While we probably agree to not let computers and #AI take over geopolitics and play #WarGames, politicians are too likely to use #LLMs as shadow advisors.
Large Language Models are no reasoning machines, and while the simulation setup is rather complex, real-world complexity far exceeds the simulation.
Much more text is written on escalation than de-escalalation, creating a potential explanation for the models' escalation tendencies.
Add comment