third of a million billable hours of boring legal contract work (link: Bloomberg news article).

I'm not talking sentient AI here. Not yet. No androids (though people are working on those, too). I'm only talking about AI systems that can do one particular thing very well, probably better than an expert human.

Today we're seeing the advent of Internet of Things (IoT), where just about everything is attached to the net. Your fridge. Your tv. Your lights. Your thermostat. Your car. Your phone. The button next to the washer that orders more laundry detergent.

I think the next step is to see "AI of Things." That is, everything is programmed with a certain level of intelligence. It not only makes smart decisions for us, but it also communicates with us (probably verbally) and makes sure it's doing a good job to understand our needs.

AI and the game

There were good reasons not to have clever AI in the Traveller universe. Foremost, it takes away player opportunities. Is it fun to have Pilot 1 or Gunnery 2 when the computer pilots the ship and fires all the weapons?

It's getting harder to ignore the fact that AI is going to do a lot of things better than we can.

1. What does the Traveller universe look like when AI is everywhere?

2. Can Traveller have AIoT without losing the element that makes it fun to play?

 

[Adam Dray]

I submit this idea:

In the future, AI doesn't replace people; it makes them smarter.

Gamewise, you can handle ubiquitous AI in a couple ways.

1. AI systems enhance human systems by lowering the difficulty. You need an 8 on your pilot roll normally but it's 6 with the AI engaged.

(Feel free to flip that around and treat the AI as a +2 skill enhancement instead of a difficulty modifier.)

2. AI systems grant rerolls for the human using them. This is like a skill bonus, sorta.

3. AIs are still limited in how smart they can be. Say, they max out at skill level 2. Traveller characters, then, are the people who are better than the computers. Change the generation rules to produce truly remarkable characters with a few skills in the 3-5 range, then throw problems in their way that require those higher skills.

 

[mike wightman]

I would go the other way.

I think smart machinery and AI assistance are routinely built into TL9+ machinery, which is why skill rolls are so low and very technical tasks like piloting a starship and maintaining its engines and systems (engineering) are reduced to very little more than playing the video games of today.

Keep the skill rolls/task target numbers what ever you use at their given values, and make it harder if you are trying to do something without the AI assistance.

Defining when machines become sentient is the hard part, which I would put into the TL16+ range.

 

[kilemall]

Consider the droids in Star Wars- full-fledged characters, personality, but definitely an underclass. That can keep them of limited utility.

Holding an owner responsible for robot actions keeps the liability up and the ubiquity down.

People can be more expensive to train and pay, but cheaper maintenance.

IMTU some people invest in robots to do work for them, thus earning a profit on money down. With the cheapness of software, switching them to better skill markets is easy. But with the limited lifespan especially of lower TL robots, making back that capital investment can be costly, much less living entirely off their labor.

 

[JimMarn]

I can see keeping combat and pilot skills. I'll semi-quote/paraphrase a line from Robert A. Heinlein's Starship Troopers.

"When the computer, or their combat suits, go down, you still have to fight."

This was in a dicussion on why they still had to learn hand to hand combat, fire a vintage rifle accurately, etc. when their combat suits did everything for them.

 

[nobby-w]

Originally I used this concept to explain away the retro computer tropes of a traditional Traveller setting.

An alternative view of AI from the history of MTU: Butlerian Jihad.

If we take a look at today's A.I. technology, the killer app of A.I. is automated mass surveillance. The majority of funding for A.I. research is from companies who want to sell targeted advertising or customer data or provide automated mass monitoring facilities to other parties, state or otherwise.

At some point in the future, the 'Aibatsus' and state surveillance agencies get so powerful and intrusive that they generate massive social unrest and anybody who is associated with the technology has to lie low or risk publically ostracised or even lynched. A.I. largely dies off as a commercial discipline and state actors who get caught implementing automated mass surveillance platforms tend to get forcibly removed from power, either through legal or violent means. Talking ideologically about A.I. gets people branded as a wingnut in the way that someone advocating eugenics would be perceived now.

