Isifinyezo: Uhlaka olusha lwethiyori luphikisa ukuthi ukwehlukana osekunesikhathi eside phakathi kokusebenza kwezibalo kanye nemvelo yebhayoloji kugeja ukuthi ubuchopho bangempela buhlanganisa kanjani ngempela.
Ababhali bahlongoza “i-biological computationalism,” umbono wokuthi ukubala kwe-neural akunakuhlukaniswa namandla obuchopho angokomzimba, ayingxubevange, namandla acindezelayo esikhundleni se-algorithm engaqondakali esebenza kuhadiwe. Kulo mbono, izehlakalo ze-neural ezihlukene kanye nezinqubo zomzimba eziqhubekayo zakha isistimu ehlanganiswe ngokuqinile engakwazi ukuncishiswa ekucubunguleni ulwazi olungokomfanekiso.
Ithiyori iphakamisa ukuthi i-digital AI, naphezu kwamakhono ayo, ingase ingaphinde isenze isitayela esibalulekile sokuhlanganisa esinikeza ulwazi olubalulekile. Esikhundleni salokho, ukuqonda okufana nengqondo kwangempela kungase kudinge amasistimu okwakha okubala kwawo okuvela ku-dynamics yenyama efana naleyo etholakala ebuchosheni bebhayoloji.
Amaqiniso Abalulekile:
- I-Hybrid Dynamics: Ukubalwa kobuchopho kuvela kuma-spikes ahlukene ashumekwe ngaphakathi kwezinkambu zamakhemikhali nezikagesi eziqhubekayo.
- I-Multi-Scale Coupling: Izinqubo ze-Neural zihlala zixhumene ngokujulile kuwo wonke amazinga, okusho ukuthi ama-algorithms awakwazi ukuhlukaniswa nokusetshenziswa ngokomzimba.
- Izithibelo Ezinamandla: Imikhawulo ye-Metabolic ilungisa i-neural computation, ithonya ukufunda, ukuzinza, nokugeleza kolwazi.
Umthombo: I-Estonian Research Council
Njengamanje, impikiswano mayelana nokuqaphela imvamisa izwakala iqhwa phakathi kwezikhundla ezimbili ezigxilile.
Ngakolunye uhlangothi kuhlezi ukusebenza kwekhompyutha, okuphatha ukuqonda njengento ongayichaza ngokugcwele mayelana nokucutshungulwa kolwazi olungacacile: thola inhlangano esebenzayo efanele (kungakhathalekile ukuthi isebenza ngaphi) futhi uthola ukwazi.
Ngakolunye uhlangothi i-biological naturalism, egcizelela ukuthi ukwazi akunakuhlukaniswa nezakhiwo ezihlukile zobuchopho nemizimba ephilayo: i-biology ayiyona nje imoto yokuqonda, iyingxenye yalokho ukuqonda. kuyinto.
Ikamu ngalinye lithwebula okuthile okubalulekile, kodwa ukuvimba kusikisela ukuthi kukhona okushodayo esithombeni.
Ephepheni lethu elisha, siphikisa indlela yesithathu: i-biological computationalism. Umbono uvusa ngamabomu kodwa, sicabanga, uyacacisa. Isimangalo sethu esiyinhloko ukuthi ipharadigm yendabuko yekhompyutha iphukile noma okungenani ayifani neze nendlela ubuchopho bangempela obusebenza ngayo.
Emashumini eminyaka, bekulokhu kulinga ukucabanga ukuthi ubuchopho “bubala” cishe ngendlela efanayo namakhompiyutha ajwayelekile: njengokungathi ukuqonda kuyisoftware, esebenza nge-neural hardware. Kodwa ubuchopho abufani nemishini ye-von Neumann, futhi ukuyiphatha njengokungathi iyasiphoqa ukuba senze izingathekiso ezingaqondakali kanye nezincazelo ezingaqondakali.
Uma sifuna ithiyori engathi sína yokuthi ubuchopho buhlanganisa kanjani nokuthi yini ezodingeka ukwakha izingqondo kwamanye ama-substrates, sidinga ukunweba esikushoyo ngokuthi “ukubala” kwasekuqaleni.
