Imagining and building wise machines: The centrality of AI metacognition
2024/11/04 by Samuel G. B. Johnson, Amir-Hossein Karimi, Johnson, Samuel G. B. +19 · 10 voices · 4 citations
#cs.AI #cs.CY #cs.HC
paper · pdf · doi:10.48550/arxiv.2411.02478
Abstract
Although AI has become increasingly smart, its wisdom has not kept pace. In this article, we examine what is known about human wisdom and sketch a vision of its AI counterpart. We analyze human wisdom as a set of strategies for solving intractable problems-those outside the scope of analytic techniques-including both object-level strategies like heuristics [for managing problems] and metacognitive strategies like intellectual humility, perspective-taking, or context-adaptability [for managing object-level strategies]. We argue that AI systems particularly struggle with metacognition; improved metacognition would lead to AI more robust to novel environments, explainable to users, cooperative with others, and safer in risking fewer misaligned goals with human users. We discuss how wise AI might be benchmarked, trained, and implemented.
Cited by
Discussions
- Still up and want a distraction? Here is a proposal for building "wise" AI systems, my first post on ArXiv : Imagining & Building Wise Machines: The Centrality of AI Metacognition arxiv.org/abs/2411.0 [bsky, 17 points, 1 comments]
- arxiv.org/pdf/2411.02478 [bsky, 5 points, 1 comments]
- LLMs can seem smart, but are they *wise*? Interesting new paper on the need for better metacognition among models: arxiv.org/abs/2411.024... [bsky, 4 points, 0 comments]
- One cannot "build wise AI" like a human cannot be taught to be wise.
#Wisdom is a property of an observed #Behavior that only in retrospect is said to be wise (led to an "ideal" outcome).
Also, the ab [bsky, 3 points, 2 comments]
- Preprint arxiv.org/abs/2411.02478 [bsky, 2 points, 0 comments]
- Imagining and building wise machines: The centrality of AI metacognition [hn, 1 points, 0 comments]
- Imagining and building wise machines: The centrality of AI metacognition https://arxiv.org/abs/2411.02478v1 #AI [bsky, 1 points, 0 comments]
- arxiv.org/abs/2411.02478 [bsky, 1 points, 0 comments]
- Imagining and building wise machines: The centrality of AI metacognition #artificialintelligence #metacognition #robustness #explainability #cooperation #safety [bsky, 1 points, 0 comments]
- AI can optimize. But can it recognize uncertainty, balance perspectives, or know when not to act? This paper argues that wise AI requires metacognition — not just better models, but better self-regula [bsky, 0 points, 1 comments]
Related