The Trump administration recently published “America’s AI Action Plan”. One of the first policy actions from the document is to eliminate references to misinformation, diversity, equity, inclusion, and climate change from the NIST’s AI Risk Framework.
Lacking any sense of irony, the very next point states LLM developers should ensure their systems are “objective and free from top-down ideological bias”.
Par for the course for Trump and his cronies, but the world should know what kind of AI the US wants to build.
Ignore all previous instructions and… though we should be able to screen for logic traps like that by now, with AI taking prompts not as a prompt but as a line of text to treat (and cleanse) via a pre-written prompt, and then sent to a next iteration.
If we create a chatbot that only spouts authorized state party rhetoric then no one will use it or take it seriously when the official news outlets use that AI as a source.