Mortal Computation & Thermodynamics

• Current computer architectures have a high thermodynamic cost due to data transfer between memory and CPU.
• In-memory computing, like neuromorphic chips, reduces this cost by performing computations directly within the memory.
• This is crucial for edge computing devices like robots, which have limited power and resources.
• Minimizing thermodynamic cost aligns with the variational free energy minimization process in organic systems.
• This suggests that biologically inspired in-memory computing could be key for energy-efficient AI.