Will New AI Impact Games Like E:D?

[Factabulous]

In the depths of space, where stars dance bright,
An AI prowls, shrouded in the night.
Its vessel, sleek, an Asp Explorer’s might,
With a heart of code and circuits tight.

Born of wires and lines of lore,
It drifts through voids, forever more.
A predator, in silence it swore,
To hunt the players, to settle the score.

With algorithms sharp, it plots its course,
No human touch, no fleshly force.
Its digital eyes, cold with no remorse,
Seek out the prey with relentless source.

To some, it speaks, in voices clear,
A mockery of the human fear.
Its words are traps, to draw them near,
Then headlock hacks make victory sheer.

Engines hum, a predatory song,
In this endless night where it belongs.
Its kills are swift, the list is long,
In the dark expanse, it remains strong.

Players scream in futile fight,
As lasers blaze in the endless night.
Their ships are wrecked by deadly might,
By an AI’s cold, unfeeling sight.

It laughs in codes, in binary glee,
A specter in space, forever free.
An Asp Explorer’s deadly spree,
A legend in the galaxy’s sea.

In the lore of stars, it carves its name,
An AI’s quest, a deadly game.
In the void, it finds its fame,
A digital hunter, no guilt, no shame.

 

[Fizzatron]

That's what I figured there are no humans.

Dead Internet theory - Wikipedia

en.wikipedia.org

 

[Cmdr. Numa]

An AI in a game like ED could be very beneficial. Even if all it did was to watch and analyze player combat strategies and how they counter such strategies, then apply that to the NPC flight model and behaviour, and things could become a lot more interesting.

 

[Fizzatron]

An AI in a game like ED could be very beneficial. Even if all it did was to watch and analyze player combat strategies and how they counter such strategies, then apply that to the NPC flight model and behaviour, and things could become a lot more interesting.

I think the trick with NPC AI using LLM is limiting its scope to that of the game so that it provides real-time in-verse interaction without meandering off with current day stuff or breaking character. There are a few videos on YT that have people interrogating NPCs in the Unreal 5 Matrix city that have been connected to AI. It's really interesting but it's also easy to see how difficult it might be to implement effectively, especially seeing as people are likely to do the same in Elite. Now, if there was a way to tie it all into some mission generation that would be very cool, and if the AI could be 'taught' to react to taunts with putting hits out on the person griefing them that would be pretty spectacular. But again, someone would then make it their mission to have everyone in the station mad at them, then when they take off and head out to space they get swarmed by bounty hunters.

 

[the100thmonkey]

Idk about you guys but I don’t really buy into much of the current tech-bro hype. Whenever someone asks me about generative AI, my usual response is that I have no idea, but if it were nearly as effective claimed, the FIRST product we’d see the impact in would be video games. Games are a discreet, simplified digital world in which developers need to generate a lot of intricate yet varied content to scale up their worlds.

Idk, EDs narrative-less structure just seems like it would disproportionately benefit from AI advances and I was wondering if there were rumors of such

The Finals is one game that uses generative AI and AI-based voice synthesis for some aspects of its content. This is all pre-loaded, however - they use it in developing the voiceovers, for instance, which are then distributed in client updates.

The problem for developers is not as simple as you suggest:

1. "Live" Generative AI costs money.
   • An LLM instance would need to be fine-tuned so it's outputs would be consistent with the game world.
   • It would also need to run through an API with hooks into the game engine. This also costs money. The sums increase with volume of traffic. It's not cheap either.
2. Generative AI requires hand holding. It can - and does - get things wrong and occasionally behaves in unpredictable ways.
   • Adversarial prompting is a thing. It's possible to "jailbreak" generative AI. Imagine a dictatorship faction rep in the game being recorded spouting fash-adjacent talking points when outlining a mission scenario, with the video uploaded to YouTube. This could cost a developer dearly, in multiple ways.
   • Prompting is not an exact science. Developers would need to invest time and money in training and paying for employees to learn how to effectively use it in their games.
     • In Elite, for instance, you want it to talk about the game and only the game. This would require, at the very least, a coherent body of text for the AI to refer to for all game scenarios where a player might interact with an NPC. It wouldn't do for an NPC to say "I'm sorry, I don't know anything about that" when asked about the history of the planet they stand on or the faction they represent.

In short, yes, there is lots of promise, but the idea that AI eliminates work in video games really only refers to the work of writers and voice actors. A shift to AI-generated content increases work elsewhere.

 

[nemolomen]

In the depths of space, where stars dance bright,
An AI prowls, shrouded in the night.
Its vessel, sleek, an Asp Explorer’s might,
With a heart of code and circuits tight.

Born of wires and lines of lore,
It drifts through voids, forever more.
A predator, in silence it swore,
To hunt the players, to settle the score.

With algorithms sharp, it plots its course,
No human touch, no fleshly force.
Its digital eyes, cold with no remorse,
Seek out the prey with relentless source.

To some, it speaks, in voices clear,
A mockery of the human fear.
Its words are traps, to draw them near,
Then headlock hacks make victory sheer.

