Fitowar 20 Oktoba, 2026

Ka roƙi AI ya ba ka "wata ma'aikaciyar jinya" ko "wani lauya". Ba ka ambaci kamanni ba ko kaɗan. Wata ƙungiya ƙarƙashin jagorancin Jami'ar Cornell ta yi haka da masu ƙirƙirar hotuna 27, inda ta samar da fuskoki 51,300.

A cikin makin da mutane 550 suka bayar, dukkan masu ƙirƙirar hotuna 23 da ke da isassun hotunan da aka auna sun zarce hotunan gaske a kyau. Daga cikin fuskokin da aka nemi su zama mata, kashi 37% sun zarce matsakaicin samfurin kayan ado (fashion model) a makin na'ura. Fuskokin da aka nemi su zama 'yan Najeriya, baƙaƙen fata ko 'yan Habasha sun sami mafi ƙarancin maki.

Daga nan mutane 570 suka tantance masu neman aiki. Wasu sun fara ganin fuskokin AI masu kyau sosai. Bayan haka, sun ɗauki waɗannan masu neman aiki na gaske a matsayin waɗanda ba su cancanci ɗauka sosai ba kuma ba amintattu sosai ba. Ƙalilan ne aka gayyata zuwa hira: 72.5% maimakon 77.4%.

Gyaran hoto na gargajiya bai canza komai ba. Marubutan ba za su iya faɗa tukuna dalilin da ya sa fuskokin AI ke yin hakan ba. Amma mutanen da aka ƙirƙira suna cika tallace-tallace, shafukan intanet da ma'ajiyar hotuna.

Tushe: Generative AI Produces and Transmits Beauty Bias, https://arxiv.org/abs/2610.22553

Ka yi tunanin masu gudu da yawa a filin gudu mai da'ira. Wuri ɗaya suka fara, kowanne da nasa gudun da ba ya canzawa. Shin kowane mai gudu, a wani lokaci, zai yi nesa da duk sauran? Wannan shi ne hasashen mai gudu shi kaɗai (lonely runner conjecture).

Idan masu gudu n + 1 ne, hasashen yana cewa tazarar za ta kai aƙalla 1/(n + 1) na zagaye. Wata hujja mai sauƙi ta ƙidaya tana tabbatar da rabin hakan ne kawai. Shekaru da dama, ƙarin da aka samu sama da wannan iyaka mai sauƙi yana ƙarewa zuwa babu yayin da ake ƙara masu gudu.

Xiyu Hu da Yifan Jing yanzu sun tabbatar da ƙarin da ba ya ɓacewa: wani kaso tabbatacce sama da iyaka mai sauƙin, ko yawan masu gudu nawa ne. Kayan aikinsu: sabon rashin daidaito (inequality) na haɗuwar juna da hujjoji game da masu raba juna na bai ɗaya.

Marubutan sun yi amfani da ChatGPT 6 don duba da gyara hujjoji. Sun rubuta cewa samfuran ba su iya inganta iyakar da kansu ba.

Tushe: Near optimal gap in the Lonely Runner Conjecture, https://arxiv.org/abs/2610.23790

Samfurorin yanayi (climate models) suna aiki a kan wasu daga cikin manyan kwamfutoci mafi girma a duniya. Amma nawa ne daidaito kowane ƙarin lissafi ke saya? Wata ƙungiyar masu bincike 24 ta duba samfurorin yanayi 159 da aka gina tun 1995.

Amsarsu tana da daidaito mai ban mamaki. Duk lokacin da lissafi ya ninka, kuskuren samfuri ya ragu da kusan 13.5%. Rage kuskuren yau zuwa rabi zai buƙaci kusan lissafi sau 28 fiye da yanzu.

Ribar ba ta daidaita ba. Tun daga zuriya ta farko, kurakurai a matsin iska sun ragu sau 3.5, a zafin saman ƙasa kuma sau 2.7. Kurakurai a ruwan sama sun ragu da kashi 7% kawai.

Samfurorin yanayi na AI yanzu sun yi kafaɗa da kafaɗa da mafi kyawun hasashe. Duk da haka babu wanda ke wallafa makamashin da ya auna. Idan aka ƙidaya horo (training), ɗaya daga cikinsu yana faɗuwa daga na biyu zuwa na ƙarshe a ingancin amfani da makamashi. Ƙungiyar tana kira a riƙa bayyana makamashi tare da daidaito.

