But the OP was lamenting the turn times as well. There's no free lunch, here. Maybe 5 years in the future it starts to become feasible to use networks for parts to help out with opponent behavior. And it definitely should be. I agree. But I also think that the strength of Civ is that the breadth of systems and the interaction of them between themselves and the player generate surprising/fun results. And I wish they'd focus more on that. Make the "meta" less deterministic and boring.Turns in Civ always take several seconds to complete, sometimes more than a minute in low end hardware. It never was limited by render time.
Yes, I'm talking about something trained in a neural network. Not just something programed by humans.
The AI in Civ games has always been very lacking. So with the advent of proper AI, we might finally get something decent.
Turns in Civ always take several seconds to complete, sometimes more than a minute in low end hardware. It never was limited by render time.
Yes, I'm talking about something trained in a neural network. Not just something programed by humans.
The AI in Civ games has always been very lacking. So with the advent of proper AI, we might finally get something decent.
Yes. Technically, the design and commercial model would have to be severely compromised to make neural net CPU players a priority. Let's assume that google alpha-whatever just decide for shits and giggles to train an AI on Civ 7 and share all their technical knowledge with Firaxis. Forget the game requiring a 4090 to process turns in under ten minutes, which is just speculation, one problem would be having to retrain and test the AI for every patch or microtransaction DLC. Then how big would the weights file for the AI be? Every patch you have to deliver a 100 gigabyte file to users? Would they be happy with that? OK, so you reduce the weights file to 4 gigabytes, then it starts acting more even more weird and has to be retested again. Seems like a lot of hassle compared to a big human-readable switch statement written in C.Are you expecting Civ 7 to still use just a bunch of if statements, not proper modern AI?
The thing is most people don't like to lose to AI, so in order to have good AI it also needs to be bad. Especially in strategy games, where the AI can calculate billions and billions of possible outcomes in seconds.
The AI in chess games peaked 20-30 years ago, and I don't think the AI in strategy games is any different. All examples of AI that learns and grows against you as you play it, like the AI mod for Dawn of War RTS, is super niche and definitely not what is going to sell your product imo.
I think the reason the game is being ported to the Switch is because they believe it represents a large number of sales. And that bodes well for the Steam Deck then too, so I'm happy. Then again I don't really feel like the series has become that much better since the original Civ.
Given games support across all platforms MS may ditch the Xbox Series S successor for good.I think everyone wants last gen to end, clearly it’s not an emotional thing I spoke with someone who still plays on 7th generation.
We don't need to have just one level on a trained AI.
We can have several difficulties from which the player an choose from.
And AI can make mistakes. Just look at how bad Google's Gemini AI is.
But even ChapGTP makes mistakes.
So I'm all for having an organic, AI that runs at different levels, but I don't think there's anything to that which would preclude a Switch, hell even a Gamecube. Stuff like Star Dock's Gal Civ 2 released 2006 and that hardware was more than sufficient for all kinds of AI, overkill even. You can go further back to like Master of Magic or Heroes of M&M2, and I don't really feel AI has significantly advanced in strategy games from those times, personally.
AI in vast majority of games is heavily CPU dependent for calculations. So it’s very much dependent on hardware."Good" AI is a software and player satisfaction problem, not hardware.
That hardware doesn't have support for accelerating matrix multiplication, add and fusion.
And yet, the games from 20 years ago still can have AI that is written to achieve higher difficulty than what we have today. I mean I understand you can't have lots of things in regards to AI on older hardware, but I just fail to see how that does anything useful in terms of this game. I've played countless strategy games over the decades and AI is one thing that really never evolved, like graphics has. In many ways, I feel AI has gotten worse in newer games vs older ones.
You are misconstruing AI written by hand, by a human, with matrix generated AI.
I think I simply don't care as a player. Why would I play this game over a more enjoyable older one that costs a fraction of the price? Having matrix generated AI doesn't matter to me, if a new Heroes of Might and Magic game released with it, that doesn't really mean it would be better than 2, 3 or 5, which didn't have it. I guess I have yet to be sold on the benefits of matrix generated AI when difficulty has never really been the limiting or deciding factor for me in a TBS or 4X game. Artwork, game balance, research, and exploration all play bigger roles. If anything you could argue a hand written AI can be implemented to provide specific challenges and the need to overcome them in a way a matrix generated one may not.
I guess what is lacking is a clear example of a TBS game which is an actual gem - that is beyond fun to play, super engaging, and is something that is not possible or ever been done before due to the AI. I guess it just doesn't even come in to my radar that being on the Switch is a reason to be disappointed in a game.
Then why are you doing in the Civ VII thread?
Clearly I disagree with the premise that a TBS / 4X game existing on the Switch is crap. And I have yet to see an argument of actual evidence to convince me otherwise.