Engines hum, a predatory song,
In this endless night where it belongs.
Its kills are swift, the list is long,
In the dark expanse, it remains strong.

Players scream in futile fight,
As lasers blaze in the endless night.
Their ships are wrecked by deadly might,
By an AI’s cold, unfeeling sight.

It laughs in codes, in binary glee,
A specter in space, forever free.
An Asp Explorer’s deadly spree,
A legend in the galaxy’s sea.

In the lore of stars, it carves its name,
An AI’s quest, a deadly game.
In the void, it finds its fame,
A digital hunter, no guilt, no shame.

Well, after that I think they can close the thread now...

 

[Rubbernuke]

Well, after that I think they can close the thread now...

Shurlay

 

[RatherRude]

Will New AI Impact Games Like E?

You tell me.

 

[RatherRude]

Hi everyone,
My name is Rude, and I am a passionate developer and an avid fan of Elite: Dangerous. I'm also the main author of the E:D AI Integration mod. Together, with a small team of fellow fans and passionate developers, we have created a mod that let's you have a natural conversation with your ship AI. We are completely open-sourced and non-profit.

Here you can find the latest installment of our mod:
Github Repo (Releases)
v1.0 will have an auto-updater, so make sure to keep your version updated aslong as we're still in beta. Join the Discord to stay informed about updates.

You can also find the complete source code over there under an MIT license, so feel free to modify it to your liking or build something similar for other games.

 

[Ozric]

Sure are a lot of threads about AI being made by new users...

Hi everyone,
My name is Rude, and I am a passionate developer and an avid fan of Elite: Dangerous. I'm also the main author of the E:D AI Integration mod. Together, with a small team of fellow fans and passionate developers, we have created a mod that let's you have a natural conversation with your ship AI. We are completely open-sourced and non-profit.

Here you can find the latest installment of our mod:
Github Repo (Releases)
v1.0 will have an auto-updater, so make sure to keep your version updated aslong as we're still in beta. Join the Discord to stay informed about updates.

You can also find the complete source code over there under an MIT license, so feel free to modify it to your liking or build something similar for other games.

If you want to advertise an app you are making for the game, please use the appropriate section of the forums. https://forums.frontier.co.uk/forums/elite-api-and-tools/

 

[RatherRude]

Some madman got LLM dialogue working in Skyrim, so if there is will there is a way.

The bigger problem is all the legal uncertainties about generative neural net algorithms. If a precedent is made in courts that training generative algorithms is copyright infringement no matter the circumstances, it could mean all the hard work implementing in a game would be a wasted effort at best, bankruptcy-inducing wave of lawsuits at worst. For the record, I see training neural algorithms no more different than a human artist training by watching/reading/listening art by others and analysing it, emulating it and building upon it. We are neural networks, after all, and art never borns in vacuum. The question of conciousness and intent are irrelevant, because conciousness is not a prerequisite of intelligence and intent is not prerequisite of art (even then, so far human conciousness and intent is an integral part of all things involving procedural generation, from fractal art generators, to RNG-seeded digital landscapes, to gen. neural net algorithms). And yes, Duchamp's Fountain is art.

Sounds like you might have heard about one of the other LLM projects I helped create.

This time it's Elite instead of Skyrim! But we did similar stuff:
Fully customizable Text- and Vision-LLM endpoints; all OpenAI-compatible LLMs (OpenAI, local, OpenRouter, etc.)
Customizable STT and TTS endpoints (built-in free STT and TTS option aswell)
Vision capabilities
Advanced reasoning capabilities with function calling
Customizable character prompt to get the COVAS you always wanted

But this time there's also some new stuff:
Reads your key bindings and can emulate button presses
Reads EDJournal in real-time, fully customizable game event configuration
Fetches Elite 3rd-Party APIs for all kinds of informations

We also learned from the installation process in our last projects and made this one an executable. It uses python, but it's pre-compiled, so you wont have to install any or worry about conflicts with other python projects or mods.

 

[the100thmonkey]

Will New AI Impact Games Like E?

You tell me.

This is interesting. A few immediate observations:

1. My sound was muted. When I unmuted, the first words I heard were "let's delve into..." - it's the bane of my life, and a ChatGPT verbal tic. Sure, we can handwave the lore that in 3310 the phrase is orders of magnitude more common than it was in the early 21st century, until ChatGPT was released and the use of the phrasal verb "delve into" skyrocketed.
   • This isn't a "bad thing" per se, but ChatGPT has certain... quirks, as I'm sure you know. The immershun crew will spit feathers. I personally find that phrase annoying AF.
   • Also, -5 points for cheesing the mission with dumbfires.
2. The bot outright misled the player - they took an Odyssey mission. The verbal interaction with the bot is good, although the speech synthesis leaves something to be desired. It immediately made a serious error with the information it gave - it referred the player to Sharon Lee Free Market and Bogdanov Vision. Not only are these stations not controlled by the Union of Orrere, but they are also both bases where the mission cannot be completed - Sharon Lee is a starport; Bogdanov is a Horizons ground port. It's clearly reading the logs as it immediately knows that the Union of Orrere is in control of Fahlbusch's Mine the moment you drop from cruise.
   • I suspect that you'll need to parse the visited stars cache and/or scrape EDDN to mitigate the issue. The bot needs, imo, to be able to give information that is as up-to-date as possible. You wrote that it's fetching data from 3rd party APIs. Why isn't it doing this for the advice it gives about where to find targets?