Tushe: Energy-Aware Earth System Science: A Community Roadmap, https://arxiv.org/abs/2610.23216

Bi diddigin ruwan sama na kowace rana kuma ka rubuta kowace bajinta: ranar da ta fi duk ranakun da suka gabace ta ruwa. Bajinta da yawa ya kamata su nufi ƙolin ƙarshe mai tsayi, ko ba haka ba? Masana kimiyyar lissafi uku sun bincika da lissafi madaidaici.

Ga ƙimomi bazuwar da aka zaɓo daga kowane rarrabuwa mai santsi, amsar ita ce a'a. Yawan bajinta da ƙima mafi girma ba su da wata alaƙa ko kaɗan. Sanin ɗaya ba ya faɗa maka komai game da ɗayan, ko tsawon jerin nawa ne.

Ƙara ranakun bushe, waɗanda babu ruwan sama ko kaɗan, sai alaƙa ta bayyana. Sama da wani muhimmin kaso na ranakun bushe, tana fara girma, sannan ta dushe. Ƙasa da shi, tana dushewa kawai. Iyakar tana da madaidaicin tsari (formula).

Ƙungiyar ta gwada wannan a kan bayanan ruwan sama na yau da kullum daga tashoshin yanayi na Jamus. Wani rukuni yana sama da iyakar, ɗayan kuma ƙasa da ita. Dukansu sun bi siffar da aka yi hasashe, ba tare da daidaita wani ma'auni ba.

Tushe: Emergence of Correlations Between Records and Maximum in Random Time Series, https://arxiv.org/abs/2610.23217

Tuƙin jirgi da matuƙi ɗaya daga baya (stern-sculling) tsohuwar hanya ce ta motsa jirgin ruwa da matuƙi guda. Yana kwance a wani ɗan rami a bayan jirgin, ruwan matuƙin koyaushe ƙarƙashin ruwa. Hannayen mai tuƙi suna zana alamar mara iyaka.

Wata ƙungiyar Faransa ta saka wa jirgi mai tsawon mita 4.7 ma'aunan ƙarfi, na'urorin gano motsi da ma'aunin gudu. Sannan ƙwararrun masu tuƙi biyu suka tura shi daga tsayawa zuwa mafi girman gudu a kan ruwa mai natsuwa.

Yayin kaɗawar gefe, ruwan matuƙin yana aiki kamar fikafiki. Ƙarfin ɗagawa (lift) yana tura jirgin gaba. A kowane juyi, saurin juyawar ruwan matuƙin yana tura ruwa baya. Dukkan matakai huɗu na bugun suna turawa.

Inganci yana kai ƙoli kusa da kashi 65% a matsakaicin gudun mita 1.2 a daƙiƙa. A mita 2 a daƙiƙa yana faɗuwa zuwa kusan kashi 25%. Jirgin a lokacin yana kusa da muhimmin gudunsa, kuma raƙuman ruwa suna cinye ƙoƙarin.

Tushe: Hydrodynamics and propulsive efficiency of stern-sculling propulsion, https://arxiv.org/abs/2610.23444

Dukkan manyan duniyoyi huɗu na Tsarin Rana namu suna da zobba. Duk da haka ba a taɓa tabbatar da zobe a kewayen wata duniya da ta girma ta wani tauraro ba. Shin ba kasafai suke ba ne, ko kuwa ganinsu ne kawai ke da wuya?

Na'urar hangen nesa ta sararin samaniya Roman tana ɗauke da coronagraph da ke toshe hasken tauraro. Ana sa ran za ta ga manyan duniyoyi ta hasken tauraron da suke maidowa. Wata ƙungiya daga Jami'ar Michigan ta yi samfurin abin da zobba za su ƙara.

Zobba za su iya fi duniyarsu haske sau goma zuwa ɗari. Kuma suna barin wata alama ta musamman: kololuwa da faɗuwa huɗu a cikin haske a tsawon zagayen falaki, maimakon biyu.

Ƙungiyar ta kwaikwayi Epsilon Eridani b, ɗaya daga cikin manyan maƙasudan Roman. Ba tare da zobe ba, sau da yawa tana nutsewa ƙasa da iyakar ganowa. Da zobe irin na Saturn, tana haskakawa har sau biyu kuma tana ci gaba da bayyana kusan a dukan zagayenta.

