Constraints of Biological Neural Networks and TheirConsideration in AI Applications

Biological organisms do not evolve toper fection, but to out compete others in their ecological niche, and therefore survive and reproduce. This paper views the const r aints i mposed on imper fec t organisms, p ar ticularly on t heir neur al systems and abilit yto capture a nd process infor mation accur ately. By understanding biological constraints of the physical properties of neurons,simpler a nd more e fficient ar t ificial n eu r al n etwor ks can be made (e.g ., spiking networ ks w i ll t r ansmit less infor mation than g r adedpotential n etwor ks, spikes only occur i n n ature due to limitations of car r y ing elect r ical ch arges over large distances). Fu r t her m ore,understanding the b ehav iour al and e colog i cal const r aints on animals allows an understanding of the l imitations of bio-inspiredsolutions, but also an understanding of why bio-inspired solutions m ay fail and how to cor rect t hese failures.