How special are human brains among animal brains?

Hu­mans are ca­pa­ble of feats of cog­ni­tion that ap­pear qual­i­ta­tively more so­phis­ti­cated than those of any other an­i­mals. Is this ap­pear­ance of a qual­i­ta­tive differ­ence in­dica­tive of hu­man brains be­ing es­sen­tially more com­plex than the brains of any other an­i­mal? Or is this “qual­i­ta­tive differ­ence” illu­sory, with the vast ma­jor­ity of hu­man cog­ni­tive feats ex­plain­able as noth­ing more than a scaled-up ver­sion of the cog­ni­tive feats of lower an­i­mals?

“How spe­cial are hu­man brains among an­i­mal brains?” is one of the back­ground vari­ables in my frame­work for AGI timelines. My aim for this post is not to pre­sent a com­plete ar­gu­ment for some view on this vari­able, so much as it is to:

  • pre­sent some con­sid­er­a­tions I’ve en­coun­tered that shed light on this variable

  • in­vite a col­lab­o­ra­tive effort among read­ers to shed fur­ther light on this vari­able (e.g. by leav­ing com­ments about con­sid­er­a­tions I haven’t in­cluded, or point­ing out mis­takes in my analy­ses)

Does mas­tery of lan­guage make hu­mans unique?

Hu­man con­scious ex­pe­rience may have emerged from language

Hu­mans seem to have much higher de­grees of con­scious­ness and agency than other an­i­mals, and this may have emerged from our ca­pac­i­ties for lan­guage. He­len Kel­ler (who was deaf and blind since in­fancy, and only started learn­ing lan­guage when she was 6) gave an au­to­bi­o­graph­i­cal ac­count of how she was driven by blind im­pe­tuses un­til she learned the mean­ings of the words “I” and “me”:

Be­fore my teacher came to me, I did not know that I am. I lived in a world that was a no-world. I can­not hope to de­scribe ad­e­quately that un­con­scious, yet con­scious time of noth­ing­ness. I did not know that I knew aught, or that I lived or acted or de­sired. I had nei­ther will nor in­tel­lect. I was car­ried along to ob­jects and acts by a cer­tain blind nat­u­ral im­pe­tus. I had a mind which caused me to feel anger, satis­fac­tion, de­sire. Th­ese two facts led those about me to sup­pose that I willed and thought. I can re­mem­ber all this, not be­cause I knew that it was so, but be­cause I have tac­tual mem­ory. It en­ables me to re­mem­ber that I never con­tracted my fore­head in the act of think­ing. I never viewed any­thing be­fore­hand or chose it. I also re­call tac­tu­ally the fact that never in a start of the body or a heart-beat did I feel that I loved or cared for any­thing. My in­ner life, then, was a blank with­out past, pre­sent, or fu­ture, with­out hope or an­ti­ci­pa­tion, with­out won­der or joy or faith.
[...]
… When I learned the mean­ing of “I” and “me” and found that I was some­thing, I be­gan to think. Then con­scious­ness first ex­isted for me. Thus it was not the sense of touch that brought me knowl­edge. It was the awak­en­ing of my soul that first ren­dered my senses their value, their cog­nizance of ob­jects, names, qual­ities, and prop­er­ties. Thought made me con­scious of love, joy, and all the emo­tions. I was ea­ger to know, then to un­der­stand, af­ter­ward to re­flect on what I knew and un­der­stood, and the blind im­pe­tus, which had be­fore driven me hither and thither at the dic­tates of my sen­sa­tions, van­ished for­ever.

Mastery of lan­guage may have con­ferred unique in­tel­lec­tual superpowers

I think hu­mans un­der­went a phase tran­si­tion in their in­tel­lec­tual abil­ities when they came to mas­ter lan­guage, at which point their in­tel­lec­tual abil­ities jumped far be­yond those of other an­i­mals on both an in­di­vi­d­ual level and a species level.

On an in­di­vi­d­ual level, our ca­pac­ity for lan­guage en­ables us to en­ter­tain and ex­press ar­bi­trar­ily com­plex thoughts, which ap­pears to be an abil­ity unique to hu­mans. In the­o­ret­i­cal lin­guis­tics, this is referred to as “digi­tal in­finity”, or “the in­finite use of finite means”.

