Summary of Beyond Preferences in Ai Alignment, by Tan Zhi-xuan et al.
Beyond Preferences in AI Alignment
by Tan Zhi-Xuan, Micah Carroll, Matija Franklin, Hal Ashton
First submitted to arxiv on: 30 Aug 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper challenges the dominant preferentist approach to AI alignment, which assumes preferences are an adequate representation of human values and AI systems should be aligned with human preferences. The authors argue that this approach neglects the semantic content of human values and ignores the incommensurability of those values. They critique the normativity of expected utility theory (EUT) for humans and AI, highlighting its limitations and silence on which preferences are acceptable. Instead, the paper proposes reframing AI alignment to align systems with normative standards that promote mutual benefit and limit harm, negotiated by all relevant stakeholders. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary AI alignment is about making sure artificial intelligence (AI) systems behave safely and in a way that respects human values. Right now, most people think this means aligning AI with what humans want or like. But some researchers are questioning whether this approach is good enough. They’re saying that it doesn’t take into account the complexity of human values and might not even be the right goal for AI alignment. Instead, they suggest we should focus on aligning AI with certain standards that promote fairness and prevent harm, regardless of what humans like or dislike. |
Keywords
» Artificial intelligence » Alignment
