I admit, this would be a huge overhaul, and maybe it would be a better goal for a Planet Zoo 2, but it would be so helpful.
What if an AI neural network was used to calculate how guests or animals navigated around/interacted with different objects? The most perfect application of this would be for calculating traversable areas that actually make sense, but another could be for detecting things like custom bridges (or even just custom paths) for guests to use without actually needing to use paths. Of course, this could potentially get out of control really fast, so maybe the AI won't try to detect anything in a given area unless a player tells it to.
Of course, there's the concern that the AI could start making things happen that are totally incorrect, but here's the best part: AI neural networks work off of training data. Before the release of the feature, users can submit an endless amount of designs to be used as training data, and after the release, if a user runs into an instance of the AI making a mistake, they can submit it to Frontier, Frontier tells the AI how the scene is supposed to be handled, and voila! Both parties benefit: the user gets their scene fixed, and Frontier gets more training data that will improve the algorithm.
If anyone watches the YouTube channel Two Minute Papers, they will see AI neural networks doing stuff easily as complicated as this and more. This is what AI neural networks are perfect for. Additionally, this offers a perfect design solution to something that has been consistently bugging me about Planet Zoo: players want more and more things to work, but the only feasible solution has often been that Frontier just gives them the ability to flat out override the system. Vista Points are a perfect example; rather than producing a better algorithm for calculating where guests would look at a habitat from, Frontier just gave users the ability to more or less directly interact with the underlying system, which gives the players more power, but also kind of feels like moving in the direction of making the player one of the programmers just making their own thing, rather than a customer enjoying a finished product made by a separate company; in other words, having players get their hands dirty when that's not their job. Using an AI neural network would be a perfect solution to this, making players just the designers who lay out their ideas, and the game makes it all actually work.
Again, I'm sure it would be a huge overhaul, but I feel having users submit training data on an ongoing business would simplify a lot of concerns.
What if an AI neural network was used to calculate how guests or animals navigated around/interacted with different objects? The most perfect application of this would be for calculating traversable areas that actually make sense, but another could be for detecting things like custom bridges (or even just custom paths) for guests to use without actually needing to use paths. Of course, this could potentially get out of control really fast, so maybe the AI won't try to detect anything in a given area unless a player tells it to.
Of course, there's the concern that the AI could start making things happen that are totally incorrect, but here's the best part: AI neural networks work off of training data. Before the release of the feature, users can submit an endless amount of designs to be used as training data, and after the release, if a user runs into an instance of the AI making a mistake, they can submit it to Frontier, Frontier tells the AI how the scene is supposed to be handled, and voila! Both parties benefit: the user gets their scene fixed, and Frontier gets more training data that will improve the algorithm.
If anyone watches the YouTube channel Two Minute Papers, they will see AI neural networks doing stuff easily as complicated as this and more. This is what AI neural networks are perfect for. Additionally, this offers a perfect design solution to something that has been consistently bugging me about Planet Zoo: players want more and more things to work, but the only feasible solution has often been that Frontier just gives them the ability to flat out override the system. Vista Points are a perfect example; rather than producing a better algorithm for calculating where guests would look at a habitat from, Frontier just gave users the ability to more or less directly interact with the underlying system, which gives the players more power, but also kind of feels like moving in the direction of making the player one of the programmers just making their own thing, rather than a customer enjoying a finished product made by a separate company; in other words, having players get their hands dirty when that's not their job. Using an AI neural network would be a perfect solution to this, making players just the designers who lay out their ideas, and the game makes it all actually work.
Again, I'm sure it would be a huge overhaul, but I feel having users submit training data on an ongoing business would simplify a lot of concerns.
