Could you elaborate more? From my very amateur point of view, fishing seems like something that could be completely automated pretty easily, but it seems like I'm not seeing the complexity.
A large part of it is liability. In theory you could just send a self-driving boat out with a giant net to scoop up a few tons of marine life to bring back to shore and stuff in a processing plant. Or even better maybe make a self-sufficient platform that uses current to turn turbines and just scoops up fish, sorts them, and puts them on ice for cargo carriers to come by and pick up.
The problem is the same one as self driving cars on a more massive scale - accountability. When you have human bodies at every step of the process you have someone to blame if you "accidentally" kill a protected species, overrun your quotas, put tainted meat in the market, etc - if everything is operating on its own you get the same situation where self driving car operators are going to try to blame the software engineers if something goes wrong, while the company gets to take the fallout.
And the fines on illegal fishing are actually legitimate compared to the token paltry fines in other industries (depending on the country, of course).
The economics are completely on automations side here, you could get 97% good fish for a tenth the price with some substantial up front investments today. But that 3% false positive / sorting error case with the inability to blame line workers kills the business.
The ocean is a dynamic environment. Nets get tangled, lines get tangled, wildlife interferes, weather and sea conditions are harsh and unpredictable. The ocean isn't a nice flat square corn field that a combine can lazily navigate through. We've just barely figured out how to automate picking strawberries in a nice safe greenhouse.