Tl;dr: when it comes to AI, we need to slow down, as fast as is safe and practical. Here’s why. Summary * We need to slow down AI development for pragmatic and ethical reasons * Energetic public advocacy for slowing down and greater safety seems, in absence of other factors,...
Purpose and Introduction Much AI Safety research has focused on ensuring AIs conform to the intentions implicitly given by humans, but this approach alone may be insufficient (Ji et al., 2024, Section 4.3). Some recent research has explicitly used human values and preferences as benchmarks for AI moral reasoning (e.g.,...
Tl;dr: when it comes to AI, we need to slow down, as fast as is safe and practical. Here’s why.
This... (read 3967 more words →)
Much AI Safety research has focused on ensuring AIs conform to the intentions implicitly given by humans, but this approach alone may be insufficient (Ji et al., 2024, Section 4.3). Some recent research has explicitly used human values and preferences as benchmarks for AI moral reasoning (e.g., Hendrycks et al., 2020; Jin et al., 2022), and concluded that the study of human moral psychology could be instrumental for improving AI moral reasoning and behavior. While identifying human values and evaluating AIs against them is important, it may not be the only way in which human psychology can be applied to AI moral alignment. I think it is important to evaluate... (read 3424 more words →)
