# This one equation may be the root of intelligence

• Can we not do click­bait ti­tles on linkposts, please? Let’s use the Hacker News rule—de­fault to the ar­ti­cle ti­tle, but if it’s not a good rep­re­sen­ta­tion of the con­tent of the ar­ti­cle (e.g. it’s click­bait), change it to some­thing de­scrip­tive.

• And in gen­eral, can we NOT try to evolve in the HuffPo di­rec­tion?

• You won’t be­lieve this life chang­ing equa­tion!

• Given that this was posted to LW, you’d think this link would be about a differ­ent equa­tion..

• Namely? Bayes? (TBH I wouldn’t ex­pect bayes be­cause that’d be wrong, I think—you can have “dumb” in­tel­li­gence based on re­in­force­ment learn­ing)

• This equa­tion is sim­ply the sum of each x = i choose k for k in [ 1, i ].

So what he’s say­ing is that the neu­ral cir­cuits that fol­low the prin­ci­ples he de­scribes have one neu­ron to rep­re­sent ev­ery pos­si­ble com­bi­na­tion of on/​off states in the set of in­puts. It’s the most brain-dead way you could pos­si­bly im­ple­ment a clas­sifier sys­tem.

• Does the mag­i­cal 2^i-1 equa­tion pre­dict that the hu­man brain with cca 85-86 billion neu­rons can only con­tain 36 differ­ent con­cepts?

• From a pa­per by Dr. Tsien, re­trieved from http://​​www.au­gusta.edu/​​mcg/​​dis­cov­ery/​​bbdi/​​tsien/​​doc­u­ments/​​the­o­ry­of­con­nec­tivity.pdf

Fifth, this power-of-two math­e­mat­i­cal logic con­fines the to­tal num­bers of dis­tinct in­puts ( i ) com­ing into a given micro­cir­cuit in or­der to best uti­lize the available cell re­sources. For in­stance, as a re­sult of its ex­po­nen­tial growth, at a mere i = 40, the to­tal num­ber of neu­rons ( n ) re­quired to cover all pos­si­ble con­nec­tivity pat­terns within a micro­cir­cuit would be more than 10^12 (already ex­ceed­ing the to­tal num­ber of neu­rons in the hu­man brain). For Caenorhab­di­tis el­e­gans – which has only 302 neu­rons, limit­ing i to 8 or less at a given neu­ral node makes good eco­nomic sense. Fur­ther­more, by em­ploy­ing a sub-mod­u­lar ap­proach (e.g., us­ing a set of four or fi ve in­puts per subn­ode), a given cir­cuit can greatly in­crease the in­put types it can pro­cess with the same num­ber of neu­rons. ′

He also men­tions cor­ti­cal lay­er­ing. It seems like he’s en­vi­sion­ing the brain as a for­est of smaller, rel­a­tively shal­low net­works fol­low­ing the prin­ci­ples he de­scribes, rather than one tree where all neu­rons are wired to­gether in a uniform way.

• “In stark con­trast, Tsien pre­dicts the brain runs on a se­ries of pre-pro­grammed, con­served net­works. Th­ese net­works are not learned; in­stead, they’re made up of pre-es­tab­lished neu­ral net­works, wired ac­cord­ing to a sim­ple math­e­mat­i­cal prin­ci­ple.

In other words, at a fun­da­men­tal level the brain’s wiring is in­nate — the mo­tifs, es­tab­lished by ge­net­ics, un­der­lie our abil­ity to ex­tract fea­tures, dis­cover re­la­tional pat­terns, ab­stract knowl­edge and ul­ti­mately, rea­son.”

Brain Com­pu­ta­tion Is Or­ga­nized via Power-of-Two-Based Per­mu­ta­tion Logic

http://​​jour­nal.fron­tiersin.org/​​ar­ti­cle/​​10.3389/​​fn­sys.2016.00095/​​full

″ the unify­ing math­e­mat­i­cal prin­ci­ple upon which evolu­tion con­structs the brain’s ba­sic wiring and com­pu­ta­tional logic rep­re­sents one of the top most difficult and un­solved meta-prob­lems in neu­ro­science”

“This sim­ple math­e­mat­i­cal logic can ac­count for brain com­pu­ta­tion across the en­tire evolu­tion­ary spec­trum, rang­ing from the sim­plest neu­ral net­works to the most com­plex.”

• Thanks! Very in­ter­est­ing!

• And the an­swer to the ques­tion about Life, the Uni­verse, and Every­thing is… 42.