Forcing the Babylon Bee to spoil its jokes is just the setup. Control is the punchline
The most dangerous thing about artificial intelligence may not be what the machine thinks. It may be who gets to tell it what is true.
TL;DR
- The primary danger of AI may be who dictates its understanding of truth, not what the machine itself thinks.
- A lawsuit against New Mexico over AI-generated satire disclaimers illustrates the struggle over who classifies information.
- Efforts to 'debias' AI systems involve human judgment, substituting one set of biases for another.
- Historical and contemporary examples show a pattern of controlling information through labels like 'trust and safety' or 'disclosure'.
- The principle of contestability, where evidence and assumptions are open to challenge, is proposed as a solution to institutional bias.
- Making bias visible through comparing sources and exposing assumptions is more effective than pretending bias can be eliminated.
- True self-governance requires citizens to examine evidence, distrust authority, and form their own conclusions.
- The ultimate threat is not AI thinking, but granting someone the power to define permissible thought.