I was looking through some of @jeffclune's tweets about creative ways AIs have solved problems and was reminded of this book. Yukikaze is a fighter plane AI that gradually gains sentience and independence, but does so while remaining thoroughly alien. Whereas her human pilot is very attached to her, there's no indication that she is in any way attached to her human pilot, or even perceives him as anything other than one more variable in her missions.
Then there's @JanelleCShane's tweets. This one, I think, is particularly relevant to Yukikaze: "If 'kill all humans' is the easiest solution to a problem, then machine learning will do that unless prevented." (Tied in with this tweet, which is a little like a scene that actually happens in Yukikaze: "Another algorithm discovered that rather than minimizing force, it could apply such a huge force that it overflowed the simulator's memory and registered as zero instead. Of course, the pilot would die, but hey. Perfect score.")
I had been planning to give Good Luck, Yukikaze, the second book, another go starting today. It might be best to work on my interlibary loan books, though. I haven't decided yet.