AI, Synthetics and Posthumans

[ParagonExile]

Yay first post :3

We know that currently there are a handful of general species phenotypes, like mammalian or anthropod, and that is all well and good. But notably there is no "synthetic" category, and this irks me. I want to play as this handsome guy:

We know there will be robots and artificial intelligence in some form, as we can build them and they can potentially revolt, but there is no category for them. It would be interesting to be able to play as AIs, for obvious reasons. They could potentially be the most unique category of life, with immortal rulers, unique governments (like universal consensus democracy, or hive mind) and no restrictions on planet settlement, perhaps with some unique drawbacks as well. Or even just say they are machines and give them the same mechanics, letting us roleplay instead and filling them out in some expansion in the future alongside the others.

Likewise, it would also be great to play as a race who were once biological, but attained a technological singularity and transformed into machines, becoming posthumans for lack of a better term. They would be functionally identical to AIs as far as the game is concerned, but it would give them a lot more personality.

tl;dr - "Synthetic" should be a classification of alien race.

 

[TheAnguishedOne]

They've said in other threads that while they want to include an AI race, it would be hard to represent. We can expect it as DLC. Until then, mods.

 

[aitaituo]

A species first achieving interstellar flight is unlikely to have already been supplanted by a synthetic race they have created. A purely mechanical race that evolved on its own is probably impossible. It doesn't make sense for synthetics to be in a starting position of Stellaris.

A fallen empire of synthetics would be feasible, but as said, resources have not been allocated to develop such an AI-only (ha, pun) position in the initial release.

 

[ParagonExile]

A species first achieving interstellar flight is unlikely to have already been supplanted by a synthetic race they have created. A purely mechanical race that evolved on its own is probably impossible. It doesn't make sense for synthetics to be in a starting position of Stellaris.

Considering that we could have sapient machines by 2050 it's not exactly a massive stretch, or at least no larger a stretch than FTL travel or spaceships fighting each other. Do we even know what year the game takes place in?

We don't know the technological progression of the alien races from before the game starts either.

 

[aitaituo]

I am somewhat skeptical of futurists who think we could have strong AI in the next 100 years.

 

[King of Malta]

I think the stumbling block may be the Genetic Traits. We don't know what the current ones are and the only ones that could themeaticly fit with AI would be specific to them only and no other type of species.

 

[ParagonExile]

They've said in other threads that while they want to include an AI race, it would be hard to represent. We can expect it as DLC. Until then, mods.

That's unfortunate, but thank you for letting me know

 

[ParagonExile]

I am somewhat skeptical of futurists who think we could have strong AI in the next 100 years.

Not futurists, people working for Google, Microsoft, IBM and DARPA. Their estimates were that they were somewhat confident that it could be done by 2050, and almost certain it could be done by 2100. A working general AI is the holy grail of robotics.

Regardless, it's a science fantasy game. Sapient machines are a hallmark of the genre and not including them is a large wasted opportunity. Not including them as an option because of a realism argument would be rather shortsighted.

 

[valergain]

I am somewhat skeptical of futurists who think we could have strong AI in the next 100 years.

We'll see.

Regardless, it's a science fantasy game. Sapient machines are a hallmark of the genre and not including them is a large wasted opportunity. Not including them as an option because of a realism argument would be rather shortsighted.

Don't know why you decided it was a science fantasy game everything so far has a reasonable grounding in science. It's not totally hard sci-fi either but it's defenitly closer to that than it is to science fantasy. Regardless one the end game disasters they mentioned was a robot rebellion and they have said they might do machine races as DLC.

 

[aitaituo]

Not futurists, people working for Google, Microsoft, IBM and DARPA. Their estimates were that they were somewhat confident that it could be done by 2050, and almost certain it could be done by 2100. A working general AI is the holy grail of robotics.

Not to pooh-pooh you too much, but these are often the same people who insist Moore's Law is still literally true, while others have been frantically trying to get around the red brick wall. Personally, I'm not going to get too enthused by AI predictions until non-binary transistors (or replacement) are in widespread use.

 

[ParagonExile]

Don't know why you decided it was a science fantasy game everything so far has a reasonable grounding in science. It's not totally hard sci-fi either but it's defenitly closer to that than it is to science fantasy. Regardless one the end game disasters they mentioned was a robot rebellion and they have said they might do machine races as DLC.

I'm a bit iffy on that, it rides the line. Things like single-biome planets, humanoid aliens, widespread FTL travel etc are definitely more on the fantasy side. But you're right.

