Jump to content

Tasirin muhalli na hankali na wucin gadi

Daga Wikipedia, Insakulofidiya ta kyauta.
Tasirin muhalli na hankali na wucin gadi
Bayanai
Ƙaramin ɓangare na Abubuwan da suka shafi muhalli
Fuskar Kirkirar Basira (Artificial Intelligence)
Muhimmin darasi Kirkirar Basira (Artificial Intelligence) da Abubuwan da suka shafi muhalli
Ana lura da kwakwalwan GPU da aka yi amfani da su a cikin AI don buƙatar ƙarin makamashi da sanyaya idan aka kwatanta da guntu na CPU na gargajiya.

Tasirin muhalli na hankali na wucin gadi ya haɗa da Amfani da makamashi mai yawa don horo da amfani da samfuran ilmantarwa mai zurfi, da kuma sawun carbon da amfani da ruwa. Wasu masana kimiyya sun ba da shawarar cewa hankali na wucin gadi (AI) na iya samar da mafita ga matsalolin muhalli.

Saurin carbon

[gyara sashe | gyara masomin]

AI tana da mahimman sawun carbon saboda karuwar amfani da makamashi, musamman saboda horo da amfani.[1][2] Masu bincike sun yi jayayya cewa ya kamata a yi la'akari da sawun carbon na samfuran AI yayin horo yayin ƙoƙarin fahimtar tasirin AI.[3] Wani binciken ya ba da shawarar cewa nan da 2027, farashin makamashi don AI na iya ƙaruwa zuwa 85-134 Twh, kusan 0.5% na duk amfani da wutar lantarki na yanzu. Horar da manyan samfuran harshe (LLMs) da sauran AI mai samarwa gabaɗaya suna buƙatar ƙarin makamashi idan aka kwatanta da gudanar da tsinkaya ɗaya akan samfurin da aka horar.[4] Yin amfani da samfurin da aka horar akai-akai, duk da haka, yana iya sauƙaƙe farashin makamashi na tsinkaya.[4] Lissafin da ake buƙata don horar da samfuran AI mafi ci gaba yana ninka sau biyu a kowane watanni 3.4 a matsakaici, wanda ke haifar da amfani da wutar lantarki mai yawa da kuma haifar da sawun carbon.[5] Bugu da ƙari, algorithms na wucin gadi da ke gudana a wurare galibi ta amfani da man fetur don makamashi zai yi amfani da sawun carbon mafi girma fiye da wuraren da ke da tushen makamashi mai tsabta. Ana iya canza waɗannan samfuran don ƙananan tasirin muhalli a farashin daidaito, yana jaddada muhimmancin samun daidaituwa tsakanin daidaito da tasirin muhallu.

Horar da babban samfurin AI yana buƙatar makamashi mai yawa. An kiyasta cewa horar da cikakken samfurin AI yana samar da kusan 626 000 lbs (283 Tons) na carbon dioxide. Ya yi daidai da zirga-zirgar tafiye-tafiye 300 tsakanin New York da San Francisco, ko kusan sau 5 na rayuwar rayuwar matsakaicin mota.[6]

, samfurin harshe da aka horar a cikin 2019, yana buƙatar "makamashi na jirgin tafiya na zagaye" don horar da shi.[7] GPT-3 ya fitar da tan 552 na carbon dioxide a cikin yanayi yayin horo, "daidai da motocin fasinja 123 da ke amfani da man fetur da aka fitar na shekara guda". [7] [8] [9] Yawancin farashin makamashi saboda gine-ginen da ba su da inganci.[7] Ɗaya daga cikin samfuran da ake kira BLOOM, daga Hugging Face, an horar da shi tare da kwakwalwa masu inganci kuma, sabili da haka, kawai ya saki tan 25 na CO2.[8] Haɗakar da farashin makamashi na ƙera kwakwalwan kwamfuta don tsarin ya ninka sawun carbon, zuwa "daidai da kusan jirage 60 tsakanin London da New York. " An kiyasta BLOOM na yau da kullun don saki sawun carbon daidai yayin tuki mil 54. [8][8]