Ukubala kwebhayoloji, njengoba sikuchaza, kunezici ezintathu ezichazayo.
Okokuqala, iyi-hybrid: ihlanganisa imicimbi ehlukahlukene kanye namandla aqhubekayo. Ama-Neurons spike, ama-synapse akhulula ama-neurotransmitters, namanethiwekhi abonisa uguquko olufana nomcimbi, nokho konke lokhu kushumekwe emikhakheni eguquguqukayo yamandla kagesi, ama-chemical gradients, i-ionic diffusion, kanye nokushintshashintsha kwesikhathi.
Ubuchopho abuyona nje idijithali, futhi akuwona nje umshini we-analog. Kuyisistimu enezingqimba lapho izinqubo eziqhubekayo zilolonga khona izehlakalo ezingacacile, futhi izehlakalo ezingacacile zilolonga kabusha ukubukeka kwezwe okuqhubekayo, ku-loop yempendulo engashintshi.
Okwesibili, ayihlukaniseki ngezikali. Kukhompyutha evamile, singadweba umugqa ohlanzekile phakathi kwesofthiwe nehadiwe, noma phakathi “kweleveli yokusebenza” kanye “nezinga lokusetshenziswa.” Ezingqondweni, lokho kwehlukana akuhlanzekile nhlobo.
Awukho umngcele ohlelekile lapho singasho khona ukuthi: nansi i-algorithmfuthi laphaya kukhona izinto ezibonakalayo ezenzekayo ukuze uziqaphele. Indaba eyimbangela ihamba ngezikali eziningi ngesikhathi esisodwa, kusukela kumashaneli e-ion kuye kuma-dendrite kuya kumasekhethi kuya ku-dynamics yobuchopho obuphelele futhi amaleveli awaphathi njengezendlalelo zemojuli kusitaki.
Ukushintsha “ukusebenza” kushintsha “ukubala,” ngoba ezinhlelweni zebhayoloji, lezo zibambene ngokujulile.
Okwesithathu, ukubala kwebhayoloji kusekelwe ku-metabolic. Ubuchopho buyisitho esilinganiselwe amandla, futhi inhlangano yabo ibonisa lokho kucindezela yonke indawo. Okubalulekile, lokhu akuwona nje umbhalo waphansi wobunjiniyela; ilolonga lokho ubuchopho obungakumelela, ukuthi bufunda kanjani, yimaphi ama-dynamics azinzile, kanye nokuthi ulwazi olugeleza luhlelwa kanjani.
Kulo mbono, ukuhlangana okuqinile kuwo wonke amazinga akuyona inkimbinkimbi yephutha. Kuyi- isu lokuthuthukisa amandla: indlela yokukhiqiza ubuhlakani obuqinile, obuguquguqukayo ngaphansi kwemikhawulo enzima ye-metabolic.
Lezi zici ezintathu ziholela esiphethweni esingazizwa singakhululekile uma sijwayele ukucabanga ngamagama ezibalo zakudala: ukubala kwengqondo akukona ukukhohlisa kophawu. Akuyona nje indaba yokuhlanganisa izethulo ngokwemithetho esemthethweni, indlela engokoqobo enikezwe “ukusebenzisa nje.”
Esikhundleni salokho, ekubalweni kwebhayoloji, i-algorithm yi-substrate. Inhlangano ebonakalayo ayigcini nje ukusekela ukubala; yona yakha yona. Ongqondongqondo abamane baqhube uhlelo. Bona kukhona uhlobo oluthile lwenqubo yomzimba eyenza ukubala ngokwembuleka ngesikhathi.
Lokhu futhi kugqamisa umkhawulo obalulekile wokuthi sivame ukukhuluma kanjani nge-AI yesimanje. Amasistimu amanje, ngawo wonke amandla awo, alingisa kakhulu imisebenzi. Balinganisela ukwakheka kwemephu kusuka kokufakwayo kuye kokuphumayo, ngokuvamile ngokujwayelekile okuhlaba umxhwele, kodwa ukubala kuseyinqubo yedijithali esetshenziswa kuzingxenyekazi zekhompuyutha ezidizayinelwe isitayela sekhompyutha esihluke kakhulu.
