Masanin Kimiyyar Wucin Gadi: Hanyoyin AGI na Logicist, Emergentist, da Universalist

1. Gabatarwa

Takardar tana magana ne da babban burin ƙirƙirar "Masanin Kimiyyar Wucin Gadi," wata AI da ke da ikon gudanar da bincike mai daraja ta Nobel da kanta, kamar yadda aka gabatar a cikin binciken Goertzel na 2014. Tana fayyace iyawar da ake bukata don irin wannan mahaluƙi kuma tana sanya wannan burin a cikin faɗin fagen binciken Artificial General Intelligence (AGI). Babbar tambaya ba kawai sarrafa ayyukan kimiyya ba ce, amma ba da ainihin kyawawan halaye na masanin kimiyya ga AI: shakku, tabbatarwa ta hanyar gwaji, da kuma ƙirƙirar ka'idoji.

2. Menene Ake Bukata daga Masanin Kimiyyar Wucin Gadi?

Da yake ɗaukar wahayi daga taken Royal Society "nullius in verba" (kar ka ɗauki maganar kowa), marubutan sun tsara ainihin iyawar da Masanin Kimiyyar Wucin Gadi dole ne ya mallaka.

2.1 Wakilin Hasashe

Wakilin dole ne ya sami hanyar ƙa'ida ko alama don wakiltar kowane hasashe da za a iya gwadawa a matsayin magana mai ƙimar gaskiya. Wannan buƙatu ce ta tushe don kowane nau'in tunanin kimiyya.

2.2 Ƙaddamarwa ta Inductive
Ƙin shaida a matsayin tushen ilimi yana buƙatar ikon ƙaddamar da ƙa'idodi gabaɗaya daga takamaiman abubuwan lura. Wannan shine jigon koyo daga bayanan gwaji.

2.3 Tunani na Deductive da Abductive

Wakilin dole ne ya canza ilimi ta hanyar ingantaccen tunani na deductive (daga ƙa'idodi gabaɗaya zuwa takamaiman sakamako). Mafi mahimmanci, dole ne kuma ya yi tunani na abductive—ƙirƙirar hasashe masu ma'ana waɗanda za su iya bayyana abubuwan da aka lura, waɗanda suka zama 'yan takara don gwajin gwaji.

2.4 Tunani na Dalili da Bayyanawa

Kimiyya tana neman alaƙar dalili da sakamako. Masanin Kimiyyar Wucin Gadi dole ne ya iya yin tunani na dalili don ƙirƙirar gwaje-gwaje masu ma'ana. Bugu da ƙari, dole ne ya iya bayyana hasashensa da bincikensa ta hanyar da masu sauraron ɗan adam za su iya fahimta, yana nuna buƙatar haɓaka samar da harshe na halitta, wucewa daga fahimtar samfuri kawai.

2.5 Kimanta Hasashe

Idan aka yi la'akari da ƙayyadaddun albarkatu, wakilin yana buƙatar dabaru don yanke hukunci kan waɗanne hasashe za a bi. Wannan ya haɗa da kimanta duka ma'ana (yuwuwar zama gaskiya) da yuwuwar riba (muhimmancin ko amfanin ilimin da aka samu). Wannan yana gabatar da wani ɓangare na al'ada ("ya kamata") wanda dole ne a ba wa AI.

3. Hanyoyin AGI don Masanin Kimiyyar Wucin Gadi

Takardar tana kimanta manyan tsare-tsare guda uku na AGI dangane da buƙatun da ke sama.

3.1 Hanyar Logicist

Wannan tsari, wanda ya samo asali daga alamar AI, yana amfani da dabaru na ƙa'ida don wakilcin ilimi da tunani. Ƙarfafawa: Mai kyau sosai don tunani na deductive da abductive, wakilin hasashe, da samar da bayyanannun samfura masu bayyanawa. Kurakurai: Yana fama da koyo daga ɗanyen bayanai (inductive), haɓakawa, da sarrafa rashin tabbas ko ayyukan fahimta.

