Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Radio Controlled]

Me and syv have hit a bit of a wall on some points in regard to that self-building colony idea I told you about a while ago. Is it ok if we work out some details in here?

sure

Those organic computers from the planetoid, the very adaptable ones. Remember those? The idea is to use them to create the colony's central organizing structure due to them being so adaptable to different circumstances. Would this work? And would it work better than using a regular silicon computer controller, or one based of wetware tech? In general, what are your thoughts on the needs and options for the central controlling unit? 

Hmm. Well it might.

The thing the organic computers do great is changing themselves. They're sort of swiss army knives of computers and can change and adapt.  When you say organizing structure, does that require more an ability to utilize lateral thinking, or more the ability to rapidly change the entire computational infrastructure to handle multiple platforms?

Well, the thing is that the organizing structure would need to perform such a wide range of tasks, from long term planning to short term management to interfacing with different kinds of tech (be it Heabi or sharkmist or whatever), and in such different environments and circumstances, it would really need to be very adaptable and be capable of doing a large amount of different possible tasks (perhaps by 'splitting' itself up into specialized subcomputers).

For example, in one universe it might decide that direct remote control of the robutts is better, and so needs to set up the hard and software to manage lots of different units at once in a central way, while in another it might decide giving the robutts independence would work better (eg cause the cavern walls of that universe interfere with EM waves) meaning it would need to set up something to design and update independent software algorithms to program its servants with.

Of course a lot of creative problem solving would be needed, but looking at the vast amount of possible demands a particular environment in a particular universe might pose, it seems to me the adaptability seems most needed. But remember, I'm not married to this exact concept, I'm just trying to figure out what works best. If you tell me a wetware/silicon 'core' that then uses organic computer subunits for delegating specific tasks to works better, then I'm all for that. What I personally thought is that we could try to 'program in' the needed lateral thinking into the organic computer by taking a chunk of them and running them through a whole bunch of (VR and/or live) sims challenging them with unexpected and difficult situations and crises while managing a colony and seeing how well it fares, as well as maybe presenting them with a bunch of examples of unconventional problem solving to give them a good 'seed' to work off of.

Does this all make sense, or should I explain what I mean more in depth?

[piecewise]

How long does pocket dimensional matter take to break down? minutes? seconds?

how much control do we have over the size of the portal?
is it set in stone, or is there a certain range of sizes to choose from?

speaking of setting things in stone, if we embedded the portal in a large slab if rock then placed said slab on a wagon... could we make our portals mobile?

Depends on the size. It's a volume over time deal, so the bigger it is, the longer it lasts. But still, a few minutes at best, and thats for the entire thing. You generate a tank in there and it will probably break down in seconds after coming out because of all the complex parts in it that are dissolving.

Set in stone, but can be modified when bought. Radio did that.

• Would it be feasible, with ERtech™, would it be possible to create a computerized scope (like you put on weapon, or possibly as binoculars) that automatically analyze someone's general emotional state and propensity to violence using blood pressure, heart rate, breathing patterns etc.?
  
• Would the above be possible to make as a software update for cameyes (though I would suggest that this is less efficient than a dedicated item)?
  

Uh, well heart rate and blood pressure I kind of hard to gauge from a distance.

With future-tech? If MS Kinect can claim to measure someone's heart rate with a camera, I'd think ERtech™ would be able to pull it off reliably. I'm no expert though.

Can the Kinect measure it reliably from a distance? If it can then ER should be able to.

Can inactive teammates be controlled by active players under some circumstances?  If so, under what circumstances is this permitted? 

If I recall correctly, this happened a few times with Feyri during M18, so there is some precedent for this.

I generally let people control other people if
1. the absent person said to do so
2. the absent person just vanished mid mission and now the survival of other characters depends upon their actions.

Me and syv have hit a bit of a wall on some points in regard to that self-building colony idea I told you about a while ago. Is it ok if we work out some details in here?

sure

Those organic computers from the planetoid, the very adaptable ones. Remember those? The idea is to use them to create the colony's central organizing structure due to them being so adaptable to different circumstances. Would this work? And would it work better than using a regular silicon computer controller, or one based of wetware tech? In general, what are your thoughts on the needs and options for the central controlling unit? 

Hmm. Well it might.

The thing the organic computers do great is changing themselves. They're sort of swiss army knives of computers and can change and adapt.  When you say organizing structure, does that require more an ability to utilize lateral thinking, or more the ability to rapidly change the entire computational infrastructure to handle multiple platforms?