The phenomenon, nicknamed 'Butlerian Jihad' for a reference to a classic work of pre-stellar literature, makes A.I. extremely unpopular with the general population. Through its association with ubiquitous totalitarian surveillance and social control, it is viewed as barbaric in much the way that we would view roman circuses or medieval torture instruments as barbaric.

An OTU parallel might be the negative attitude towards the study of psionics in the Third Imperium.

This reverts business practices back to a larger prevalence of manual tasks as an excessively computerised or networked infrastructure is looked on with suspicion. It holds back robotics, automated guidance systems and many other branches of technology that could benefit from A.I. capabilities.

Implementing A.I. technology is largely only done in secret or on a limited basis within accepted social norms.

 

[Carlobrand]

Consider the droids in Star Wars- full-fledged characters, personality, but definitely an underclass. ...

Probably with good reason. At some point, AI starts threatening to replace people who have enough power to do something about the threat. It's one thing to replace your butler, or your law clerk. It's quite another to replace your judge or your CEO. Powerful people will act to protect their power, while the common people may be content to be served by machines but they'll object to being judged or led by them.

 

[Adam Dray]

http://www.travellerrpg.com/CotI/Discuss/showthread.php?t=37369

Um, I don't understand. That thread is titled "PCs and TL" and doesn't contain the terms "artificial intelligence" or "AI".

 

[flykiller]

it wanders into ai.

 

[Lycanorukke]

I think smart machinery and AI assistance are routinely built into TL9+ machinery, which is why skill rolls are so low and very technical tasks like piloting a starship and maintaining its engines and systems (engineering) are reduced to very little more than playing the video games of today.

Which is how I think of it.

Imperium 'Narrow AI' is basically does most of the 'grunt work' that computers handle, with the human providing direction and oversight. All of it is hidden behind the scenes, so nobody really thinks about the fact that a computer is doing most of the work. So while 'General AI' is considered a boogyman to the Imperium, most don't realise that a specialised types of AI are pretty much running everything in a high tech world.

If anything switching off (or destroying) the AI should give a penalty to skills. Now you have to work a lot harder covering the tasks required without computer enhancement. The pilot now doesn't just set a destination with the computer cranking out the optimal flight path, tweaking the engines, adjusting the inertial systems etc - he has to do all this on the fly - think what happens when the 'dumb' flight computer shuts down in modern airlines.

And by the same token, it's a good idea to know how to do things 'manually' (at least on a basic level) for the times the computer does break down or locks up.

 

[BytePro]

1. What does the Traveller universe look like when AI is everywhere?

2. Can Traveller have AIoT without losing the element that makes it fun to play?

A TU with AI is more high tech - and more dependent on such.

If AI is everywhere - its a level playing field. Fun should not be negatively affected. Many Players will probably expect AI as part of the setting.

When it is not, there were will be advantages ... and disadvantages. AIs require systems - and systems inherently have weaknesses.

As for AI's supplanting expert skills - that is equivalent to NPCs doing the same.

 

[JimMarn]

In both Hammers Slammers tanks and replacements limbs and Ghost in the Shell cyberization, best high tech arms and legs, and other such gear requires specialized maintanence. Such a tech level may not be available every where even on the same planet.

Around 10 or 15 years ago some other folks I know brought up, can a robot maintain itself ? One person, pointed out that in order to maintain itself, it would also have to have a copy of its own systems and the maintenance diagrams and supply parts. So if the maintenance part of the robot was damaged, another robot would have to do the repairs. So the second robot, if not the same type, would have to carry two types of repair instructions. Ad infinitum.

Since then, I realized, hopefully correct, that a maintanence robot could handle several robots... unless it was damamged. There would then have to be a robot to just maintain it. So, how is the later of those two kept repaired?

Or have I misinderstood the consequences/requirements of self-repairing robots ?

:CoW:

 

[kilemall]

In both Hammers Slammers tanks and replacements limbs and Ghost in the Shell cyberization, best high tech arms and legs, and other such gear requires specialized maintanence. Such a tech level may not be available every where even on the same planet.