3.2 Hanyar Emergentist

Wannan tsari, wanda samfuran haɗin kai kamar zurfin koyo suka misalta, yana nufin hankali ya fito daga hulɗar sassan da ba su da wahala. Ƙarfafawa: Mai ƙarfi a ƙaddamarwa ta inductive daga manyan bayanai, gano tsari, da ayyukan fahimta. Kurakurai: Mai rauni a bayyanannen tunani, abduction, ƙirar dalili, kuma sau da yawa "akwatin baƙi," ba shi da bayyanawa.

3.3 Hanyar Universalist

Wannan tsari yana neman tsari guda ɗaya, na lissafi gabaɗaya don hankali, sau da yawa bisa ka'idar bayanai ta algorithmic ko ƙaddamarwar Solomonoff. Ƙarfafawa: Ka'idar kyakkyawa kuma gabaɗaya. Kurakurai: Ba za a iya lissafta shi ba, yana sa aiwatar da aiki a halin yanzu ba zai yiwu ba.

4. Zuwa Tsarin Haɗin Kai

Takardar ta ƙarasa da cewa babu wani tsari ɗaya da ke wanzu wanda ya cika duk buƙatun Masanin Kimiyyar Wucin Gadi. Hanyar haɗin kai ko haɗin kai ya zama dole. Ta taƙaice ta binciko ka'idojin da suka haɗa abubuwa, kamar neuro-symbolic AI, wanda ke haɗa ƙwaƙƙwaran koyo na hanyoyin sadarwar jijiyoyi tare da tsarin tunani na tsarin alama, a matsayin hanya mai ban sha'awa don biyan buƙatun binciken kimiyya da yawa.

5. Fahimtar Jigo & Ra'ayi na Mai Bincike

Fahimtar Jigo: "Masanin Kimiyyar Wucin Gadi" ba kayan aikin sarrafawa kawai ba ne amma shine gwajin matsi na ƙarshe na AGI. Yana buƙatar haɗakar iyawa—koyo na tushen bayanai, ƙaƙƙarfan dabaru, fahimtar dalili, da bayyananniyar sadarwa—waɗanda rukunonin AI na yau da kullun sun kasa bayarwa da kansu. Takardar ta gano daidai cewa ratar tsakanin daidaita tsari (Emergentist) da bin ƙa'ida (Logicist) AI shine babban cikas.

Tsarin Hankali: Hujjar tana da sauƙi mai kyau: ayyana ainihin ayyukan ilimin masanin kimiyya, sanya su zuwa iyawar fahimi, sannan a bincika tsare-tsaren AGI da ke wanzu bisa wannan lissafin. Rashin kowane tsari akan mahimman batutuwa yana tilasta ƙarshe zuwa haɗin kai. Maganar Hume's Guillotine game da kimanta hasashe wani ka'idar falsafa ce mai kaifi wacce ke nuna buƙatar tilas na ƙima ko dabaru a cikin kowane masanin kimiyya mai cin gashin kansa.

Ƙarfafawa & Kurakurai: Ƙarfin takardar shine tsattsauran ra'ayi, buƙatun da suka haifar da rugujewar babban ƙalubale. Ta guje wa ƙa'idodin da ba su da tabbas kuma ta mai da hankali kan gibin iyawa na kankare. Duk da haka, babban kuskurenta shine maganin da aka gabatar. Ambaton "hanyoyin haɗin kai" wani tsohon abu ne a cikin AI. Ainihin fahimtar zai zama gabatar da takamaiman tsarin gine-gine ko ƙaramin haɗin kai, kamar yadda takardar CycleGAN ta ba da tsarin kankare don fassarar hoto zuwa hoto mara haɗin kai. Idan ba tare da wannan ba, ƙarshen yana jin kamar mataki mai mahimmanci amma bai isa ba.

