Big AI to humanity: drop dead
6 hours ago
- The author, a lawyer and creator, filed the first lawsuits against generative AI companies for using copyrighted works without permission; there are now 142 such cases in the US.
- Media portrayals of AI catastrophe often rely on sci-fi scenarios (the 'Skynet fallacy'), which are unhelpful for realistic policy change.
- A global ban on AI research is impractical because society has already invested over a trillion dollars in AI, and no technology has ever been preemptively banned globally.
- Existing laws like the Computer Fraud and Abuse Act already cover malicious AI activities, but enforcement is lacking; new laws are empty gestures without enforcement.
- Open-weight language models already pose risks, so halting research would not reduce threats from already-circulating models.
- An NTSB-style investigative body for AI would have limited impact without an FAA-style regulatory agency to impose and enforce safety regulations.
- Requiring public disclosure of AI training runs faces legal hurdles for private companies, and national-security models would be exempt anyway.
- AI 'kill switches' are sci-fi fantasy; AI-generated malware can propagate beyond any cutoff, as seen in historical incidents.
- Big AI companies use contradictory narratives: they downplay risks to avoid liability but exaggerate threats to justify their own protection role, while resisting accountability.
- The author argues for enforcing existing laws against AI companies, prosecuting violations, and applying the rule of law—the most concrete next step.
- The increasing alignment of government with Big AI, such as the House Speaker's call for 'partnership', threatens democratic accountability ('politics without people').