It's a common mistake to assume that gun buybacks are being proposed as a solution. The solutions being proposed are a set of laws/policies to tighten gun controls, like who's allowed to buy guns, what guns are allowed to be owned and how many, improving checks and mitigating newer loopholes.
Tighter gun controls are shown to reduce mass shootings. In Australia, the laws have loosened a lot since the big wave of gun laws in 1996. The buyback program is a consequence of bringing people in line with the new laws.
The realistic goal is not to make it absolutely impossible for a motivated extremist with lots of resources to plan and commit a mass shooting, it's to make it much harder to prepare to do and to create more opportunities to notice their preparation.
Right, but in the same spirit, we're not just talking about LLMs. If we're being accurate, the common interface we're used to with ChatGPT or Gemini, they are a system of different models including LLMs and other models for images, or sound.
If we're talking AI in movies and music, besides LLMs for writing, we're mainly concerned with diffusion models
If we're talking AI for wearables... That's usually more on the sensor/classification side of ML, so it's not even generative.