CHATPGT-4 in stellaris

[Stellaris enjoyer]

Is anyaone can integrate Chatgpt-4 in stellaris. For example you can communicate with other country as you want and say them what you want. Their answer will be based on the principles of the state.

[LuvianQ]

Oh, yeah. That's another one. Things are moving pretty fast in this space

[Lambv]

Αναρτήθηκε αρχικά από Immortalis:

It won't work. At least not with this level of technology.

Chatgpt is not a lite-program of a few MBs, it's a massive database that relies on an equally massive infrastructure (let alone the discussion about having it in multiple languages). You'd have to create a mod / DLC / update (call it however you want) of hundreds of GBs to accompany a game measuring in the low double digit (right now my folder is barely above the 16 GBs mark).

And even if you manage something like that, it would slow down the game so much that it will turn into a slogfest as soon as you meet an alien empire. And the more the game goes on the more interactions the system would need to record.

And that's leaving aside the entire argument about "How would the chat influence the other empire". For example, if I tell them "I won't attack you" do you want that sentence to translate into what? A non aggression pact? A simple promise that I'm free to break? Would the empire trust me and leave the border undefendend or dismantle a few ships...


Honestly, it's not a "just put the square in the appropriately shaped hole and call it a day" approach

I think you're wrong on this one. The technology is evolving so rapidly, that many people are predicting that you will have something of gpt-4 equivalence running on your PHONE by 2025, and if you look at the speed of optimization, it's not that unbelievable.

Wizard-Vicuna, 13b could probably be put into stellaris and add a very convincing chat system in game for relatively cheap resources... *its 15 ish gigs, and runs on a 12 gb vram card* (still will require either a dedicated processor, or a decent gpu running solely for the AI) but for a game like stellaris... I mean, It hardly even utilizes my 3080.

If things continue the way they are, you will probably see chatgpt type ai, in games, with customized character profiles by 2025-6.

[corisai]

Αναρτήθηκε αρχικά από Lambv:

The technology is evolving so rapidly

We promised with fusion reactors near for a century. Rapidly. Yean.

Αναρτήθηκε αρχικά από Lambv:

that many people are predicting that you will have something of gpt-4 equivalence running on your PHONE by 2025

It could "run" even here and now - in form of client software :)

Sorry, but such "predictions" are just big fat lies. CPU are already no logner doubling their performance each 2 years - so called "Moore's law" is broken.

Αναρτήθηκε αρχικά από Lambv:

and if you look at the speed of optimization

Actually "speed of optimization" is negative one. Our current phones are far more powerful that central computers of interplantary probes and current software waste even more resources. One of the best examples of resource hog is Microsoft Office - with more or less same functionality each new iteration consume more and more resources.

Αναρτήθηκε αρχικά από Lambv:

If things continue the way they are, you will probably see chatgpt type ai

It isn't AI and more likely a direct opposite of it. It's just a glorified chat bot that writting responses using a semi-random method (so we even unable use to it make interactive Wiki as it don't ashamed for "imagine" things).

[rafael.ramus]

Αναρτήθηκε αρχικά από corisai:

Αναρτήθηκε αρχικά από Lambv:

If things continue the way they are, you will probably see chatgpt type ai

It isn't AI and more likely a direct opposite of it. It's just a glorified chat bot that writting responses using a semi-random method (so we even unable use to it make interactive Wiki as it don't ashamed for "imagine" things).

Interestingly enough, that was my one and only conversation with Chat GPT: are you AI? The answer was negative (it even babbled about the links between intelligence and consciousness as I pushed it a little. It was adamant it itself is no true AI).

[corisai]

Αναρτήθηκε αρχικά από rafael.ramus:

It was adamant it itself is no true AI.

Hard codded response :)

And easily to check - after "eating" a whole Wiki it still unable to solve properly complex yet basic math (I mean any prolonged combination of four basic operations). It guess fine on small values but usually go off on anything bigger then 9-12 digits.

I mean - c'mon, proper AI will be capable to learn at least basic math, isn't it?

[MP]

Because it is not AI. It is a LLM - Large Language Model. It is a neural network "trained" on a huge amount of data.

It doesn't think - it just predicts output based on input. It is little more advanced than a circuit made of logic gates; there is no thinking at all - it is plain simple input - output, just on a greater scale with billions of parameters and probability distribution.

It doesn't do any actual thinking nor deduction - it just predicts responses based on prompts. In most cases, it gets predictable input and produces expected output. Works well for established knowledge.

But human is not human if he doesn't do stupid stuff, like asking unpredictable prompts (read: stupid questions) and post predictably erroneous output for internet points.

