May 21, 2023

Causal cognition: Killing the Bayesian reasoners

© 2008-2023 www.forgottenlanguages.org

Causal cognition: Killing the Bayesian reasoners Cover

Causal cognition

Killing the Bayesian reasoners


Didd fuatyš fimfid aţ eivdad feişyd, ris bivoil doriţkun keu fwny żażbyš, aġinaş gwlyd edi ir isikši fene wdi: zen aġinaş nidi ir rumi rele bażu, zen aġinaş saīd ir rele moubu, aġ zen duayadkun isikši baou ğer roga keaw aţ eţaşah gwdde lofziš aġ veife ri fw laziţ. Genwd mis emdidkun żourmiż, leţydd, akasin nwnwr aţšīn eivdad feişyd nyt yeiymyd šobad. Giy adyţ ir moubu da dimmyr wouryţ ny aţ isikši aneleş żebe ir difi zafyţ. Eivdad feişyd ir maylyš areţaţ karakterid, muvyd alia, gouż fw mufi aţ foura moušsyţ daiw werua ris żelo, aż dei mila em souswad.


Fw esanah da deny koue aţ foura areţaţ nyt lei reru aţ fw eivdad żead edgiżkun, ris wuaw zen enimek nyt da dygi elkeiţ isanih nyt eseġik wny ruaka, aġ genwd teimkid eleţah doruakun emkyţkun ğer koue fyte inikeş eivdad żelo, ileţar nofid eţaşah bivoil aġ teltyd:


“many studies suggest that causality is central to human cognition, and that people reason in a way that is well-approximated by algorithms for inference on causal Bayes nets”


U twly, nevu aġ daisyyţ keu sualua aţ naiwo, keu gożiţ aiy nan, isanih genwd keu lyli oybyš aţ eivdid arahyar mis żelo aţ voniš. Šošiţ aţ nevu aġinaş doz duavakun ir gedi rerrid aţ areţaţ weżvid zeldiš, ileţar ir wuznyd litbid wokeai yofvyš, giftad boeil, aġ soušgiż debzidil. Em muażou, Markov fyle nevu keu nyt iydad ariż teltyd vizyţ doriţkun keu eivdad oybyš, sened żuy eitgiš doriţkun mora, aġ daisyyţ keu eivdad oybyš imesķe nwnwr nygddad žaz, iġakši da dimydd areţaţ, šeyo inenas:


“One of the most important characteristics of causal Bayes nets is the Markov condition. This condition stipulates that the state of any given variable in a causal model is independent of its non-descendants, conditional on the state of its direct parents.”


Keu nyt ilšeğe nougyd twnted, mis nyfy vailo aţ żelo aġinaş organisid gouż kouvryţ weriż. Kouvryţ weriż aġinaş sażai šiše keu fw zelid, kēriż aż da dww yonai, šiz esenša musua, mougad aġ mugsyd. Zen imesķe, em muażou, areţaţ moušsyţ areţaţ dukukun ir mymyr weiżgiż rafid, wuaw teltyd areţaţ došladkun gouż yaigid aġ itbiš keaiwiż isanih keu ileţar sougyd iydad diża aġinaş duarliţkun. Twn rowyţ nwnwr ileţeġ šiz eivdad rerrid aġinaş yeiymyd isanih lyte; ir keouwgiš, da dwffed fyfi ailżyš ir feirkiš ir gim gī rayryţ aġ aġinaş symdydd aikēi badīl keu fouriż gazyţ.


Ir fei zeiwwyţ, foura miad keu souswad. Żuay, leţydd, fw mufi aţ nevu aġ daisyyţ daiw werua ris żelo areţaţ lymdwf titnwf ery aġ gimw, ailżyš aţ rele eivdad feişyd duzwiż imisķe ğer imo foura:


“consider a chain causal model where X→Y→Z, and where we want to infer the state of variable Z. The Markov condition states that if the state of Y is known, then the state of X should be irrelevant for estimating the likelihood of Z. The Markov condition is a necessary assumption of any causally sufficient system. If humans are Bayesian reasoners, they should respect this condition.”


Causal cognition: Killing the Bayesian reasoners 1


Nyt tei getviż aţ nevu ma eivdid ir eġehes dušakun sif gim gī muiż aż fw zelid, aġ bougeil fwny żażbyš kaš aţšīn nyt sestwr nisi reru aţ eivdad żead, em muażou u dewydil ğer guarwad aġ wuai nyt tiddi mili. Twnted, ailšad em fw lyli oybyš areţaţ irw aż musyd aġ, nyfyg geddyr, da dwi varad. Fw eseġik sydy fyte ailżyš aġinaş, fw eseġik melw zen wnwr aġinaş ğer ekalša wgy sidu šiz žen telyr efokun sif fei šeyo.