Well, the thing is that the organizing structure would need to perform such a wide range of tasks, from long term planning to short term management to interfacing with different kinds of tech (be it Heabi or sharkmist or whatever), and in such different environments and circumstances, it would really need to be very adaptable and be capable of doing a large amount of different possible tasks (perhaps by 'splitting' itself up into specialized subcomputers).

For example, in one universe it might decide that direct remote control of the robutts is better, and so needs to set up the hard and software to manage lots of different units at once in a central way, while in another it might decide giving the robutts independence would work better (eg cause the cavern walls of that universe interfere with EM waves) meaning it would need to set up something to design and update independent software algorithms to program its servants with.

Of course a lot of creative problem solving would be needed, but looking at the vast amount of possible demands a particular environment in a particular universe might pose, it seems to me the adaptability seems most needed. But remember, I'm not married to this exact concept, I'm just trying to figure out what works best. If you tell me a wetware/silicon 'core' that then uses organic computer subunits for delegating specific tasks to works better, then I'm all for that. What I personally thought is that we could try to 'program in' the needed lateral thinking into the organic computer by taking a chunk of them and running them through a whole bunch of (VR and/or live) sims challenging them with unexpected and difficult situations and crises while managing a colony and seeing how well it fares, as well as maybe presenting them with a bunch of examples of unconventional problem solving to give them a good 'seed' to work off of.

Does this all make sense, or should I explain what I mean more in depth?

No, that makes sense. What we'll need to do is imbue it with a set of objectives for it to work towards that are universal across all its different splits and iterations.

[AoshimaMichio]

• Would it be feasible, with ERtech™, would it be possible to create a computerized scope (like you put on weapon, or possibly as binoculars) that automatically analyze someone's general emotional state and propensity to violence using blood pressure, heart rate, breathing patterns etc.?
  
• Would the above be possible to make as a software update for cameyes (though I would suggest that this is less efficient than a dedicated item)?
  

Uh, well heart rate and blood pressure I kind of hard to gauge from a distance.

With future-tech? If MS Kinect can claim to measure someone's heart rate with a camera, I'd think ERtech™ would be able to pull it off reliably. I'm no expert though.

Can the Kinect measure it reliably from a distance? If it can then ER should be able to.

And then there's this alternative way: https://www.youtube.com/watch?v=e9ASH8IBJ2U

[NAV]

What you're trying to do Egan has in fact always been a standard feature in Mk3 suits.
Link to post because you can't quote from locked threads.

[Egan_BW]

What you're trying to do Egan has in fact always been a standard feature in Mk3 suits.
Link to post because you can't quote from locked threads.

Huh. So it is.

Using this quote from you from back when the game was just getting started, I'm going to presume that this is still how it works and that you're absolutely on board with this due to GM infallibility.
Can we take basically that and cram it into a long-range scope?

[Radio Controlled]

No, that makes sense. What we'll need to do is imbue it with a set of objectives for it to work towards that are universal across all its different splits and iterations.

We could try to come up with a set of objectives ourselves and test them, and then alter them if it doesn't work, but I feel that that'd be very difficult to get right.

However, another way to do it is to first give it some basic instructions and show it some simulations of how a colony 'should be' run (in what direction it should think, what to prioritize, ...). Then we let it manage its own colonies in a sim. We throw a bunch of disasters and circumstances at it, and observe the solutions it comes up with, and whenever we see it isn't making the right decisions (eg solving a problem in a way that leads to long term doom when a better alternative is possible) we deduct points from it's internal evaluating system (and add points when it's doing things right), up until it behaves in a way that is satisfactory in most situations we throw at it.

We then look at the end result (objectives, guiding principles, decision trees) and perhaps modify it a bit if necessary. So basically, instead of trying to come up with the objectives ourselves and hoping it will lead to the right behavior, we observe its behavior given the objectives it has devised on it's own, punish/reward accordingly so that it will change these objectives, and let it evolve to its 'own' set.

Does that work? And does it work better than the alternative (can you think of a better way to do it)? If yes, shall we move on to the details of the training regimen?

[NAV]

The inner layers (of a battlesuit) are a more traditional straight meta material that seems designed more to protect against brute force.

Are these inner layers more or less protective against kinetic projectiles then the outer layers?

Is the white hot club hot enough to melt through our spacesuits?

[Moopli]

snip

So, you're essentially saying you should do hyperparameter (or hyperhyperhyperhyperparameter ) tuning over the learning algorithm?

If so, assuming you come up with a good method to tune hyperparameters (and at first, even a good way to select those hyperparameters), then you're probably ok. Better hyperparameter tuning, less overfitting, but lots more training time.

[piecewise]

What you're trying to do Egan has in fact always been a standard feature in Mk3 suits.
Link to post because you can't quote from locked threads.