[corisai]

Αναρτήθηκε αρχικά από MP:

In most cases, it gets predictable input and produces expected output. Works well for established knowledge.

And even there we have a problem - there is not exist a "base of established knowledge".
Even Wikipedia contain a few errors + have a lot of political biases (and their spread is irregular among topics).
Any kind of additional censorship is only making it worse.

[Stormwind]

Αναρτήθηκε αρχικά από corisai:

Αναρτήθηκε αρχικά από MP:

In most cases, it gets predictable input and produces expected output. Works well for established knowledge.

And even there we have a problem - there is not exist a "base of established knowledge".
Even Wikipedia contain a few errors + have a lot of political biases (and their spread is irregular among topics).
Any kind of additional censorship is only making it worse.

Αναρτήθηκε αρχικά από Simpson3k:

Αναρτήθηκε αρχικά από Immortalis:

It won't work. At least not with this level of technology.

Chatgpt is not a lite-program of a few MBs, it's a massive database that relies on an equally massive infrastructure (let alone the discussion about having it in multiple languages). You'd have to create a mod / DLC / update (call it however you want) of hundreds of GBs to accompany a game measuring in the low double digit (right now my folder is barely above the 16 GBs mark).

And even if you manage something like that, it would slow down the game so much that it will turn into a slogfest as soon as you meet an alien empire. And the more the game goes on the more interactions the system would need to record.

And that's leaving aside the entire argument about "How would the chat influence the other empire". For example, if I tell them "I won't attack you" do you want that sentence to translate into what? A non aggression pact? A simple promise that I'm free to break? Would the empire trust me and leave the border undefendend or dismantle a few ships...


Honestly, it's not a "just put the square in the appropriately shaped hole and call it a day" approach

Well a Stellaris chat gpt does not neet the whole information the one accessable on OpenAI uses. It only needs Stellaris lore from each and every content in the games code so that a conversations can be made about it. No leader in stellaris needs to talk about the trade value per year of south carolina or the reason for kidney stones.

Galciv IV will be equipped with a version of chatgpt to create your own empire just from a promt so i think alot is possible with that feature.

I don't think its that easy. LLM needs a huge training model to work from. You arent going to get it from a few pages of text, if you could, people would have done chatgpt 30 years ago.

[Stormwind]

Αναρτήθηκε αρχικά από -=EDF=- Kouta Seto:

Laugh now or later? Those AI stuff and programs are so inferior to humans it actualy HURTS! I thought we would be FAR MORE ADVANCED as a society... yea i was wrong hard time...

You want that? Then as mod. Nothing else. I dont want that ♥♥♥♥ called AI in my games safe the one that controlls the empires etc. aka THE BASE GAME ENGINE!

Also what @Immortalis said. The game is simply not designed for that NOR will it ever be. And seriously? Thats a good thing. If AI deems it logical to basicaly SPREAD FAKE NEW'S i dont want that to actualy EXIST!

I understand your concerns about the implementation of AI in games like Stellaris. You're right that there's no substitute for human creativity, intuition, and emotion, and AI is not without its limitations. As it stands, AI like GPT-4 or ChatGPT can generate text based on its training data, but it doesn't have the ability to understand context or nuance in the same way a human does.

Regarding the fear of AI spreading 'fake news', it's crucial to clarify that AI doesn't have any beliefs or intentions; it generates responses based on its programming and training data. Any false information it might produce is not out of a desire to deceive, but simply a limitation of its ability to discern fact from fiction.

When it comes to gaming, AI can be an effective tool for enhancing certain aspects of the game, like creating varied and dynamic NPC interactions. However, its implementation should always consider the game's design and player experience. Ultimately, the decision to use mods or AI enhancements should be left to individual players. Everyone has different tastes and preferences when it comes to gaming, and that diversity is part of what makes the gaming community so vibrant and exciting.

[corisai]

Αναρτήθηκε αρχικά από Stormwind:

Regarding the fear of AI spreading 'fake news' ...

We should stay strong and admit that this is just a censorship and nothing else.

Αναρτήθηκε αρχικά από Stormwind:

When it comes to gaming, AI can be an effective tool for enhancing certain aspects of the game, like creating varied and dynamic NPC interactions.

If we're speaking about modern finished AI systems - in same time such interactions will be fruitless or boring like countless randomly generated quests "kill X rats".

Making LLM capable to introduce player to new quests and parts of story during casual banter will require a lot of work from devs (and at least part of this work couldn't be bypassed by using a ready product).

[Stormwind]

Αναρτήθηκε αρχικά από corisai:

Αναρτήθηκε αρχικά από Stormwind:

Regarding the fear of AI spreading 'fake news' ...