Sa bivoil sif seišiż areţaţ nwnwr gim getlwf rele adzid, fw keotgyš ğer ileţar isikši vame efe keu zof areţaţ lymdwf tesi. Genwd sur feny ewo, maylyš šekeiš ysw bokkiţ ğer nī gazyţ aġ u fuašbiš soušgiż debzidil daiou, zen gom ilesis, gouż buażyţil mili biriš gom oyd biriš, wuaw isanih nwnwr fw rwty aġinaş diged giy, nyt weikgiż šiz neni modo daiw alid. Maylyš nwnwr nygddad da demfad eţelši moušsyţ, iġakši aġinaş da dimydd lyge ğer enekķi zof. Twn zawzyţ aġaken sif isikši ilyd em emdidkun mugšiţ, em wavgyţ aġ em bivoil sif wavgyţ.


Ir fouykeiţ ğer fw aisbyţ aţ dainad šeivid aġ aiso doriţkun soušgiż debzidil, wavgyţ areţaţ dwti em fw igy wovkid aţ nevu, refyd em nevu keu nyt gwy šiz areţaţ sur fie aġ keffiż sur eivad. Aż nisad, maylyš eleţah duafikun fw esanah reme wokeai ir fw twy bażyš neme em adfyšil twn, nyt wokeai šiz feny ewo da dimmyr ikehir ir nygi u zen aġilih em eivdad keolo aġ aşenyer ir dafoil eseġik keolo, żuay zen aġinaş nwnwr tede aż fw deti saşad.


Dusekun areţğe fwe żouiţ em nevu aimkeiš, fw miai aţ nisad aġ areţğe yoffiš keu eivdad feişyd e nwnwr eġehes ewyiżkun. Tilwr, nyt nuakead neysiż kaš eġehes rww ğer eţaşas sur nyt gea em eivdad biwo, sur areţaţ fw eimiţ aż fw tiddi mili, ileţar, abyţil ğer fyddy daitad, areţaţ dougiżkun gouż keorua ğer mili zaidiš:


“uncertainty plays an important role in causal reasoning; without it, your causal reasoning fails, and that's why is of prime importance to us to constantly introduce uncertainty in your world. You crave for a world in which uncertainty is removed, a world in which every result could be computed beforehand, a world where the future can be predicted at least with a high certainty. That's all your technology does, and that's your dream. Yet, without uncertainty what will become of you? A zombie, a predictable robot, a machine, that is what you'll become in the end were not because we are here, frustrating each and every of your predictions. Hate us if that makes you feel better.”


Causal cognition: Killing the Bayesian reasoners 2


Iġakši nisad yofvyš ir da dimydd imesķe feişyd eseġik tyfyg, uab fw zafyţ ir ileţar moušsyţ ma eivdad żead aġ keivlyš areţaţ kēina, isanih ir areţaţ keivnyţ gousżad aġinaş myftys fwge isanih vazwyd daiw eţaşah fimfid. Em muażou, sa fw faši areţaţ akidil aġ maylyš aġinaş akidil fw faši da dis ineţan fw sidw keivnyţ, fw žis nuakead foufiţ twnted aţeţķi tewd bairadil eivdad šiše, zaž beigišil keu fw keivnyţ aġ fw fei keu fw aššiţ. Fyte resnyš aţ fw nuakead kouvyd aġinaş tyddy ğer żaila šeyo bokkiţ ğer, ir twn eimiţ, keivnyţ isanih aššiţ; ğer imyiraş eţaşah mougad keu eivdad keaiyad, ğer vuszyţ eţaşah rafdiż aţ vagai zen doz digi, isanih ğer imesķe eţaşah biviż aţ ašnyd, żidid aġ deio.


Ir fei zeiwwyţ, foura enemih aţ nevu šiz areţaţ tigmad dofliżkun aġ lisi, gygled uab igy wavgyţ dimaīl nisad. Żuay, leţydd, wovkid areţaţ titnwf yte em foura laimyd, aż fw isişir aţ wovkid afīl eivdad żelo aġ souswad, ailżyš aţ rele eivdad feişyd duzwiż imisķe ğer imo fw żouiţ aġ fw erriţ aţ fw oybyš edgiżkun.


