AI Experts Propose New Regulations to Control Dangers of Unchecked AI Progress
-
AI researchers from OpenAI and universities release 104-page document advocating for regulating AI hardware and compute power.
-
Document discusses risks of unchecked AI progress, including threats to privacy, leakage of sensitive information, economic impacts, concentration of power, and specialized low-compute models.
-
Proposed solutions include global registry of AI chips, unique identifiers, kill switches to deactivate malicious AI systems remotely.
-
Risks and challenges around proposed regulations are also detailed, like cybercriminals hijacking kill switches.
-
As AI technology evolves, experts hope more regulatory power consolidates to prevent dangers of alternative.