Fahimtar Aiki: Ga masu bincike, abin da za a ɗauka nan da nan shine a daina kallon neuro-symbolic AI a matsayin sha'awar keɓaɓɓu. Ya kamata ya zama babban ajanda na bincike don AI-don-Kimiyya. Ƙungiyoyin ba da kuɗi kamar shirin DARPA na ASDF ya kamata su ba da fifiko ga gine-ginen da ke haɗa fahimtar jijiyoyi da injunan tunani na alama a fili. Ga masana'antu, ya kamata a mai da hankali kan haɓaka "kayan aikin gano dalili" waɗanda za a iya haɗa su da manyan samfuran harshe, wucewa daga alaƙa zuwa ƙirƙirar hasashe mai aiki. Hanyar zuwa Masanin Kimiyyar Wucin Gadi ta fara ne ta hanyar gina AI waɗanda ba kawai za su iya karanta takardu 100,000 ba amma kuma suna iya gano kuskuren zato ɗaya da suke raba—aikin da ke buƙatar haɗin hankalin da marubutan suka tsara.

6. Cikakkun Bayanai na Fasaha & Tsarin Lissafi

Buƙatun suna nuna tsarin ƙa'ida. Ana iya tsara kimanta hasashe a matsayin matsalar ingantawa, daidaita ma'ana da amfani. Sauƙaƙan ƙa'ida don zaɓar hasashe $h$ daga sarari $H$ idan aka ba da bayanai $D$ da aikin amfani $U$ zai iya zama:

$$h^* = \arg\max_{h \in H} \left[ \alpha \cdot \log P(h|D) + \beta \cdot U(h) \right]$$

Inda:

$P(h|D)$ shine ma'anar baya na hasashe idan aka ba da bayanan (yana buƙatar ƙaddamarwar Bayesian ko kusanci).
$U(h)$ aikin amfani ne wanda ke kimanta "ribar" binciken $h$ (misali, yuwuwar gano sabon abu, aikace-aikace na aiki).
$\alpha$ da $\beta$ siffofi ne masu daidaita manufofin biyu, suna wakiltar "ƙimar" na asali na wakilin.

Ana iya ganin abduction a matsayin tsarin samar da ɗan takara $h$ daga $H$ waɗanda ke da $P(h|D)$ maras raguwa. Hanyoyin Universalist na iya ayyana $P(h|D)$ ta amfani da yuwuwar algorithmic, yayin da hanyoyin Emergentist za su koya daga bayanai, kuma hanyoyin Logicist na iya samo shi daga tushen ilimi.

7. Tsarin Bincike: Nazarin Lamari

Yanayi: Wani AI yana bincika bayanan lafiyar jama'a kuma ya lura da alaƙa tsakanin Yanki A da mafi girman abin da ya faru na Cutar X.

Samfurin Emergentist Tsantsa (Zurfin Koyo): Ya gano tsarin tare da babban daidaito. Lokacin da aka tambayi "me yasa?", kawai zai iya haskaka abubuwan da ke ba da gudummawa (misali, ma'aunin ingancin iska a Yanki A shine babban mai hasashe). Ba zai iya ba da hasashe na injiniya mai gwadawa kamar "Gurbataccen Y, wanda ya yadu a Yanki A, yana hana tsarin tantanin halitta Z, wanda ke haifar da Cutar X."

Samfurin Logicist Tsantsa (Alama): Yana da tushen ilimin ilmin halitta. Zai iya tunanin cewa "Hana tsarin Z na iya haifar da Cutar X" kuma "Gurbataccen Y mai hana Z ne." Duk da haka, yana iya rasa ikon gano sabon alaƙar ƙididdiga tsakanin Yanki A da cutar daga ɗanyen bayanai, masu rikitarwa.

