Halittu › Alamomin sani
Alamomin sani ga AI
A 2023 masu bincike goma sha tara a neuroscience, falsafa da koyon inji sun lissafa halayen da manyan ka’idojin kimiyya suka ce tsarin mai sani zai kasance da su. Sannan suka duba AI na yau a kan wannan jeri.
Ka’idar sarrafawa mai maimaitawa
- RPT-1 Ɓangarorin shigarwa da suke amfani da maimaituwar algorithm
- RPT-2 Ɓangarorin shigarwa da suke ƙirƙirar wakilcin ji da aka tsara kuma aka haɗa
Ka’idar filin aiki na duniya (global workspace)
- GWT-1 Tsarin ƙwararru da yawa masu iya aiki a lokaci ɗaya (ɓangarori)
- GWT-2 Filin aiki mai iyakataccen ƙarfi, wanda ke haifar da matsatsi a gudanar bayani da tsarin kulawa mai zaɓi
- GWT-3 Watsawa ta duniya: samuwar bayani a filin aiki ga dukan ɓangarori
- GWT-4 Kulawa mai dogara da yanayi, da ke haifar da iya amfani da filin aiki don tambayar ɓangarori a jere don yin ayyuka masu sarƙaƙiya
Ka’idojin tsari mafi girma na lissafi
- HOT-1 Ɓangarorin ji masu ƙirƙira, daga sama zuwa ƙasa ko masu hayaniya
- HOT-2 Sa ido na metacognitive da yake raba amintattun wakilcin ji da hayaniya
- HOT-3 Aiki da tsarin ƙirƙirar imani na gama gari da zaɓin aiki ke jagoranta, da ƙarfin ɗabi’a na sabunta imani bisa sakamakon sa ido na metacognitive
- HOT-4 Sparse da smooth coding da ke ƙirƙirar “sararin inganci”
Ka’idar tsarin kulawa (attention schema)
- AST-1 Samfurin hasashe da ke wakilta kuma yana ba da damar sarrafa halin da hankali yake ciki a yanzu
Hasashen sarrafawa (predictive processing)
- PP-1 Ɓangarorin shigarwa da suke amfani da predictive coding
Aiki da jiki
- AE-1 Aiki: Koyo daga martani da zaɓar sakamako don bin manufofi, musamman inda hakan ya haɗa da amsawa mai sassauƙa ga manufofin da ke gogayya
- AE-2 Jiki: Ƙirƙirar samfurin alaƙar sakamako da shigarwa, har da wasu tasirin tsari, da amfani da wannan samfuri a ji ko sarrafawa
Ƙarshensu
Kimantawarsu: “babu wani tsari na yanzu da ya yi kama da ɗan takara mai ƙarfi na sani.” Suna ƙara da cewa, idan computational functionalism gaskiya ne, tsarin AI masu sani “za a iya gina su da gaske a nan kusa.”
Butlin P, Long R, Elmoznino E, Bengio Y, Birch J, et al. (2023). Consciousness in Artificial Intelligence: Insights from the Science of Consciousness. arXiv:2308.08708. https://arxiv.org/abs/2308.08708 (an karanta rubutun Table 1 kai tsaye daga PDF)