We should stay strong and admit that this is just a censorship and nothing else.

Αναρτήθηκε αρχικά από Stormwind:

When it comes to gaming, AI can be an effective tool for enhancing certain aspects of the game, like creating varied and dynamic NPC interactions.

If we're speaking about modern finished AI systems - in same time such interactions will be fruitless or boring like countless randomly generated quests "kill X rats".

Making LLM capable to introduce player to new quests and parts of story during casual banter will require a lot of work from devs (and at least part of this work couldn't be bypassed by using a ready product).

You've raised a great point, and I completely understand your perspective. A fundamental challenge of integrating AI into a game like Stellaris is ensuring that it doesn't just churn out repetitive or uninspiring content. Here's how ChatGPT, or a similar language model, might be employed while avoiding these pitfalls:

1. Dynamic and Responsive Dialogue: ChatGPT can generate dialogue that adapts to players' input. This means the NPC interactions could be more varied and immersive than simple "kill X rats" quests. For instance, it could introduce complex moral choices or decision-making quests based on players' past actions or stated preferences.

2. Enhancing Storytelling: The AI could be employed to enrich the lore and backstory of the game universe. It could provide interesting titbits, tales, or mythologies that are unique to each playthrough.

3. Procedurally Generated Content: ChatGPT could also be used to create procedurally generated storylines or quests, which would vary greatly from game to game. This wouldn't replace the handcrafted quests or storyline but would supplement them to provide a richer, more diverse gaming experience.

4. Learning and Improving: While current AI models, like GPT-4, don't 'learn' from new data after training, future models could potentially adapt to player feedback over time, allowing the game to improve and evolve.

Indeed, integrating AI would require significant work from the developers. It's not just about plugging in the AI and letting it loose, but meticulously crafting guidelines, rules, and boundaries for the AI to operate within to maintain the game's core essence. It's also important to thoroughly test and refine the system to ensure a smooth gaming experience.

At the end of the day, it's all about enhancing the player experience. If AI can contribute to that in a meaningful and enjoyable way, then it's worth considering. If not, the traditional methods of game development continue to provide fantastic gaming experiences.

[Maya-Neko]

Αναρτήθηκε αρχικά από Immortalis:

Let's hear it. I'd also like a list of software companies that not only have their product linked to a third party server they have absolutely no control or oversight over but specifically make it an integral part of that product.

never heard about things like Denuvo, VAC or Easy Anti-Cheat? And then we've also Clients like Steam or Epic, where the devs rely on these companies ability to provide the informations needed to get people into multiplayer. Or we can also take a look into engines like Unity or Unreal, which is used by many developers. And in terms of infrastructure, many devs already go with google or amazon cloud servers, since it's cheaper than building your own server rooms and keeping it up.

[Sputnik]

While I certainly expect humanity to be finally snuffed out by AI, I still look forward to having a little fun with it along the way.

1. Personality algorithms are very likely looming as the greatest revolution seen be the games for the games industry (Yeah yeah, they said First Person Shooters were a pipe dream when I was kid).
   
2. Most likely they'll be streamed from a server, rather than sitting on your hard drive.
   
3. In multiplayer games you'll never quite be certain just who is real, and who is a NPC.
   
4. In open world RPGs, the procedural generation of quests will be possible as NPC enlist players to serve their interests, With the changing fortunes of NPCs, so too will the politics inside your game world change. RPG worlds will finally "come alive".
   
5. Heck, maybe even the Total War battle AI will move its troops out of my artillery fire!

Just someone tell me when I can get the USB port into my neck and these Meta Quest Pro goggles are straight in the bin. Then it's just a simple matter of an online ordering for urinary catheters and we can all make way for the AI Revolution!

[ScreamCon]

Αναρτήθηκε αρχικά από Checkered:

Is anyaone can integrate Chatgpt-4 in stellaris. For example you can communicate with other country as you want and say them what you want. Their answer will be based on the principles of the state.

what... really are that many people addicted to the software?
Even if you could incorporate the feature would only work if you had internet access, because it would be communicating in the back end. And even then the developers would have to flavor the software for each empire you communicate with.

[Twelve-Oh-One Alarm]

Αναρτήθηκε αρχικά από ScreamCon:

Αναρτήθηκε αρχικά από Checkered:

Is anyaone can integrate Chatgpt-4 in stellaris. For example you can communicate with other country as you want and say them what you want. Their answer will be based on the principles of the state.

what... really are that many people addicted to the software?
Even if you could incorporate the feature would only work if you had internet access, because it would be communicating in the back end. And even then the developers would have to flavor the software for each empire you communicate with.

If you're going to join the discussion, at least know how the ChatGPT APIs work