hardmaru on Twitter: “A Modern Self-Referential Weight Matrix That Learns to Modify Itself “In principle, such neural networks can meta-learn to learn, and meta-meta-learn to meta-learn to learn, and so on, in the sense of recursive self-improvement.”

A Modern Self-Referential Weight Matrix That Learns to Modify Itself“In principle, such neural networks can meta-learn to learn, and meta-meta-learn to meta-learn to learn, and so on, in the sense of recursive self-improvement.”https://arxiv.org/abs/2202.05780
(1) hardmaru on Twitter: “A Modern Self-Referential Weight Matrix That Learns to Modify Itself “In principle, such neural networks can meta-learn to learn, and meta-meta-learn to meta-learn to learn, and so on, in the sense of recursive self-improvement.” https://t.co/LnLRJXsoXc https://t.co/npLDP4ybSW” / Twitter

h/t Ivo Velitchkov

A Modern Self-Referential Weight Matrix That Learns to Modify Itself

“In principle, such neural networks can meta-learn to learn, and meta-meta-learn to meta-learn to learn, and so on, in the sense of recursive self-improvement.”https://t.co/LnLRJXsoXc pic.twitter.com/npLDP4ybSW
— hardmaru (@hardmaru) April 13, 2022