Algorithms waɗanda ke da ƙarancin farashin makamashi amma suna gudana miliyoyin sau a rana na iya samun mahimman sawun carbon.[8] Haɗin AI a cikin injunan bincike na iya ninka farashin makamashi sosai, [7] tare da wasu kimantawa da ke ba da shawarar farashin makamashin da ke ƙaruwa zuwa kusan kWh biliyan 30 a kowace shekara, sawun makamashi ya fi ƙasashe da yawa girma.[10] Wani kimantawa ya gano cewa haɗa ChatGPT a cikin kowane binciken Binciken Google zai yi amfani da TWh 10 a kowace shekara, daidai da amfani da makamashi na shekara-shekara na mazaunan Tarayyar Turai miliyan 1.5.[10]

Once the model is trained, it consumes significantly less energy, however it still requires a high amount of electricity. Researchers have estimated that a ChatGPT query consumes about five times more electricity than a simple web search.[11] In June 2025, OpenAI executive Sam Altman stated that the average ChatGPT query used about 0.34 Wh (1.2 kJ) of electricity and 8.5×10−5 US gal (0.32 ml) of water.[12]

Manazarta

[gyara sashe | gyara masomin]
  1. ↑ Toews, Rob. "Deep Learning's Carbon Emissions Problem". Forbes (in Turanci). Archived from the original on 14 June 2024. Retrieved 4 July 2024.
  2. ↑ Heikkilä, Melissa (5 December 2023). "AI's carbon footprint is bigger than you think". MIT Technology Review (in Turanci). Archived from the original on 5 July 2024. Retrieved 4 July 2024.
  3. ↑ "Achieving net zero emissions with machine learning: the challenge ahead". Nature Machine Intelligence (in Turanci). 4 (8): 661–662. 30 August 2022. doi:10.1038/s42256-022-00529-w. ISSN 2522-5839. Archived from the original on 22 April 2024. Retrieved 4 July 2024.
  4. ↑ 4.0 4.1 Desislavov, Radosvet; Martínez-Plumed, Fernando; Hernández-Orallo, José (1 April 2023). "Trends in AI inference energy consumption: Beyond the performance-vs-parameter laws of deep learning". Sustainable Computing: Informatics and Systems. 38. arXiv:2109.05472. Bibcode:2023SCIS...3800857D. doi:10.1016/j.suscom.2023.100857. ISSN 2210-5379. Archived from the original on 6 July 2024. Retrieved 3 July 2024.
  5. ↑ Sundberg, Niklas (12 December 2023). "Tackling AI's Climate Change Problem". MIT Sloan Management Review (in Turanci). Archived from the original on 27 June 2024. Retrieved 4 July 2024.
  6. ↑ Cooley, Amy. "Research Guides: A Guide to Artificial Intelligence (AI) for Students: Environmental Impacts". libguides.ecu.edu (in Turanci). Retrieved 2025-04-01.
  7. ↑ 7.0 7.1 7.2 7.3 Saenko, Kate (25 May 2023). "A Computer Scientist Breaks Down Generative AI's Hefty Carbon Footprint". Scientific American (in Turanci). Archived from the original on 6 July 2024. Retrieved 3 July 2024. Cite error: Invalid <ref> tag; name ":0" defined multiple times with different content
  8. ↑ 8.0 8.1 8.2 8.3 8.4 Heikkiläarchive, Melissa (14 November 2022). "We're getting a better idea of AI's true carbon footprint". MIT Technology Review (in Turanci). Archived from the original on 3 July 2024. Retrieved 3 July 2024. Cite error: Invalid <ref> tag; name ":1" defined multiple times with different content
  9. ↑ Coleman, Jude. "AI's Climate Impact Goes beyond Its Emissions". Scientific American (in Turanci). Archived from the original on 27 June 2024. Retrieved 3 July 2024.
  10. ↑ 10.0 10.1 Calvert, Brian (28 March 2024). "AI already uses as much energy as a small country. It's only the beginning". Vox (in Turanci). Archived from the original on 3 July 2024. Retrieved 3 July 2024.
  11. ↑ "Explained: Generative AI's environmental impact". MIT News | Massachusetts Institute of Technology (in Turanci). 2025-01-17. Retrieved 2025-04-01.
  12. ↑ Altman, Sam (10 June 2025). "The Gentle Singularity". Posthaven. Retrieved 12 June 2025.
Daga "https://ha.wikipedia.org/w/index.php?title=Tasirin_muhalli_na_hankali_na_wucin_gadi&oldid=691416"