Around 10 or 15 years ago some other folks I know brought up, can a robot maintain itself ? One person, pointed out that in order to maintain itself, it would also have to have a copy of its own systems and the maintenance diagrams and supply parts. So if the maintenance part of the robot was damaged, another robot would have to do the repairs. So the second robot, if not the same type, would have to carry two types of repair instructions. Ad infinitum.

Since then, I realized, hopefully correct, that a maintanence robot could handle several robots... unless it was damamged. There would then have to be a robot to just maintain it. So, how is the later of those two kept repaired?

Or have I misinderstood the consequences/requirements of self-repairing robots ?

:CoW:

<Shrug> same thing as medics.

Remember, doesn't have to be mechanical. A biobot would just be a squishy alternative form of medical.

 

[aramis]

In both Hammers Slammers tanks and replacements limbs and Ghost in the Shell cyberization, best high tech arms and legs, and other such gear requires specialized maintanence. Such a tech level may not be available every where even on the same planet.

Around 10 or 15 years ago some other folks I know brought up, can a robot maintain itself ? One person, pointed out that in order to maintain itself, it would also have to have a copy of its own systems and the maintenance diagrams and supply parts. So if the maintenance part of the robot was damaged, another robot would have to do the repairs. So the second robot, if not the same type, would have to carry two types of repair instructions. Ad infinitum.

Since then, I realized, hopefully correct, that a maintanence robot could handle several robots... unless it was damamged. There would then have to be a robot to just maintain it. So, how is the later of those two kept repaired?

Or have I misinderstood the consequences/requirements of self-repairing robots ?

:CoW:

once you have 3 robots that can self repair, and are of the same type, they can keep each other maintained. A repairs/maintains B, B repairs/maintains C, C repairs/maintains A. The more of a single type, the more efficient this gets.

If your maintenance bot suite is self-repairing (and there's no reason it shouldn't be), you just need the diagrams and spares for the others.

Note that once you get to 30+ or so, if one or two go down, they're parts. As long as the problem isn't a virus, or an environmental interaction that all are subject to (*cough*cough*EMP*Cough) they're stable. When 21 goes down, if you have spares, you fix it. If not, you store it, so when 16 goes down, cross-check 16 and 21 to see which needs fewer parts from the other to return to duty.

Likewise, a general repair bot need not be programmed with the repair schematics of each unit - provided each unit carries a set accessible to the repair bots.

The issue is spares. But the Maker units can (eventually) churn out the needed spares given the specifications... so in a fleet context, again, not an issue.

Which leads to, at some point, "No more humans needed except for directions" - per even T5, that's above TL15. Realistically, it's looking like it might be as low as TL 9...

 

[Lycanorukke]

Which leads to, at some point, "No more humans needed except for directions" - per even T5, that's above TL15. Realistically, it's looking like it might be as low as TL 9...

The limit though, is how smart the 'diagnostic suite' is and how modular the robots are.

Even if the parts are modular, if the repair system can't work out why Unit 21 blue screened (except by switching out modules and hoping it fixes the problem) economics would say keep some trained meat around to dig into the more complicated faults and find the underlying problem.

At least until the electronic brains start catching up to organic ones, which usually starts to appear to be around TL-12.

 

[aramis]

The limit though, is how smart the 'diagnostic suite' is and how modular the robots are.

Even if the parts are modular, if the repair system can't work out why Unit 21 blue screened (except by switching out modules and hoping it fixes the problem) economics would say keep some trained meat around to dig into the more complicated faults and find the underlying problem.

At least until the electronic brains start catching up to organic ones, which usually starts to appear to be around TL-12.

Wrong. Computerized diagnostics are actually already far more capable than the average repair technician. A robot with two probes and a multimeter can much more readily, with much less risk, already test each component individually save IC's... give it 30 probes (which no single tech can do), and you no longer even need to pull many of the IC's to test them.

The best repairmen now, at least in electronics, are utterly reliant upon diagnostic tools, and the spec sheets. The repair bot is highly likely to be able to read the specs faster, and have a far lower handling time for the readings, and be able to test a board in 1/100 the time or less. 0.1 sec vs 5-20 sec per item tested.