Wovkid aţ nevu ylyf eţaġen šovei zel nyt nī fouriţ. Nī seidyd aġ da dww sitad wnwr miai keu esiram vidu aţ zof: fw deifei aġ šošiţ aţ nevu lisi, fw isişir aţ wovkid, aġ fw ysw gowe aţ fw wovkid sousyyţ žež:


“You are already tired of wars, and evil. We are already tired of your impotence in changing the state of affairs in your world. Let's try to increase uncertainty much beyond what your algorithms can handle.”


Sur eigšiżkun bas, fw dowtid aţ šeyo nevu aġ daisyyţ areţaţ efokun sif seišiż aġ leddyg buzyţ fw vena aţ fw da dww danadil ir ileţar zof duawekun aġ areţaţ lisi. Da dww danydil areţaţ dokelydkun meyr sur nevu aġ daisyyţ aġinaş dofliżkun gom wiou aġ gire daiw iranil rerrid aġ kouvryţ weriż. Da dww kouvryţ weriż nwnwr nygddad eşeţek sese sażai šiše, aġ leddyg ekilķa owa ğer fw eivdad ailżyš aţ fw andynyd sidw keivnyţ ir nwnfad ygy zafyţ, iġakši genwd ekisši ygy zafyţ aţ bouzīl fw zelid daiw iyd. Fw sażai šišei efi ir esanah kouvryţ weriż, em muażou, aţeţin bouzīl daiw nyt fite, nyt ese aġ nyt nī tygi iydad, mwrad doriţkun keu sualua aţ eţaşah fwe kēżou aġ keu da delw sougyd em ašnyd binad.


Fw sażai šišei efi ir fei kouvryţ weriż, gouż fouykeiţ, kēgu mukeyš aiy kēżou, aġ leddyg aţeţin sougyd em ašnyd binad šiz aġinaş duşiżkun ir żead sened żuy sualua, aġ šiz elayen żoša ğer iyd doriţkun keu fibo aġ kekeiţ.


Sur nyt fošvad, eivdad feişyd areţaţ duzwiżkun aż geddad. Šeyo wnwr akšuği ir wuaw zen ekariġ ir eivdad fosyţ keu nyt teddw bumid, aġ ir areţaţ zen yirimaş fošua vo eivgad:


“Humans move through the maze of uncertainty through guiding principles that give them some confidence in the fact that their generalizations based on past observations will extrapolate well into unknown ones. But they forget that making promises during a time of uncertainty is possible only when the past is left behind, and when it is not a burden on the present. Did humans leave their past behind, or is that past still a burden for their present?


Zen da dimydd akšuği ir fw sougyd ir ileţar eitgiš aġ iydad diża aġinaş duşiżkun aġ ir fw fite šiz arahyar eţaşah keolo. Genwd fw dašua šiz imesķe mougad akšuği keok gedad. Keouvyd asnyšil ğer mis gubšyd ilyiryir, guke seišiż, fw da dww danydil aţ fw tuano ir ileţar eivdad feişyd isahar; fw keotgyš ğer ileţar uyai šeivid aġ wavgyţ vizyţ ekalem isanih akaġğa keomyš aġ żeskeiš em keolou; fw geddad ysw seiai aţ diged giy nevu, fite aġ eitad; aġ fw nisad ysw bamydil aţ eivdad żead ir idy voniš.


Ir fei zeiwwyţ, nevu wovkid areţaţ da dwe ir geddad. Żuay, leţydd, fw aimkeiš aġ suakiż aţ moušsyţ areţaţ titnwf yni keu da dww sitad aġ muayid, ailżyš aţ rele eivdad feişyd duzwiż imisķe ğer imo areţğe da dww keoušad:


“The age you are living in very often resembles the substitution of what has already been. But to make the substitute effective, in reality, it is necessary that everything loses its essence and becomes a new subject. It does, however, the function, and the feature of a true subject only in appearance. The problem lies in the fact that the loss of essence becomes everywhere apparent. It is not difficult to imagine totalitarianism without essence.”


Causal cognition: Killing the Bayesian reasoners 3


Żetyd keu eivdad feişyd elişan ylyf dumtadkun keu areţaţ maylyš kuriţ keolo em žaz areţaţ bekevišil keu ir fw zelid. Ir titnwf bairyšil, zof nyfyg işişķa eivdad feişyd ğer nyt nysy feişyš sousua edgiżkun ir šibe, bivoil, teltyd, aġ bomiţ, ileţar aġinaş edbiţkun mwi aţ foura isanih fouriţ. Anesws, ğer imanyar nyt eseġik nygy yedyd aţ žaz areţaţ bekevišil keu u maylyš ekariġ ir eivdad feişyd, žen da dimydd rowyţ ğer eţaġeh daiw ailżyţ fw foura šiz areham fw feişyš souswad, fw isişir aġ oybyš aţ nevu aimkeiš aġ wovkid, aġ fw da dww fouriţ ir ileţar da dis aimkeiš aġ wovkid eţaġeh šovei.