Huh. So it is.

Using this quote from you from back when the game was just getting started, I'm going to presume that this is still how it works and that you're absolutely on board with this due to GM infallibility.
Can we take basically that and cram it into a long-range scope?

In that case, sure.

The inner layers (of a battlesuit) are a more traditional straight meta material that seems designed more to protect against brute force.

Are these inner layers more or less protective against kinetic projectiles then the outer layers?

Is the white hot club hot enough to melt through our spacesuits?

More.

Yes, but not instantly. It would take a bit and you'd have to be keeping it directly on the suit.

No, that makes sense. What we'll need to do is imbue it with a set of objectives for it to work towards that are universal across all its different splits and iterations.

We could try to come up with a set of objectives ourselves and test them, and then alter them if it doesn't work, but I feel that that'd be very difficult to get right.

However, another way to do it is to first give it some basic instructions and show it some simulations of how a colony 'should be' run (in what direction it should think, what to prioritize, ...). Then we let it manage its own colonies in a sim. We throw a bunch of disasters and circumstances at it, and observe the solutions it comes up with, and whenever we see it isn't making the right decisions (eg solving a problem in a way that leads to long term doom when a better alternative is possible) we deduct points from it's internal evaluating system (and add points when it's doing things right), up until it behaves in a way that is satisfactory in most situations we throw at it.

We then look at the end result (objectives, guiding principles, decision trees) and perhaps modify it a bit if necessary. So basically, instead of trying to come up with the objectives ourselves and hoping it will lead to the right behavior, we observe its behavior given the objectives it has devised on it's own, punish/reward accordingly so that it will change these objectives, and let it evolve to its 'own' set.

Does that work? And does it work better than the alternative (can you think of a better way to do it)? If yes, shall we move on to the details of the training regimen?

Oh thats basically what I want, just a check list of the sort of things that the colony would need like:

1. Create breathable atmosphere of the following composition (data)
2. Secure protective field to reduce radiation to X level

That sort of thing.

[Egan_BW]

What you're trying to do Egan has in fact always been a standard feature in Mk3 suits.
Link to post because you can't quote from locked threads.

Huh. So it is.

Using this quote from you from back when the game was just getting started, I'm going to presume that this is still how it works and that you're absolutely on board with this due to GM infallibility.
Can we take basically that and cram it into a long-range scope?

In that case, sure.

*ER wanders off, giggling evilly.

[Radio Controlled]

snip

So, you're essentially saying you should do hyperparameter (or hyperhyperhyperhyperparameter ) tuning over the learning algorithm?

If so, assuming you come up with a good method to tune hyperparameters (and at first, even a good way to select those hyperparameters), then you're probably ok. Better hyperparameter tuning, less overfitting, but lots more training time.

Yeah, we'll need to pick and refine some hyperparameters at the start either way, though their 'tuning' can be done during the learning process itself I think. Training time isn't all that big a problem I think; seeing as how the Sword now needs several years to speed up just to use the jump points, we can probably even afford a decade or more training time if needed. And that's just for making it work 'good enough', we can continue training (supplemented with real field data) after we sent the first of these out into the wild.

Spoiler (click to show/hide)

No, that makes sense. What we'll need to do is imbue it with a set of objectives for it to work towards that are universal across all its different splits and iterations.

We could try to come up with a set of objectives ourselves and test them, and then alter them if it doesn't work, but I feel that that'd be very difficult to get right.

However, another way to do it is to first give it some basic instructions and show it some simulations of how a colony 'should be' run (in what direction it should think, what to prioritize, ...). Then we let it manage its own colonies in a sim. We throw a bunch of disasters and circumstances at it, and observe the solutions it comes up with, and whenever we see it isn't making the right decisions (eg solving a problem in a way that leads to long term doom when a better alternative is possible) we deduct points from it's internal evaluating system (and add points when it's doing things right), up until it behaves in a way that is satisfactory in most situations we throw at it.

We then look at the end result (objectives, guiding principles, decision trees) and perhaps modify it a bit if necessary. So basically, instead of trying to come up with the objectives ourselves and hoping it will lead to the right behavior, we observe its behavior given the objectives it has devised on it's own, punish/reward accordingly so that it will change these objectives, and let it evolve to its 'own' set.

Does that work? And does it work better than the alternative (can you think of a better way to do it)? If yes, shall we move on to the details of the training regimen?

Oh thats basically what I want, just a check list of the sort of things that the colony would need like:

1. Create breathable atmosphere of the following composition (data)
2. Secure protective field to reduce radiation to X level

That sort of thing.