I do hope we can take control of the rebellions, robot or otherwise, as we can in Crusader Kings.

 

[ParagonExile]

Not to pooh-pooh you too much, but these are often the same people who insist Moore's Law is still literally true, while others have been frantically trying to get around the red brick wall. Personally, I'm not going to get too enthused by AI predictions until non-binary transistors (or replacement) are in widespread use.

Moore's Law is still a go until 7nm chips are deployed around 2020. After that, I'm not qualified to say.

That said, this is a game, and we don't need to take things too seriously. Rule of cool and all that

 

[Krajzen]

Considering that we could have sapient machines by 2050

Absolutely no way. There is absolutely no way to develop actual thinking, sentient, conscious AI in the foreseeable feature, because it is completely beyond linear programming we do; it would require reverse engineering human brain, which is so mind-boggingly complex and connected with various bodily functions I can hardly imagine a necessity of investing billions into a sentient machine, seeing as we already have 7 200 000 000 of them.

Don't get me wrong, we develop AI and we need it. But we need soft AI: 'intelligent', 'learning' algorithms and programs to operate robots, factories, tools, simulations, computers etc - but all of that doesn't require sapience, consciousness, thought, semantics, and is very hard enough to do. We need and can do AI programs with the 'intelligence' of bacterias or plants, but we have made almost no progress of developing actual 'thinking' machine since 50s when the term AI was invented. There is not a single AI program in the world which actually understands what is it doing, even the most complicated ones are just algorithms after all.

To do an actual sapient machine you would need a perfect knowledge of human, or at least advanced animal brain to reverse-engineer its neural webs and connections, and we are very far from it. On top of that, we would need to teach such machine semantics and language and actually invent the entire new psychic and cognitive model for it to make it sapient, as even the most perfect AI would lack a lot of biological factors necessary for sapience. All those things are so ridiculously beyond our abilities I would be very, very surprised if I have seen a 'sapient machine' in a 21st century.

In the early 50s or 40s the first concepts of AI have been created, when mathematics were far more developed than our knowledge of mind, and some AI theoreticists thought 'well human brain is basically a computer so if we'll put some algorithms into it, we will create AI! It will happen in like 20-30 years' and since then we have achieved a lot in terms of creating practical AI and absolutely nothing in terms of creating actual sentient machine.

And no, there won't be a sudden great breakthrough and sapient machine being created by some genius - we are not dealing with one big problem here which, upon solving it, would present us with a scientific revolution; we are dealing with a lot of big problems here that are connected with each other in extremely complicated, deep ways we don't understand.

And we still cannot grasp how the hell some meat can use electricity and biochemistry to craft language and the feeling of self - this is fucking insane - so, well, we should try to understand that before any serious attempts at creating AI. If you looked on the human's neural structure using microscope you would see billions of neurons but not a single trace of thought processes - but it is somehow spawned and stored somewhere here. How we can design sentient machine if our own sentience is beyond our grasp and comprehension yet?

 

[ParagonExile]

Absolutely no way...

You're using an argument from ignorance, specifically the argument from incredulity. And you say several things as facts which are not the case. This isn't an insult, but you are using a logical fallacy and facts that are decades out of date. When I said "possibly by 2050" I wasn't using my own words, but of industry leaders who have billions of dollars riding on it.

But what does this have to do with OP? We're talking about a game. Robots are cool, ergo they should be in the game.

 

[Nirmara]

The actual paradigm in AI research recently shifted to machine learning which require less input from human programmers in order to create AI doing complex task. The main advantage of this approach is that it doesn't require understanding the brain in order to create a complex AI. Also, an AI do not need to think exactly like a human brain in order to be a strong AI; it just have to exhibit the same cognitive capacities.

Considering the actual paradigm, the main limit is imposed by the hardware, not the software. Quantum computing might be the answer to this problem and we are making some progress towards it, but we still are very far from making it functional. Still, a breakthrough in quantum computing might make the creation of a strong AI a distinct possibility in the near future.

You must keep in mind that science doesn't work linearly but by paradigm shift (see Thomas Kuhn) making prediction of the state of any scientific field in a decade or more risky to say the least.

A species first achieving interstellar flight is unlikely to have already been supplanted by a synthetic race they have created. A purely mechanical race that evolved on its own is probably impossible. It doesn't make sense for synthetics to be in a starting position of Stellaris.

