The case for self-improving hardware in full recursive self-improvement
The author of SchmidhuberAI’s post claims to have published the first concrete recursive self-improvement algorithms in 1987. They argue that full RSI will need self-improving hardware as well as software.
TLDR
The author of SchmidhuberAI’s post traces their recursive self-improvement (RSI) work to algorithms they say they published in 1987, and cites later work on self-modifying policies, neural networks and the Gödel Machine. They say software-based RSI has become practical, but argue that full RSI will also require self-improving hardware in the physical world.
Combined views
2 Sources, first seen 3h ago
reposts
