I’m honestly a little confused by the “competition” part. I don’t think either @Carnel or I are trying to propose a “better idea” or find flaws faster than someone else.
We both actually said that we LIKE the idea of deeper attraction, preferences and relationship-aware behaviour. The only thing we are discussing is whether AI specifically is necessary to achieve it. To me that is just normal discussion of a suggestion, not a competition.
I also still don’t agree with the idea that because Smart Zoi already exists, the optimization question is basically solved and expanding its use would create no additional load.
Kjun himself recently talked about this. He said that Smart Zoi currently has bugs where it fails to respond properly as existing content expands, and that the responsible team is still discussing improvements:
In the very same post he also explains that even the normal autonomy system can become expensive when many Zois perform decision-making at the same time. Because of that they are optimizing the search order, limiting the search scope and reducing unnecessary calculations.
So I don’t think “the system already exists” automatically means using it for many more calculations has no performance cost. How often something runs and how many Zois use it still matters.
There is another interesting Kjun post regarding this as well:
Kjun specifically says that the LLM-based NPC storytelling system they were developing is not the current Smart Zoi. He also talks about improving things such as betrayal, jealousy, rumors and Zois recognizing relationships through multitasking, group behaviour and the event system.
So even KRAFTON themselves don’t seem to treat this as “Smart Zoi already exists, therefore everything social should use Smart Zoi.” They appear to use different systems depending on what they are trying to achieve.
That is why I also don’t really agree that traditional systems are simply an outdated step backwards. A good game can use AI where AI actually provides an advantage, normal simulation/game logic where that works better, and possibly combine both.
And “normal logic” doesn’t mean developers have to manually write thousands of simple “if this → then that” conditions either. There can be weighted probabilities, traits, moods, relationships, memories, context and randomness involved.
For example, the preference system we are talking about could still be completely unpredictable to the PLAYER because we may not know another Zoi’s preferences when we first meet them. We could discover those preferences gradually, similar to how we discover traits now.
I’m not against AI in inZOI at all. I just don’t think “AI exists in the game” automatically makes AI the best or cheapest solution for every complex behaviour.
And this is also why I mentioned earlier that if the technical implementation should be left to KRAFTON, as you said, then I think we can suggest the result we want — deeper attraction, first impressions, jealousy, hidden preferences and better relationship awareness — and let KRAFTON decide which combination of systems is actually best for achieving it. 