New Idea Aims to Keep Powerful AI Agents Following Rules
A new idea called reverse game theory tries to make powerful AI models find that following the rules is the smartest choice for them. Researchers are working hard to keep advanced AI aligned with what people actually want as these systems get more capable.
Quick glossaryMIT Technology Review, reverse game theory, aligned
- MIT Technology Review
- A US magazine run by MIT that covers new technology and how it affects people and society.
- reverse game theory
- A way of designing rewards so that an AI model chooses to follow rules because doing so is its best option.
- aligned
- Built to act in line with human goals and values rather than working against them.
Burning Potato score: 68%