I, scientist with Balazs Kegl Alexander Ororbia
Nov 14, 2024
Alexander Ororbia, an Assistant Professor of Computer Science at Rochester Institute of Technology and director of the Neural Adaptive Computing Laboratory, dives deep into intriguing concepts like identity and mortality. He discusses how internal regulatory structures define selfhood and what happens to identity when these structures are compromised. The idea of 'mortal computation' is explored, emphasizing how programmed decay can support system survival. Ororbia also highlights the future potential of organoids and maverick computing in understanding biomimetic intelligence.
AI Snips
Chapters
Transcript
Episode notes
Hardware–Software Entanglement Drives Efficiency
- Mortal computation argues hardware and software must be co-designed for true biomimetic intelligence.
- This fusion yields new principles drawn from biophysics, cybernetics, thermodynamics, and cognitive science.
In-Memory Compute Beats Von Neumann Energy Costs
- In-memory, neuromorphic computing approaches approach thermodynamic bounds like Landauer's limit.
- Von Neumann architectures waste energy by shuttling data between memory and compute, unlike biological systems.
Identity As A Non-Equilibrium Steady State
- Free Energy Principle frames identity as maintaining a non-equilibrium steady state (NESS).
- Organisms act (active inference) to preserve their NESS through sensing and acting across a Markov blanket.