Really enjoyed this story, "Zeta-Epsilon" by Isabel J. Kim. It's about a ship AI + human cyborg, and I particularly liked how the backstory of the AI's development draws from the sorts of issues that tend to occur with current, real-life forms of machine learning:
"The function was: get from here to there. How to get from here to there? Blow up the ship here and its component molecules will end up there. The function was: get from here to there, but do not destroy the ship. How can the ship pass through a minefield when the fuel will run out if the ship goes around? It’s possible: burst fuel in a single acceleration and let the ship drift for two decades. The function was: get from here to there, do not destroy the ship, and bring the crew back alive. How to bring the crew across solar systems and back with net-zero loss of life? Replace any dead crew with new crewmembers taken from the enemy during skirmishes."
