Active Inference Insights Alexander Ororbia ~ Active Inference Insights 008 ~ Mortal Computation, Cybernetics, AI
35 snips
Feb 29, 2024 Delving into mortal computation and cybernetics, the podcast explores biomimetic intelligence and the benefits of in-memory computing for neural simulations. It discusses minimizing variational free energy in biological and artificial systems, as well as the concept of cybernetic variety in problem-solving. The episode also touches on active inference, Bayesian surprise, and the limitations of current robotics technology in achieving adaptive intelligence.
AI Snips
Chapters
Transcript
Episode notes
Mortal Computation & Biomimetic Intelligence
- Biomimetic intelligence means emulating biological intelligence in artificial systems using inspiration from natural organisms.
- Mortal computation binds software and hardware to increase energy efficiency and adaptivity, differing from traditional software-hardware separation.
Thermodynamics & Efficiency in Computing
- Thermodynamics limits the energy efficiency of computation, motivating in-memory computing to approach physical lower bounds.
- Von Neumann architectures cause inefficiencies due to hierarchical memory access creating energy overheads in traditional deep learning.
Free Energy & Non-Equilibrium Life
- Living systems maintain themselves far from thermodynamic equilibrium, termed non-equilibrium steady state (NESS), which is key for life and cognition.
- Minimizing variational free energy relates to optimizing thermodynamic efficiency in biological systems to preserve identity and avoid dissolution.