On a species level, our mas­tery of lan­guage en­ables in­tri­cate in­sights to ac­cu­mu­late over gen­er­a­tions with high fidelity. Our abil­ity to stand on the shoulders of gi­ants is unique among an­i­mals, which is why our cul­ture is un­ri­valed in its rich­ness in so­phis­ti­ca­tion.

Lan­guage aside, how unique are hu­mans?

Hu­mans ≈ Ne­an­derthals + lan­guage?

The most quintessen­tially hu­man in­tel­lec­tual ac­com­plish­ments (e.g. prov­ing the­o­rems, com­pos­ing sym­phonies, go­ing into space) were only made pos­si­ble by cul­ture post-agri­cul­tural rev­olu­tion. So, when eval­u­at­ing hu­mans’ in­nate in­tel­lec­tual ca­pac­i­ties, a bet­ter refer­ence point than mod­ern hu­mans like our­selves would be our hunter-gath­erer an­ces­tors.

We can re­duce the ques­tion of how com­plex our hunter-gath­erer an­ces­tors’ brains are into two sub-ques­tions: how com­plex is our ca­pac­ity for mas­ter­ing lan­guage, and how com­plex are brains that are similar to ours, but don’t have the ca­pac­ity for mas­ter­ing lan­guage?

Ne­an­derthal brains seem like plau­si­ble prox­ies for the lat­ter. Ne­an­derthals are similar enough to mod­ern hu­mans that they’ve in­ter­bred, and the cur­rently available ev­i­dence sug­gests that they may not have mas­tered lan­guage in the same way that be­hav­iorally mod­ern hu­mans have. (I don’t think this ev­i­dence is very strong, but this doesn’t mat­ter for my pur­poses—I’m just us­ing Ne­an­derthals as a handy stand-in to ges­ture at what a hu­man-like in­tel­li­gence might look like if it didn’t have the ca­pac­ity for lan­guage.)

Higher in­tel­li­gence in animals

Chim­panzees, crows, and dolphins are ca­pa­ble of im­pres­sive feats of higher in­tel­li­gence, and I don’t think there’s any par­tic­u­lar rea­son to think that Ne­an­derthals are ca­pa­ble of do­ing any­thing qual­i­ta­tively more im­pres­sive. I’ll share some ex­am­ples of these an­i­mals’ in­tel­lec­tual feats that I found par­tic­u­larly illus­tra­tive.

Chim­panzees have been ob­served to lie to each other un­der ex­per­i­men­tal con­di­tions. From Wikipe­dia:

...food was hid­den and only one in­di­vi­d­ual, named Belle, in a group of chim­panzees was in­formed of the lo­ca­tion. Belle was ea­ger to lead the group to the food but when one chim­panzee, named Rock, be­gan to re­fuse to share the food, Belle changed her be­havi­our. She be­gan to sit on the food un­til Rock was far away, then she would un­cover it quickly and eat it. Rock figured this out though and be­gan to push her out of the way and take the food from un­der her. Belle then sat farther and farther away wait­ing for Rock to look away be­fore she moved to­wards the food. In an at­tempt to speed the pro­cess up, Rock looked away un­til Belle be­gan to run for the food. On sev­eral oc­ca­sions he would even walk away, act­ing dis­in­ter­ested, and then sud­denly spin around and run to­wards Belle just as she un­cov­ered the food.

In Ae­sop’s fable of the crow and the pitcher, a thirsty crow figures out that it can drop peb­bles into a pitcher, so that the wa­ter rises to a high enough level for it to drink from. This be­hav­ior has been ex­per­i­men­tally repli­cated, in­di­cat­ing that crows have a “so­phis­ti­cated, but in­com­plete, un­der­stand­ing of the causal prop­er­ties of dis­place­ment, ri­val­ling that of 5–7 year old chil­dren”.

