Rare had an AMA for SoT and here are some of the replies that they gave.
"We’ve got big ambitions about our AI and we definitely want to have other living things in the game other than just skeletons & sharks. I know there are a number of people here who would love us to spend some time on animals for the world. I mean, Fable had chickens, right?"
"The current skeleton AI are just set as a base level so they provide somewhat of a challenge but don’t annoy players when exploring too much. As we add different quest types, progression etc, there will definitely be different challenges to take on"
"In terms of the performance of the game, updating physics objects is probably one of the most expensive things we do (the ship is one great big physics object containing lots of other little ones!). We did experiment with making the chest more realistic (and having it fall down cliffs and bounce around) but it’s a trade off between spending our precious CPU cycles on things like the chest, or allowing us to support more ships or other exciting things. We are always looking at the performance side of things, so we’ll never rule anything out though!"
"Well, the spy glass is something we’d definitely love to have in the game, as it’s such an iconic pirate thing to have!
The spy glass is actually something of a technical challenge for us as it breaks our island streaming code, which currently works by only loading in islands that you are close to. This means that we only load in high detailed textures and models when we are really close to an island – and as you are inevitably sailing towards an island, we’ve got time to load it in by the time you get there. If you can whip out a telescope and see islands that are further away, it means we need to have more of these high-resolution assets loaded, which means we use up more memory (and we currently use a lot of memory). So, it’s something we really want to do, but before we get to it we first need to make sure we’ve got our memory under control, and we give our island streaming code a bit of love and attention (which we are at the moment). So nothing to report right now, but hopefully we’ll have news soon."
"Well, we’d definitely love to add much more damage to the game. Our Unity prototype tested out different forms damage (like the mast coming down) but we haven’t had a chance to prototype those things in Unreal yet. The actual damage tends to be the easy bit, but we need to make sure that the players can repair the ship afterward that brings the challenge. We don’t want to do any permanent damage to things like cannons until we have a method of refreshing your ship without sinking it first!
Obviously, as we start looking into more sea based AI, we may well need a more complex damage model."
Full article: http://wccftech.com/rare-big-ambitions-ai-sea-of-thieves/