There's an entire range of depictions when it comes to AI in space-travel themed science-fiction movies and TV shows. Hal, for example, is a conscious AI - for some definition of consciousness. While this is a wonderful plot device and something tech visionaries have been saying is right around the corner, it's not how I view AI at all.
I think the show Expanse got it just right. All the ships and stations are loaded with AIs but you'd never even notice them. It's subtle. The AI does the job and gets out of the way. Take the example of intercepting missiles with PDCs or firing a rail gun. It's not a simple or trivial process (and the show goes to great lengths to make that clear as well). There are a huge variety of parameters to consider to keep the ship stable when it's firing its rail gun. This is space after all and Newton's third law can be a pita. Counteracting forces - via thrusters - are required in extreme precision to keep the ship from spinning or moving in any of the 3D space.
Similarly, the countless other things to monitor in a ship and adjust automatically is the only way space travel would be possible with a reasonable crew. A very minor angle shift could mean the difference between reaching two entirely different planets and nothing's ever stationery in space.
I also found it cool the way the crew interacts with the ship is depicted - a mix of voice commands and gestures and an AI that can interpret what they mean. At least the way it’s depicted, they can also feed it "high level programming" - like Naomi creating a firing solution.
My point is that none of this requires an AGI - Artificial General Intelligence - "just" an AI would do. It’s in by no means a ding. If we could replicate what evolution has taught animals to do, that in itself would be nothing short of a miracle. If you think about it, a hummingbird with the brain the size of a pea, can flap its wings 53 times a second, maintain near perfect stillness, see things, hear things, dodge obstacles, seek out food and recognise what is good and what is not, identify danger, identify a mate and on and on. And the whole thing operates on a few grams of sugar.
A hummingbird is not what we typically think of when we think intelligence or consciousness. However, it's precisely this type of intelligence that's going to be useful for humans and I don't see any reason why we'd first solve AGI before building these AIs.
This is happening already. Alexa/Siri/Google/Cortana on the voice side, Boston dynamics on the whole slew of graceful movements, robots that replicate human emotions, the increasing autonomous features being built to see the world around us, …etc. I don’t believe the endgame will be a skynet, but rather a whole slew specialised AIs that can do one or more things autonomously by reacting to impulses rather than through a chain of thoughtful decisions.