I did not start my PhD with the topic I eventually ended it with. I originally started it with the goal of creating some sort of AI agent that is able to act diplomatically in strategy games.
I was hoping I could push for some approach that tries to trick, cheat, lie, sometimes tell the truth, but with the goal of manipulating other players into helping out the AI unwittingly.
The universe had other things to say, so I pivoted to what I found to be more appealing at the time: genetic algorithms in balancing games. And that proved to be a prophetic choice.
A few years later LLMs came out and they’re waaaaay better at lying and tricking than anything I could ever come up with.