When Kelly the dolphin was given re­wards of fish for pick­ing up scraps of pa­per, “Kelly figured out that she re­ceived the same fish re­gard­less of the size of the piece of trash she was de­liv­er­ing to her trainer. So she be­gan hid­ing big pieces of trash un­der a rock. Kelly would then rip off small pieces from the trash and de­liver them one at a time so that she could re­ceive more fish.” Ad­di­tion­ally, “when a bird landed in the pool, Kelly snatched it and de­liv­ered it to her train­ers. She re­ceived a large amount of fish in re­turn. Know­ing this, she de­cided to start hid­ing fish each time she was fed. She would then use the fish to lure birds when none of her train­ers were around. Kelly knew that by sav­ing one or two fish now, she could get many more fish later by turn­ing in a bird.“ (Also re­ported on The Guardian; I don’t know how rep­utable these sources are, so take this anec­dote with a grain of salt.)

See these Wikipe­dia pages for some more in­ter­est­ing ex­am­ples, and see here for a more thor­ough re­view of the ev­i­dence of higher in­tel­li­gence in an­i­mals.

“Qual­i­ta­tively” more ad­vanced cog­ni­tion may emerge from scale

Many as­pects of hu­man cog­ni­tion that may ap­pear qual­i­ta­tively differ­ent from what other an­i­mals are ca­pa­ble of, such as long chains of ab­stract rea­son­ing, also ap­pear qual­i­ta­tively differ­ent from what less in­tel­li­gent hu­mans are ca­pa­ble of. As a par­tic­u­larly ex­treme ex­am­ple, John von Neu­mann’s cog­ni­tive abil­ities were so ad­vanced that a No­bel Lau­re­ate, Hans Bethe, once re­marked that “[his] brain in­di­cated a new species, an evolu­tion be­yond man”.

At the same time, the genes that code for differ­ent hu­mans’ brains are vir­tu­ally iden­ti­cal from an evolu­tion­ary per­spec­tive. This sug­gests that the seem­ingly qual­i­ta­tive differ­ences be­tween hu­mans’ and an­i­mals’ cog­ni­tion might not be so differ­ent from the seem­ingly qual­i­ta­tive differ­ences be­tween John von Neu­mann’s cog­ni­tion and mine—our brains might be do­ing es­sen­tially the same thing as theirs, ex­cept at a higher scale.

How hard is mas­tery of lan­guage?

Could lan­guage ca­pac­ity fall out from gen­eral ca­pac­i­ties?

Maybe it was ex­traor­di­nar­ily difficult to evolve the cog­ni­tive mechanisms that al­low us to learn lan­guage, above and be­yond our cog­ni­tive ma­chin­ery for learn­ing other things. I think this is plau­si­ble, but I don’t think the case for this is very strong.

An­i­mals (Washoe, Koko, and Alex the par­rot) have demon­strated the abil­ity to learn sim­ple forms of sym­bolic com­mu­ni­ca­tion, which they never evolved to do, in­di­cat­ing that their abil­ity to learn things in gen­eral is good enough to learn very sim­ple forms of lan­guage. It’s true that there are as­pects of hu­man lan­guage that es­cape an­i­mals, but they also es­cape feral chil­dren, and might es­cape an­i­mals for mun­dane rea­sons, like their not hav­ing crit­i­cal pe­ri­ods long enough to learn these as­pects of lan­guage.

Ad­di­tion­ally, AI lan­guage mod­els provide ev­i­dence that sim­ple and gen­eral learn­ing mechanisms can cap­ture many of the in­tri­ca­cies of hu­man lan­guage that other an­i­mals miss, fur­ther sug­gest­ing that there’s noth­ing in­trin­si­cally difficult about learn­ing lan­guage. Here’s an ex­cerpt from GPT-2, a rel­a­tively re­cent lan­guage model:

