The single best “use case” of LLM’s is mass surveillance and manipulation on a scale like we’ve never seen before. A personalized AI agent that sees and hears 100% of what you do and “helps” you throughout the day is the perfect way to control your every thought and action.

Open source may help a little, but think about how much big tech and state actors will want to influence those models. Either by injecting malicious code via supply chain attack, or poisoning the data sets used to train them.

Beyond that, I personally think LLMs are mediocre productivity tools.