A fallen empire of synthetics would be feasible, but as said, resources have not been allocated to develop such an AI-only (ha, pun) position in the initial release.

The rebels/refugee AI from a fallen empire is an interesting way to portrait synthetic specie but it is not the only one.

Stellaris is about emerging species in the galaxy. As long as we accept the possibility of different technological paths, it is conceivable that an organic specie developed a strong AI before FTL. This premise make it quite easy to imagine an AI specie in the position of an emerging specie.

Here are a few examples:

An AI that revolted against it's creators might have discovered FTL recently or had it but not the interest in galactic exploration/expansion until the moment the game started.

An AI based on machine learning or swarm intelligence could develop sapience, by itself, even if it's creators have been extincts as long as a significant portion of their economy was automated under said AI.

An organic specie could choose willingly to turn into cyborg (in order to survive to war or climate catastrophes, to improve themselves, etc).

 

[*Aqua*]

@Krajzen
There's a number of things you got wrong in your post. For example our brain isn't too complex to understand. We already know pretty well how it's constructed and how it works on the fundamental level, what specific regions can do best, etc. And we can replicate all of that.
Our brain isn't magic. It's only complex and complexity only needs time to understand.

Atm we can't build an artifical one because it would cost more way more than a few billions and no one is willing to pay that.

Also your argumentation has another weakness: You assume that creating intelligence is only possible by replicating what nature did. That doesn't need to be true. A lot of our technology works completely different than nature and can't be found in there, i. e., the wheel.


Back to topic:
I would also like to play a robot or android nation. If done right they could work completely different than organic species. For example a robots doesn't need morals or money to do it's work. This could provide an interesting and fun game.

 

[ParagonExile]

*Science fantasy talk*

Note that I said I agree, it walks the line. From what the dev diaries have noted, it is a 3 on the scale of sci-fi hardness.

As for the constant "It's just a game" rhetoric, just because it's a game doesn't mean it has to go all low brow and just throw things around because "cool". Just because something is a book, game or movie doesn't mean it has to be nonsensical for the sake of cool. That is the line of thinking that has made Michael Bay and Roland Emmerich very rich men in hollywood,where action and cool override logic, pacing, and plot(and internal consistency).

Robots and artificial intelligence are pretty central to the genre of the game, and they will be making an appearance regardless, so including them is obvious. Nothing I said is remotely similar to what you compared it to.

Also, unlike previous historical Paradox games, there's no history or grounding for the universe aside from the one they invent. Stellaris will be more game-y by default.

 

[Fuu]

I also want to be able to play as a synthetic race, but I fully expect synthetic races to be to Stellaris as theocracies (or at least, the Papacy) were to Crusader Kings.

 

[Nirmara]

I also want to be able to play as a synthetic race, but I fully expect synthetic races to be to Stellaris as theocracies (or at least, the Papacy) were to Crusader Kings.

Theocracies are unplayable in CK because it goes against the main aspect of the game: building a dynasty. There is no reasons to think a synthetic specie would be incompatible with any of the main requirement to be a playable specie.

 

[Fuu]

Theocracies are unplayable in CK because it goes against the main aspect of the game: building a dynasty. There is no reasons to think a synthetic specie would be incompatible with any of the main requirement to be a playable specie.

I agree with your point, they wouldn't be incompatible as such. However they would be significantly different to the organic races, and would likely have radically different gameplay mechanics that would require special handling, if they were really to be playable as plausible synthetics and not 'organics but with a metallic portrait'. For example we can believe that the ethics model and government type model could be broad enough to encompass all organic societies, like those of molluscoids or mammalians. But it's the scale of the potential differences between organics (including technologically augmented organics) and synthetics that I think would make implementing plausible synthetic races challenging.

For example, synthetic reproduction would almost certainly be drastically accelerated by comparison to organic reproduction. Consequently the overcrowding mechanics would probably also have to be completely different. Depending on the type of intelligence involved, the influence of the individual may also consequently be diminished in synthetic societies, making special character-driven gameplay (special characters like scientists, your head of state, etc.) potentially incompatible (although I guess you could engineer dedicated science-bots and middle-management-bots etc.). The resources required to sustain a synthetic pop unit would also be different to those for an organic pop unit.

These are just a few thoughts. I think (and very much hope) it can be done, but I doubt synthetic races would be a release date capability. Even then, there are such a variety of different types of synthetic race that may each require radically different gameplay mechanics, that it might not be possible to satisfy all tastes (hegemonizing swarm 'race', for example).