Ubuchopho, ngokuphambene, buqinisekisa ukubala ngesikhathi somzimba. Izinkambu eziqhubekayo, ukugeleza kwe-ion, ukuhlanganiswa kwe-dendritic, ukuhlangana kwe-oscillatory yendawo, nokusebenzisana okuvelayo kagesi akuyona nje “imininingwane” yebhayoloji esingase siyizibe ngokuphepha ngenkathi sikhipha i-algorithm engabonakali.
Ngokombono wethu, lezi zinjalo ama-primitives ezibalo yohlelo. Ziyindlela ubuchopho obuzuza ngayo ukuhlanganiswa kwesikhathi sangempela, ukuqina, nokulawula okuguquguqukayo.
Lokhu akusho ukuthi sicabanga ukuthi ukwazi kukhethekile ngokumangalisayo empilweni esekelwe kukhabhoni. Asenzi impikiswano “yebhayoloji noma lutho”.
Esikufunayo kucace kakhulu: uma ukwazi (noma ukuqonda okufana nengqondo) kuncike kulolu hlobo lokubala, kungase kudinge inhlangano yokubala yesitayela sebhayoloji, noma ngabe isetshenziswa kuma-substrates amasha.
Ngamanye amazwi, umbuzo obalulekile akukhona ukuthi i-substrate ingokwebhayoloji ngokoqobo, kodwa ukuthi ingabe isistimu iqinisa isigaba esifanele sengxubevange, esikalini esingahlukaniseki, i-metabolically (noma ngokuvamile ngokusebenza ngamandla) ukubala okusekelwe phansi.
Lolo shintsho lushintsha okuhlosiwe kunoma ubani onentshisekelo emicabangweni yokwenziwa. Uma ukubala kobuchopho kungenakuhlukaniswa nendlela okubonwa ngayo ngokomzimba, khona-ke ukukala i-AI yedijithali kukodwa kungase kunganele. Hhayi ngoba amasistimu edijithali awakwazi ukuba namandla, kodwa ngenxa yokuthi amandla ayingxenye yendaba.
Inselele ejulile ukuthi kungenzeka silungiselela into engalungile: sithuthukisa ama-algorithms sibe sishiya i-computing eyisisekelo. i-ontology okungakathintwa.
I-Biological computationalism iphakamisa ukuthi ukuze sinjiniyela amasistimu afana nengqondo ngempela, kungase kudingeke sakhe izinhlobo ezintsha zezinhlelo zomzimba: imishini ikhompuyutha yayo ingafakwanga kahle ku-software yezingxenyekazi zekhompuyutha, kodwa isatshalaliswa kuwo wonke amaleveli, ahlanganiswe ngamandla, futhi asekelwe emigomeni ye-real-time physics namandla.
Ngakho-ke, uma sifuna okuthile okufana nokuqaphela okokwenziwa, inkinga ingase ingabi ukuthi, “Iyiphi i-algorithm okufanele siyiqhube?” Inkinga ingase ibe ukuthi, “Hlobo luni lohlelo lomzimba okufanele lube khona ukuze leyo-algorithm ingakwazi ukuhlukaniswa nokuguquguquka kwayo?”
Yiziphi izici ezidingekayo—ukusebenzelana komcimbi oyingxubevange—ukusebenzelana kwenkundla, ukuhlanganisa izilinganiso eziningi ngaphandle kokusebenzelana okuhlanzekile, izithiyo zamandla ezakha ukucabanga nokufunda—njengokuthi ukubala akuyona incazelo engabonakali ebekwe phezulu, kodwa kuyisici sangaphakathi sohlelo ngokwalo?
Lokho kuyizimfuno ze-biological computationalism: ukusuka ekusesheni ilungelo uhlelo ekusesheni ilungelo uhlobo lwendaba yekhompyutha.
Imibuzo Ebalulekile Iyaphendulwa:
A: Ibhekana nokusimama phakathi kwemibono ebuka ukwazi njengokucutshungulwa kolwazi olumsulwa kanye naleyo esekela ngokuyinhloko kubhayoloji, iphakamisa imodeli ehlanganisa ukubala ngekhompyutha namandla okusebenza ngokomzimba.
