Realistic dialogue

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

@Itookausername

They’re here for the hyper-realistic simulation of biology and complex evolution that you can’t find in any other game.

As long as the realism doesn’t get directly in the way of creating a somewhat fun experience.
I wonder if anyone still remembers the “ThriveSins” list…

[Maxamax_v]

@aah31415

I wonder if anyone still remembers the “ThriveSins” list…

I would like to see this. I joined the forum less than a week ago, so I would not remember this

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

Here it is. The thread was created 2.5 years ago…

ThriveSins

Hello all, A fun little idea for a forum thread. Thrive occasionally strays away from realism a bit, more often than not due to decisions favoring fun and fair gameplay. Here, we can point out some of these inaccuracies in a harmless way, similar to the concept of CinemaSins. The goal of this thread isn’t really to suggest changes (perhaps some minor alterations or ideas can come from this), but just to point out where Thrive strays from its otherwise incredibly realistic atmosphere. So save yo…

[Maxamax_v]

Thank you I will check it out

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

I do wonder how will the list grow with the present mode of development…

[Deathwake (i nuked zenzone and will never let him forget it)]

@hhyyrylainen

I think most people will still be rocking laptops that are being sold currently in 5 years.

Yeah a modern laptop can easily run a low end AI model. The ones with accelerators obvious have the edge that they can do it while gaming and run the models with quite a few parameters, but if you don’t mind flooding your vram, a smaller model will in fact run on anything with dedicated graphics, and will “run” on literally anything, but it’s not worth it unless you have enough vram to fit the model. If we limited it to very minor rewrites, a very small model could run while gaming. You would probably have to turn down the settings to free up vram on most laptops, but any gaming laptop can handle it already. Not, like, steam deck, or my laptop, but yk, in 4 years, any old desktop, and any gaming laptop even one from today. Most people aren’t playing on pre-covid laptops, and those who are have like, a 980 TI, not a 960. And they don’t expect to have raytracing or DLSS.

@hhyyrylainen

I decided I should try the current AI offerings to have an informed opinion on them. [and the rest of the paragraph blah blah]

I meant more AI has no use in content generation than helping you autocomplete code lines. I will admit a good autocomplete can be really nice, especially when you have irritating naming schemes and loads of keywords. I’m sure the AI one’s are real nice.

@Itookausername

Well, yes. I don’t think dev time should be taken up by working on LLM integration, as that isn’t really what people got into thrive for.

YEAH!

@Itookausername

They’re here for the hyper-realistic simulation of biology and complex evolution that you can’t find in any other game.

And the simulation of the kinds of societies and governments such biology results in, don’t forget. If you play an rts game, the dialogue doesn’t matter, but the personality of the AI should make sense. In games like civ, the AI does talk to you, if we were to go toward that route, or toward, say, Spore, you do need the AI to say something. This thread is for the AI to say something they we like, and how to make that happen. If you think dialogue doesn’t matter, then this thread as a whole is kinda against that lol.

[Maxamax_v]

@aah31415

@Itookausername

I think the best way this could work is through community made AI npc mods, akin to those of skyrim and other games.

You mean the community would have a contest for the best “dialogue AI mod” and the best of such mods would then become a part of the official game?

I don’t know how i missed this idea when looking at this topic. I absolutely love this idea!

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

Should there be some cues if, say, your species communicates via other means than sound (like movement)?

[Itookausername]

Maybe look into deepseek-v3?
It’s a new open source chinese made llm that I’ve heard achieves similar or better results than the openai models, might be a promising programming aid.

[hhyyrylainen]

Doesn’t it need an absolute ton of VRAM to run? I looked at it and only like the smallest 3 variants of the model would fit in my GPUs memory. So I kind of doubt it would beat a cloud model that runs on actually beefy workstation class GPUs (that cost tens of thousands).

Also if it doesn’t integrate with my IDE I wouldn’t really use it as today I didn’t ask the AI anything but used the AI auto-complete like 20 or so times (which is almost comparable to the number of “normal” IDE autocompletes I used today).

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

For the time to come, finding new developers will continue to be more important than finding better programmer aid AIs. Well, unless something huge happens with the AIs, but I doubt that’d occur.

[Itookausername]

You could try using their website and api instead of hosting it locally

Granted, you can’t integrate that to the IDE but it might be useful for actually assisting and solving problems (rather than a really good auto complete)

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

@Itookausername

for actually assisting and solving problems (rather than a really good auto complete)

Did the AIs actually get good at this and don’t have a very bloated/“clickbaitized” reputation?

[Itookausername]

Well on ChatGPT GPT-4o model I used a specialized gpt that was fed the custom programming language and documentation for a game I was modding and it was actually fairly useful (though I still needed to correct it at times)

[hhyyrylainen]

@Itookausername

Granted, you can’t integrate that to the IDE but it might be useful for actually assisting and solving problems (rather than a really good auto complete)

This is the one thing I don’t think an AI can help with, no matter how good it is, because if I ask for some code it will obviously need to work in Thrive. So I would need to first remember all relevant files and attach those to the request and then the tool would need to generate within like the first two attempts what I want, because otherwise it is way more hassle than just writing the code in the first place. The IDE integration has a really big advantage here in terms of how much hassle it is to try to use it.

For example today, I think I wrote like a hundred lines of code, but those were split into like 20 different files and all of those individual edits were very simple as I know the context and what available methods Thrive has. The AI wouldn’t be able to get any of that done. And even if it could, it would take longer to feed the AI all the relevant files and details, and then copy-pasting the results to the right places in each file.

[Itookausername]

I think you’d have to feed an AI model with all the relevant information once, Which you can do with the custom GPT system on the openai website, you can upload a limited amount of files for it receive info from (I think higher tiers get way more file size?) and it will try to use the information it has been fed from the instructions and uploaded files.

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

I’ve been rather vary of these “AI coding tools” due to how they have been doing in non-staged coding environments (as I’ve seen from the video below, albeit it was published in september of 2023, so it can be expected that certain improvements were made since then).

Developing a game with ChatGPT 4

(spoiler, it doesn’t go well)

[Itookausername]

Tbf thats gpt-4, which is an outdated openai model. Today GPT-4o is the standard model and its much better performing.

[aah31415 (The maker of SitF, Radiostrocity, The Lifenote and TGBing; The Second Ascended...; And just maybe a security warning come alive...?)]

What happened to “GPT-5” by the way?

[Itookausername]

Its more and more expensive to train LLMs at higher parameters, so the gpt models are suffering from diminishing returns. GPT-5 wont be out for a while