[PDF] Protecting Society from AI Misuse: When are Restrictions on Capabilities Warranted?-论文阅读讨论-ReadPaper

摘要

Artificial intelligence (AI) systems will increasingly be used to cause harmas they grow more capable. In fact, AI systems are already starting to be usedto automate fraudulent activities, violate human rights, create harmful fakeimages, and identify dangerous toxins. To prevent some misuses of AI, we arguethat targeted interventions on certain capabilities will be warranted. Theserestrictions may include controlling who can access certain types of AI models,what they can be used for, whether outputs are filtered or can be traced backto their user, and the resources needed to develop them. We also contend thatsome restrictions on non-AI capabilities needed to cause harm will be required.Though capability restrictions risk reducing use more than misuse (facing anunfavorable Misuse-Use Tradeoff), we argue that interventions on capabilitiesare warranted when other interventions are insufficient, the potential harmfrom misuse is high, and there are targeted ways to intervene on capabilities.We provide a taxonomy of interventions that can reduce AI misuse, focusing onthe specific steps required for a misuse to cause harm (the Misuse Chain), anda framework to determine if an intervention is warranted. We apply thisreasoning to three examples: predicting novel toxins, creating harmful images,and automating spear phishing campaigns.

展开全部

图表提取

论文十问

参考文献

被引用

提问与回答

暂无人提供速读十问回答