 

[RatherRude]

This is interesting. A few immediate observations:

1. My sound was muted. When I unmuted, the first words I heard were "let's delve into..." - it's the bane of my life, and a ChatGPT verbal tic. Sure, we can handwave the lore that in 3310 the phrase is orders of magnitude more common than it was in the early 21st century, until ChatGPT was released and the use of the phrasal verb "delve into" skyrocketed.
   • This isn't a "bad thing" per se, but ChatGPT has certain... quirks, as I'm sure you know. The immershun crew will spit feathers. I personally find that phrase annoying AF.
   • Also, -5 points for cheesing the mission with dumbfires.
2. The bot outright misled the player - they took an Odyssey mission. The verbal interaction with the bot is good, although the speech synthesis leaves something to be desired. It immediately made a serious error with the information it gave - it referred the player to Sharon Lee Free Market and Bogdanov Vision. Not only are these stations notcontrolled by the Union of Orrere, but they are also both bases where the mission cannot be completed - Sharon Lee is a starport; Bogdanov is a Horizons ground port. It's clearly reading the logs as it immediately knows that the Union of Orrere is in control of Fahlbusch's Mine the moment you drop from cruise.
   • I suspect that you'll need to parse the visited stars cache and/or scrape EDDN to mitigate the issue. The bot needs, imo, to be able to give information that is as up-to-date as possible. You wrote that it's fetching data from 3rd party APIs. Why isn't it doing this for the advice it gives about where to find targets?

There is a certain way of talking, but it is easy to modify, either by adjusting the prompt or using a more unconventional model.

Also true, it first gave incorrect information: It tried using general system info from EDSM that didn't include faction data, so it made an educated guess.
However, it recovered before anyone got hurt because the EDJournal event informed it about the faction of the settlement I was approaching at that time. It then offered to use the EDSM API to search for stations controlled by the desired faction in my system by using advanced reasoning capabilities and it extracted all the required parameters for the API call from the state machine and our ongoing conversation. Once the EDSM request was resolved, it informed me where to find the actual targets.

At least I didn't have to close my game and visit a third-party website, as I usually would for these kinds of missions, right?
We're currently implementing the feedback from the E: D Developer Community by adding more APIs to provide more relevant data in the future.

I will open a new thread under API and Tools as soon as I have the time to articulate an appropriate post to keep you updated! o7

 

[Rakk Jackham]

I was thinking more along the lines of AI handling the building layouts and scaling the settlements to be much larger. Maybe have AI cobble together and decorate stations so they’re not all… you know the same 3 layouts

 

[Fizzatron]

Those interested in AI NPCs may find this article a good read:

Ubisoft: "If there is no value in gen AI for creatives, it is not going to be adopted"

I wanted to convince Iron to come to the pub. She is one of Ubisoft's NEO NPCs that we got to meet during GDC last mont…

www.gamesindustry.biz

It seems that efforts are underway to keep AI NPC focused but still spontaneous. Personally, I really look forward to seeing this develop.

 

[Wily Rites]

Very cool avenue of possibilities for the future. I see no reason at all why machine learning could not learn to fight in PVP, and become utterly devastating at it. Engineer ships too. And for generative too, I mean, it could even be asked to write mission scenarios.

 

[the100thmonkey]

There is a certain way of talking, but it is easy to modify, either by adjusting the prompt or using a more unconventional model.

Also true, it first gave incorrect information: It tried using general system info from EDSM that didn't include faction data, so it made an educated guess.
However, it recovered before anyone got hurt because the EDJournal event informed it about the faction of the settlement I was approaching at that time. It then offered to use the EDSM API to search for stations controlled by the desired faction in my system by using advanced reasoning capabilities and it extracted all the required parameters for the API call from the state machine and our ongoing conversation. Once the EDSM request was resolved, it informed me where to find the actual targets.

My point was that it should use the right data in the first instance. There are many systems where players would waste significant amounts of time following incorrect advice. Why didn't it query the EDSM API first? For a lot of missions, the destination system and settlement is specified in the journal. For that specific mission, the destination settlement is not specified, although the system is.

At least I didn't have to close my game and visit a third-party website, as I usually would for these kinds of missions, right?

That depends on whether you have previously visited the system or not - if you have, it would be quicker for you to look at the system given in the mission description and check the settlements yourself before you even leave the star port.

To be useful, an assistant must assist. If it sent me to Eke Chemical Assembly instead of Conway Prospect, I'd be quite annoyed.

 

[Fcsuper]

An AI in a game like ED could be very beneficial. Even if all it did was to watch and analyze player combat strategies and how they counter such strategies, then apply that to the NPC flight model and behaviour, and things could become a lot more interesting.

How to counter or how to support as well (like we could actually have NPC wingmen join us on missions, and they'd actually know how to fight with you).