Do you know the readings for expected outputs of a Z80 chip? The expected voltages in/out/thru? And that's a relatively simple IC, 8 with under 20 pins total. A human can only test it by removal to a breadboard; a suitable electronics repair unit can simply line up the needed probes and run the test in a few clock-cycles in situ.
Comparably, a modern LIF post-x786 processor still needs demounting for testing, but the bot doing it is far less likely to lift unevenly and/or bend pins on reset, and no human can test it without a dedicated test unit for that interface layout; there are simply too many test conditions needed to verify the chip for a human to do so in a reasonable time.

And for the macro-issues, the checklist is in fact quite simple to assemble, and run through.
For example, the checklist for typical users from my days in Tech Support...

1. Plugged in to wall? (Both power and ethernet)
2. Plugged in to monitor?
3. Monitor plugged in to wall?
4. Keyboard plugged in?
5. Is the power light on on both computer and monitor?
6. Does it give boot sound?
7. adjust bright and contrast on screen...

This was just de rigeur "ask these when told «my computer won't start.»"
A small neural net could verify these faster from a photo than I could by a quick glance.

 

[flykiller]

Or have I misinderstood the consequences/requirements of self-repairing robots ?

yes. the repair functionality - any functionality - is just algorithms, which can be downloaded to any machine, the only limitation being any given machine's fixed physical capabilities.

"can you fly a helicopter?" "... now I can."

 

[McPerth]

I would go the other way.

I think smart machinery and AI assistance are routinely built into TL9+ machinery, which is why skill rolls are so low and very technical tasks like piloting a starship and maintaining its engines and systems (engineering) are reduced to very little more than playing the video games of today.

Keep the skill rolls/task target numbers what ever you use at their given values, and make it harder if you are trying to do something without the AI assistance.

Defining when machines become sentient is the hard part, which I would put into the TL16+ range.

I fully agree with this. I've always assumed that the very low crew needs for Traveller ships are precisely due to the many smart systems the ship has.

Likewise, I've also ever assumed that robots are nearly omnipresent in Traveller, just not usually told about unless they have some importance in the plot or situation (as bystanders, vehicles, and many other things we usually asume to be there, but rarely talk about).

The only field where I assume robots not to be present (or at most in marginal use) are those that need interpersonnel relation.

Of course there are autodocs, that can treat you most physical damages (if they can be treated) and keep you alive, but I don't believe there are autopsyquiatrists , nor autopsycologists, as those need more interpersonnel relationto be efective, and the robots (or any smart system), not being sentient, are not effective for those cases.

 

[JimMarn]

Interesting.

I have been told I have good tech skills as I do check cables and power first. For the instances I was involved in, that was 70-80% of the problem.

The other was network connectivity.

A very few times, out of around 300 plus desktop computers, it was the motherboard, the case fan, or the case video card.

For laptops it was either not enough ram, a dead spining hard drive or a failed SSD hard drive, or the cooling fan was clogged with dirt.

For Radio frequency equipment, we had one that failed when a cable tie broke and a wire wandered, due to heat stresses, over into another area under the chassis. There was no short, but it allowed high frequency signals into the power supply.

Another puzzle I had, in the air search radar, was the signal creation for the transmitter was erratic. After several days trying to find the problem, I got permission for the FCC first class license holder, from radio repair, to assist me.

Turns out that the tiny transformer used to provide heating to the vacuum tubes, to give better frequency stabilization to the output pulse, was intermittantly breaking down. Only a megger, used to check if the case and the wire inside were conecting or not, found the problem.

A short sf story I read years ago turned on the starship's robot/computer pilot became convinced that the owner was no longer on the ship, and wouldn't take orders from anyone. The owner was still there, but due to a shipboard accident, his voice was altered. They did finally convince the ship it was okay to proceed to a planet so the owner could get medical attention, instead of beng stuck a large number of light years away for help.

It refused to let them 'fix the problem' and locked all of its spares, electrical cabinets, etc. so they couldn't open them. It did allow; food, water, and oxygen; and not much else.

 

[flykiller]

http://freebeacon.com/national-security/freighter-autopilot-hit-us-destroyer/