And do you really want these spelled out? Because these are some very detailed specifications already. I could try to hash something out, but it'd probably be a long list, especially if you want this kind of detail.

Can't we assume the colony learns these during the training period? Because I (no offense) think you're looking at this too small. For example, if a colony (carrying embryos rather than actual colonists in this example) needs a few years to set up shop before anything human-habitable starts to take form, it'd need a whole list of intermediary objectives and instructions before things like 'breathable atmosphere' come into play.

[piecewise]

snip

So, you're essentially saying you should do hyperparameter (or hyperhyperhyperhyperparameter ) tuning over the learning algorithm?

If so, assuming you come up with a good method to tune hyperparameters (and at first, even a good way to select those hyperparameters), then you're probably ok. Better hyperparameter tuning, less overfitting, but lots more training time.

Yeah, we'll need to pick and refine some hyperparameters at the start either way, though their 'tuning' can be done during the learning process itself I think. Training time isn't all that big a problem I think; seeing as how the Sword now needs several years to speed up just to use the jump points, we can probably even afford a decade or more training time if needed. And that's just for making it work 'good enough', we can continue training (supplemented with real field data) after we sent the first of these out into the wild.

Spoiler (click to show/hide)

No, that makes sense. What we'll need to do is imbue it with a set of objectives for it to work towards that are universal across all its different splits and iterations.

We could try to come up with a set of objectives ourselves and test them, and then alter them if it doesn't work, but I feel that that'd be very difficult to get right.

However, another way to do it is to first give it some basic instructions and show it some simulations of how a colony 'should be' run (in what direction it should think, what to prioritize, ...). Then we let it manage its own colonies in a sim. We throw a bunch of disasters and circumstances at it, and observe the solutions it comes up with, and whenever we see it isn't making the right decisions (eg solving a problem in a way that leads to long term doom when a better alternative is possible) we deduct points from it's internal evaluating system (and add points when it's doing things right), up until it behaves in a way that is satisfactory in most situations we throw at it.

We then look at the end result (objectives, guiding principles, decision trees) and perhaps modify it a bit if necessary. So basically, instead of trying to come up with the objectives ourselves and hoping it will lead to the right behavior, we observe its behavior given the objectives it has devised on it's own, punish/reward accordingly so that it will change these objectives, and let it evolve to its 'own' set.

Does that work? And does it work better than the alternative (can you think of a better way to do it)? If yes, shall we move on to the details of the training regimen?

Oh thats basically what I want, just a check list of the sort of things that the colony would need like:

1. Create breathable atmosphere of the following composition (data)
2. Secure protective field to reduce radiation to X level

That sort of thing.

And do you really want these spelled out? Because these are some very detailed specifications already. I could try to hash something out, but it'd probably be a long list, especially if you want this kind of detail.

Can't we assume the colony learns these during the training period? Because I (no offense) think you're looking at this too small. For example, if a colony (carrying embryos rather than actual colonists in this example) needs a few years to set up shop before anything human-habitable starts to take form, it'd need a whole list of intermediary objectives and instructions before things like 'breathable atmosphere' come into play.

Well, in reality you could condense it to

"Make things habitable for humanity"

But the reason I put it like this is because the thing is gonna take what it knows, the world around it and use evolutionary systems to try and create the best possible solution. And without any baseline parameters it might create a base without doors, because that provides the best radiation shielding; and the base is self contained so there's no reason to let anyone outside!


Really we just need a list of basic features you want all bases to have. Not like power generators, but less obvious things. Windows, or rec rooms or quarantine rooms or things like that.

((when it comes right down to it I just want an outline of what kind of base you want so I don't have to spend a while thinking about it and possibly leave things out like a tard))

[NAV]

Assume we set an automated force infuse to really really powerful. Then only charge it a bit and pull trigger. What happens? Does it fire with what power it has, or do nothing?

Tries to fire the power you set and when it doesn't have that power it often either goes unstable or can violently draw potential and kinetic energy out of the area.

1. What does "violently draw potential and kinetic energy out of the area" mean, in effects to the surrounding area and people?
2. What does "going unstable" mean?
3. What are the other possibilities when that doesn't happen?
4. Does the power or AOE scale with the amount or energy the FI needs?
5. Does it still fire the projectile using the energy it sucked out of the surrounding area?
6. What happens if there's no projectile to fire and you try to fire anyway?

[piecewise]

Assume we set an automated force infuse to really really powerful. Then only charge it a bit and pull trigger. What happens? Does it fire with what power it has, or do nothing?

Tries to fire the power you set and when it doesn't have that power it often either goes unstable or can violently draw potential and kinetic energy out of the area.