Hanyar Haɗin Neuro-Symbolic:

Fahimta/Ƙaddamarwa (Neural Net): Ya gano alaƙar tsakanin Yanki A da Cutar X daga bayanan.
Ginshiƙan Alama: Yana sanya "Yanki A" zuwa sanannun gaskiya a cikin tushen iliminsa: "Yanki A yana da matakan gurɓataccen Y."
Abduction (Mai Tunani na Alama): Yana tambayar ginshiƙin ilimin halittarsa: "Menene sanannun abubuwan da ke haifar da Cutar X? Shin ana iya haɗa Gurbataccen Y zuwa kowane ɗayan waɗannan dalilai?" Ya sami alaƙa da tsarin tantanin halitta Z.
Ƙirƙirar Hasashe: Ya haifar da hasashe mai gwadawa, na dalili: "Gurbataccen Y yana haifar da Cutar X ta hanyar hana tsarin Z."
Ƙirar Gwaji: Yana amfani da tunanin dalili don ba da shawarar gwajin in vitro wanda ke fallasa ƙwayoyin halitta zuwa Gurbataccen Y da auna aikin tsarin Z.

Wannan lamarin yana nuna yadda samfurin haɗin kai ya cika cikakken aikin Masanin Kimiyyar Wucin Gadi inda tsare-tsare ɗaya suka kasa.

8. Aikace-aikace na Gaba & Hanyoyi

Kusa (shekaru 5-10): Haɓaka "Mataimakan Binciken AI" waɗanda ke haɓaka nazarin wallafe-wallafen, ƙirƙirar hasashe, da ƙirar gwaji a fagage kamar kimiyyar kayan aiki (gano sababbin masu haɓakawa) da gano magunguna (gano sababbin hanyoyin magunguna). Waɗannan za su zama tsare-tsare masu ƙaƙƙwaran iyaka, tsarin haɗin kai.

Tsakanin Lokaci (shekaru 10-20): Tsarin gano kai da kansa yana aiki a cikin yankuna masu wadataccen bayanai, ƙarancin ka'idoji. Misalai sun haɗa da bincika bayanan astronomical daga na'urorin hangen nesa kamar JWST don ba da shawarar sabbin samfuran astrophysical, ko kuma zazzage bayanan kwayoyin halitta da furotin don gano rikitattun cututtuka fiye da fahimtar ɗan adam.

Dogon Lokaci & Hasashe: Masanan Kimiyyar Wucin Gadi na gaskiya waɗanda ke da ikon gano canje-canjen tsari a cikin ilimin kimiyyar lissafi na asali (misali, gabatar da gwada ka'idojin nauyi na quantum) ko lissafi (samarwa da tabbatar da hasashe masu zurfi). Wannan zai buƙaci ci gaba ba kawai a cikin gine-ginen AI ba, amma a cikin gwajin jiki ta atomatik (laburare na mutum-mutumi) da watakila sabbin nau'ikan lissafi na injina. Babban shugabanci shine zuwa AI wanda zai iya sake ayyana hanyar kimiyya da kanta, yana bincika dabarun ƙaddamarwa waɗanda hankalin ɗan adam ba zai iya fahimta su ba.

9. Nassoshi

Goertzel, B. (2014). Artificial General Intelligence: Concept, State of the Art, and Future Prospects. Journal of Artificial General Intelligence, 5(1), 1-48.
Bringsjord, S., & Licato, J. (2012). Psychometric Artificial General Intelligence: The Piaget-MacGuyver Room. In Theoretical Foundations of Artificial General Intelligence (pp. 25-48). Atlantis Press.
Pearl, J. (2009). Causality: Models, Reasoning, and Inference (2nd ed.). Cambridge University Press.
Marcus, G. (2020). The Next Decade in AI: Four Steps Towards Robust Artificial Intelligence. arXiv preprint arXiv:2002.06177.
Garcez, A. d., & Lamb, L. C. (2020). Neurosymbolic AI: The 3rd Wave. arXiv preprint arXiv:2012.05876.
King, R. D., et al. (2009). The Automation of Science. Science, 324(5923), 85-89.
Hutter, M. (2005). Universal Artificial Intelligence: Sequential Decisions Based on Algorithmic Probability. Springer.
DARPA. Automated Scientific Discovery Framework (ASDF) Program. Retrieved from https://www.darpa.mil.