A: Ukubala kwebhayoloji kuncike ezinqubweni zomzimba eziqhubekayo, izithiyo zamandla, nokusebenzisana kwezilinganiso eziningi okushintsha ngokuyisisekelo indlela ulwazi olumelwe futhi luguqulwe ngayo.
A: Uma ukwazi kuncike ekubalweni kwesitayela sebhayoloji, khona-ke amasistimu okwenziwa esikhathi esizayo angase adinge izakhiwo ezintsha ezibonakalayo—hhayi nje ama-algorithms edijithali anyusiwe—ukuze aphindaphinde izici ezifana nengqondo.
Amanothi Omhleli:
- Lesi sihloko sihlelwe umhleli weNeuroscience News.
- Iphepha lejenali libuyekezwe ngokugcwele.
- Okuqukethwe okwengeziwe kwengezwe abasebenzi bethu.
Mayelana nalokhu ukwazi kanye nezindaba zocwaningo lwe-AI
Umbhali: UMerilin Reede
Umthombo: I-Estonian Research Council
Othintana naye: UMerilin Reede – Umkhandlu Wocwaningo wase-Estonia
Isithombe: Isithombe sifakwe ku-Neuroscience News
Ucwaningo lwangempela: Vula ukufinyelela.
“Ekuqapheleni kwebhayoloji nokokwenziwa: Icala le-biological computationalism” ka-Jaan Aru et al. I-Neuroscience kanye nokubuyekezwa kwe-Biobehavioral
Abstract
Mayelana ne-biological and Artificial consciousness: Icala le-biological computationalism
Intuthuko esheshayo emakhonweni Ezinhlobo Zolimi Ezinkulu (LLMs) iqinise izinkulumompikiswano zomphakathi nezesayensi mayelana nokuthi izinhlelo zokwenziwa zingase ziqaphele yini ngelinye ilanga. Ithemba elikhona livame ukusekelwe ekusebenzeni kwekhompyutha: umcabango wokuthi ukwazi kunqunywa kuphela iphethini elungile yokucubungula ulwazi, ngaphandle kwe-substrate ebonakalayo.
Ngokuphikisana nalokhu, i-biological naturalism igcizelela ukuthi isipiliyoni sokuqaphela ngokuyisisekelo sincike ezinqubweni ezibonakalayo ezibonakalayo zezinhlelo zokuphila. Ngaphandle kokubaluleka kwalezi zikhundla kunkulumompikiswano yokwenziwa ukwazi, okwamanje alukho uhlaka oluhambisanayo oluchaza ukuthi ukubala kwebhayoloji kuhluke kanjani ekubalweni kwedijithali, nokuthi kungani lo mehluko ungabaluleka ekwazini.
Lapha, siphikisa ngokuthi ukungabi bikho kokuqaphela ezinhlelweni zokwenziwa akubangelwa nje ukuntuleka kwenhlangano esebenzayo kodwa kubonisa ukuhlukana okujulile phakathi kwezindlela zokubala zedijithali nezebhayoloji kanye nokuncika kwe-dynamico-structural yezinto eziphilayo.
Ngokukhethekile, siphakamisa ukuthi amasistimu ebhayoloji asekele ukucutshungulwa kokuqaphela ngoba wona (i) qinisa isikali-engahlukaniswaukucutshungulwa kwe-multiscale okuncike ku-substrate njengesu lokuthuthukisa i-metabolic, kanye (ii) kanye nezibalo ezihlukene, zenza izibalo ezinenani eliqhubekayo ngenxa yemvelo ye-fluidic substrate akhiwa ngayo.
Lezi zici – ukuhlukana kwesilinganiso futhi izibalo ezixubile – akuzona i-peripheral, kodwa zibalulekile endleleni yobuchopho yokubala.
Ngokucabangela lo mehluko, siveza izimiso eziyisisekelo zethiyori yebhayoloji yokubala futhi sichaze ukuthi kungani izinhlelo zamanje zobuhlakani bokwenziwa kungenakwenzeka ukuthi ziphindaphinde ukucubungula okuqaphelayo njengoba kuvela kubhayoloji.