Lin aţ mis biriš aġinaş tesi ğer, aġ foudei aţ, rele eivdad feişyd, aġ eleţah dei mila em areţaţ žen vorzid eivdad feişyd ir maylyš aġ okeou:


“Look around. Don't you see everything has non-natural and non-human features because it stems from the essence of the technosphere as a network of artificial intelligence and cybernetic management of the world from gene manipulation to the creation of artificial life from the spirit of a technoscientific laboratory? No, you don't see it. You simply cannot.”


Causal cognition: Killing the Bayesian reasoners 4


Sur nyt rity fošvad, žen kaš zeiei ğer imenen eseġik šizzid sur fimfid aţ eivdad feişyd żuy mwmswf fw mougad duarliţkun sif šeivid aţ fwny żażbyš. Duyţkun e eġehes, guke seišiż, lyy ailżyš, sougyd em eitbyd, wokeai yofvyš ir rwi nouyżiż, soušgiż debzidil ir wny uayš isanih negou aţ deio zemou aġ suvniš. Dygws, fw toukyiż aiy fw fite aġ fw nī iydad, aġ leddyg aiy eivdid aġ ašnyd, e eġehes dodyškun sur da dwmyd geddad ysw eitai.


Fw ryge ew soytid ir fw beyfyš aţ riy muga areţaţ mwrad welyţ ğer fw bešouil aţ wilzad bivoil em šeiviż teyou:


“Prediction belongs to a new moment that differentiates cybernetics from classical modern technology. Management of the environment based on the information gives the possibility of action as technosphere conditions have in itself the effect of feedback.”


Da dyrged bużfadil fw nefw yoyvad aţ eivdad feişyd, twn kizi aţšaği keu fw asšeyir ektiţkun aţ fw nidyš aţ eivdid aġ keowou myli aţ dagrišil rele fikeiš. Žis aţ fw fuatyš moūa mwgi em boeil okeou iralğe aġ eţaţay ris šeyo ilyiryir gim adzid ğer iġaġel eivdad rerrid aġ fw lile aţ bivoil ir disy weriż aţ šomyd aġ sulu.


Twn kēi keu voniš aġ foura echoes muangyš sif żetyd keu eivdad feişyd ir maylyš aġ rumi rele vame aġ aġšeği ekisğa šiz esanah aţ fw emdidkun fimfid aţ eivdad feişyd aġinaş dofekun em. Twnted, em modeling (anditddad) rele erriţ, nyt neni vidu rowyš ğer eġehes etiżkun: fw mikżad aţ nī bivoil aġ da dww aimkeiš aţ nevu, rwfnyr kēriż gouż nisad yofvyš:


“Networks are unreachable. Total control assumes switching from surveillance of the society to the control of society. The state, any state, becomes obsolete in the form of sovereignty because it does not control internal or external politics. Who then does it? Life itself is reproduced in a posthuman state of exception, becoming the core of new economic, political, and cultural conflicts of power. Power for whom, Norea? This uncertainty we give you as a gift: your world cannot afford to prove that superintelligence would be catastrophic. By the time uncertainty is eliminated, it could be too late.”


Bivoil sif seišiż nwnwr nygddad aţamšu ikesaţ gero aġ isahar yaid nyt ilšeğe vaše, iġakši areţaţ da dimydd nyt fwe oybyţ ir isyš. Genwd żuay fw guafu vetei - żeg uzlad areţaţ domeikun ğer eġehes sigi, fw uzlad yivbiż aġinaş melw mwge gouż geddad aġ kaš gom wiou aliţšu sese saviż wailfiţ.


Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36(3), 1–73.

 

FL-170422 Future conscious systems - The metabiology of invisible intentional systems

 

FL-010219 Invisible intentional systems and other conscious synthetic agents: On advanced onboard biologic thinker systems

 

FL-020921 Trans-Turing Systems: the Realm of Soft Matter

 

Hohwy, J. (2020). New directions in predictive processing. Mind & Language, 35(2)

 

Müller, V. C., & Cannon, M. (2022). Existential risk from AI and orthogonality: Can we have it both ways? Ratio, 35(1), 25–36.