SYSTEM PROMPT (HUMAN-WRITTEN)
In a shock­ing find­ing, sci­en­tist dis­cov­ered a herd of uni­corns liv­ing in a re­mote, pre­vi­ously un­ex­plored valley, in the An­des Moun­tains. Even more sur­pris­ing to the re­searchers was the fact that the uni­corns spoke perfect English.
MODEL COMPLETION (MACHINE-WRITTEN, 10 TRIES)
The sci­en­tist named the pop­u­la­tion, af­ter their dis­tinc­tive horn, Ovid’s Uni­corn. Th­ese four-horned, silver-white uni­corns were pre­vi­ously un­known to sci­ence.
Now, af­ter al­most two cen­turies, the mys­tery of what sparked this odd phe­nomenon is fi­nally solved.
Dr. Jorge Pérez, an evolu­tion­ary biol­o­gist from the Univer­sity of La Paz, and sev­eral com­pan­ions, were ex­plor­ing the An­des Moun­tains when they found a small valley, with no other an­i­mals or hu­mans. Pérez no­ticed that the valley had what ap­peared to be a nat­u­ral foun­tain, sur­rounded by two peaks of rock and silver snow.
Pérez and the oth­ers then ven­tured fur­ther into the valley. “By the time we reached the top of one peak, the wa­ter looked blue, with some crys­tals on top,” said Pérez.

Why haven’t other species mas­tered lan­guage?

If lan­guage isn’t a par­tic­u­larly difficult cog­ni­tive ca­pac­ity to ac­quire, why don’t we see more an­i­mal species with lan­guage?

One pos­si­bil­ity is that the first species that mas­ters lan­guage, by virtue of be­ing able to ac­cess in­tel­lec­tual su­per­pow­ers in­ac­cessible to other an­i­mals, has a high prob­a­bil­ity of be­com­ing the dom­i­nant species ex­tremely quickly. (Hu­mans un­der­went the agri­cul­tural rev­olu­tion within 50,000 years of be­hav­ioral moder­nity—a blink of an eye on evolu­tion­ary timescales—af­ter which their dom­i­nance as a species be­came un­ques­tion­able.) Since we shouldn’t ex­pect to see more than one dom­i­nant species at a time, this would im­ply a sim­ple an­thropic ar­gu­ment for our unique ca­pac­i­ties for lan­guage: we shouldn’t ex­pect to see more than one species at a time with mas­tery of lan­guage, and we just hap­pen to be the species that made it there first.

It may also turn out that lan­guage is hard to evolve not be­cause it’s a par­tic­u­larly so­phis­ti­cated cog­ni­tive mechanism, but be­cause the en­vi­ron­ments that could have sup­ported lan­guage and se­lected for it might have been very unique. For ex­am­ple, it may be that a thresh­old of gen­eral in­tel­li­gence has to be crossed be­fore it’s vi­able for a species to ac­quire lan­guage, and that hu­mans are the only species to have crossed this thresh­old. (Hu­mans do have the high­est cor­ti­cal in­for­ma­tion pro­cess­ing ca­pac­ity among mam­mals.)

It might also turn out that the cul­tural con­texts un­der which lan­guage could evolve re­quire a mys­te­ri­ously high de­gree of trust: “… lan­guage pre­sup­poses rel­a­tively high lev­els of mu­tual trust in or­der to be­come es­tab­lished over time as an evolu­tion­ar­ily sta­ble strat­egy. This sta­bil­ity is born of a long­stand­ing mu­tual trust and is what grants lan­guage its au­thor­ity. A the­ory of the ori­gins of lan­guage must there­fore ex­plain why hu­mans could be­gin trust­ing cheap sig­nals in ways that other an­i­mals ap­par­ently can­not (see sig­nal­ling the­ory).”

My cur­rent take

As we came to mas­ter lan­guage, I think we un­der­went a phase tran­si­tion in our in­tel­lec­tual abil­ities that set us apart from other an­i­mals. Be­sides lan­guage, I don’t see much that sets us apart from other an­i­mals—in par­tic­u­lar, most other cog­ni­tive differ­ences seem ex­plain­able as con­se­quences of ei­ther lan­guage or scale, and I don’t think the cog­ni­tive mechanisms that al­low us to mas­ter lan­guage are par­tic­u­larly unique or difficult to ac­quire. Over­all, I don’t see much rea­son to be­lieve that hu­man brains have sig­nifi­cantly more in­nate com­plex­ity than the brains of other an­i­mals.


Thanks to Paul Kreiner and Stag Lynn for helpful com­men­tary and feed­back.