1. What does "violently draw potential and kinetic energy out of the area" mean, in effects to the surrounding area and people?
2. What does "going unstable" mean?
3. What are the other possibilities when that doesn't happen?
4. Does the power or AOE scale with the amount or energy the FI needs?
5. Does it still fire the projectile using the energy it sucked out of the surrounding area?
6. What happens if there's no projectile to fire and you try to fire anyway?

WHY DON'T YA GET ONE AND TRY IT OUT~

[Radio Controlled]

Well, in reality you could condense it to

"Make things habitable for humanity"

But the reason I put it like this is because the thing is gonna take what it knows, the world around it and use evolutionary systems to try and create the best possible solution. And without any baseline parameters it might create a base without doors, because that provides the best radiation shielding; and the base is self contained so there's no reason to let anyone outside!


Really we just need a list of basic features you want all bases to have. Not like power generators, but less obvious things. Windows, or rec rooms or quarantine rooms or things like that.

((when it comes right down to it I just want an outline of what kind of base you want so I don't have to spend a while thinking about it and possibly leave things out like a tard))

Well, thing is, in some situations it really might be better to keep the people closed in if there's lethal radiation storms outside, whereas in other said storms might be weak enough to warrant letting people outside. I'll quote something from the tinker thread to try and give you a better idea of what I mean:

Spoiler (click to show/hide)

Okay.  For the ships which do have humans on them, how will they interact with the AIs?  Will there be a captain which has supreme authority?  Does the AI have its own moral system which can't be completely overridden by said captain?  How much can the AI say "no" to?  Can it say no to indefinitely detaining or permafreezing a human?  Murdering someone?  Murdering almost everyone on the ship?

I think a sort of 'benevolent dictatorship', like what we have going with Steve right now, would do the trick. The AI has final authority, but is programmed to allow as much freedom to colonists as possible unless it hampers the colony overly much. How the colonists organize themselves will be up to them, but again the AI acts as a sort of buffer against excesses (if the colonists prefer a hereditary 'king' like person to lead them, they can have it, but if said king starts to demand too much or hampers the optimal functioning of the colony too much, the Ai intervenes).

For example, if the colonists decide that permafreezing is the punishment for irredeemable violent criminals, the AI might allow that, but if they then start to use that for petty crimes like a drunken brawl, it will intervene to prevent a lot of useful people getting locked away.

So basically the Ai is a sort of enlightened despot (or benevolent dictator) who tries to give freedom when he can, but will reign it in when necessary (eg. it won't intervene if the colonists wanna do things in a way that's slightly less optimal, but if there's a crisis that needs everyone to act this exact way right now or the colony is destroyed, then it will go full tyrant mode until the crisis is averted).

So, it would basically try to give the colonists as much freedom and quality of life as it can afford without endangering the long-term survival of the colony.

That's why I have difficulty setting up a standard requirements list, because different universes might ask for radically unconventional base designs. From the Wikipedia page on 'Colonization of Mars':

Spoiler (click to show/hide)

Habitats
Storage facilities
Shop workspaces
Resource extraction equipment—initially for water and oxygen, later for a wider cross section of minerals, building materials, etc.
Energy production and storage equipment, some solar and perhaps other forms as well
Food production spaces and equipment.
Propellant production equipment, generally thought to be hydrogen and methane through the Sabatier reaction[39] for fuel—with oxygen oxidizer—for chemical rocket engines
Fuels or other energy source for use with surface transportation. Carbon monoxide/oxygen (CO/O2) engines have been suggested for early surface transportation use as both carbon monoxide and oxygen can be straightforwardly produced by zirconium dioxide electrolysis from the Martian atmosphere without requiring use of any of the Martian water resources to obtain hydrogen.[40]
Communication equipment

So, a list like this would be the minimal requirements for any colony, but how each item is achieved is difficult to predict.

For example, on a planet with good, livable conditions outside, the colony might decide not to build a real rec room and instead just let colonists entertain themselves outside, saving resources for other projects. But then in another universe, where due to circumstances usable space is really limited, it might decide to provide all leisure through VR machines because there just isn't any room for a real life rec room. And in a third it might decide that a regular rec room is an acceptable use of resources. But that all depends on the circumstances, which vary so wildly (at least, from what I've seen from the universes explored already) I dunno if any 'standard requirements list' would be worth anything.

If you are afraid that you won't be able to come up with a good base design for a particular universe, you could always just ask us for inspiration. Or, ask yourself the question 'what are the main difficulties and limitations for this universe? How do they relate to filling need X? How could that need still be fulfilled then?

Remember that you won't have to describe (or even work out) every colony in every universe we send these to, just when necessary.