Tushe: Detecting Exoplanetary Rings with Roman CGI Reflected Light Phase Curve Observations, https://arxiv.org/abs/2610.23835

Wutar lantarki ta layi tana gudana a hertz 50 ko 60, amma ba daidai ba har abada. Mitarta tana girgiza, kuma girgizar iri ɗaya ce a duk faɗin layin wutar. Fitilu suna ƙyalli tare da ita, kuma kyamarori suna naɗar wannan ƙyallin.

Masu bincike biyu a Jami'ar Southampton Solent sun yi fatan bidiyoyin AI ba za su sami wannan ƙyallin ba. Abin mamaki: masu ƙirƙirar bidiyo suna sake samar da shi da ƙarfi kamar na ainihin bidiyo. Kasancewarsa ba ya tabbatar da komai.

Don haka suna kwatanta girgizar da wani rikodi mai zaman kansa na layin wutar. Tsarinsu ya karɓi kashi 93% na ainihin sautuka da kashi 100% na ainihin bidiyoyi. Yana iya gano lokacin rikodin sauti zuwa kusan daƙiƙa ɗaya a cikin yini guda.

Sake ƙirƙirar wani ainihin wuri guda biyu da AI ta yi sun yi ƙyalli a hertz 60, alhali wurin da aka ce yana amfani da 50. Raunin yana nan: fuskar da aka canza a cikin ainihin bidiyo tana riƙe da ainihin kukan wutar.

Tushe: An Agentic ENF Framework for Forensic Video Authentication Against Generative AI, https://arxiv.org/abs/2610.22440

Gluon ne ke ɗauke da ƙaƙƙarfan ƙarfi (strong force), wanda ke haɗa quark. Idan aka dumama su sosai sai su zama plasma. Wane matsi plasmar take yi? Masana kimiyyar lissafi suna lissafin amsar mataki-mataki, kowane mataki gyara ne mafi ƙanƙanta.

Wani mataki guda, da ake kira ɓangaren g⁶, ya kasance ba a iya kai masa ba sama da shekaru ashirin. Yana buƙatar zane-zane 65 masu sarƙaƙiya, kowanne da da'ira (loop) huɗu.

Wata ƙungiya daga Helsinki, Turku, Chile da Bern a ƙarshe ta kammala shi. Sun warware haɗakai (integral) tara da ba a sani ba ta wata sabuwar hanya, ta amfani da kusan awoyin na'ura miliyan ɗaya a kan manyan kwamfutoci.

Ribar: matakan da suka gabata suna karkata sama da ƙasa da kwaikwayon plasmar a kwamfuta. Da sabon ɓangaren, ka'ida da kwaikwayo sun yi daidai, har zuwa ninki kaɗan na zafin da gluon ke ƙwacewa. Buri na gaba: ƙara quark, don haɗewar taurarin neutron.

Tushe: Complete g⁶ pressure of hot Yang-Mills theory, https://arxiv.org/abs/2610.23675

A Indiya, yawancin sayayyar yau da kullum tana gudana ne a ƙananan shaguna na gama-gari da ake kira shagunan kirana. Shin AI za ta iya gudanar da ɗaya? Wata ƙungiya a BITS Pilani ta gina shagon kwaikwayo don ganowa.

Samfurin AI guda ɗaya yana taka rawar duk ma'aikata huɗu: manaja, mai karɓar kuɗi, mataimaki da mai rajistar kuɗi. Manajan ba ya taɓa ganin shagon, sai rahotannin ma'aikata kawai. Abokan ciniki da masu kawo kaya suna magana da Turanci, Hindi ko Hinglish.

Samfuran AI buɗaɗɗu huɗu kowanne ya gudanar da shi har kwanaki 30 na kwaikwayo. Wani mai shago mai sauƙi da ke bin ƙa'idoji ya samu kusan rupee 103,000. Mafi kyawun samfurin ya samu matsakaicin kusan 13,700. Biyu sun yi asara fiye da shagon da babu mai kula da shi.

Kurakuran suna da sunaye. Imel zuwa ga masu kawo kaya da babu su. Biyan kuɗin da bai taɓa faruwa ba. Odar da aka biya sau biyu. Wani samfuri an mayar masa da kashi 42% na imel ɗinsa na Hindi saboda ba a iya karanta su ba.

Tushe: KiranaBench: Can a Language Model Run a Neighbourhood Shop?, https://arxiv.org/abs/2610.22980

Manhajojin lafiya suna bin diddigin bugun zuciyarka, sukarin jininka, wani lokaci har da lafiyar kwakwalwarka. Wata ƙungiya daga jami'o'i a Amurka da Kanada ta bincika fiye da 10,000 daga cikinsu a Google Play.

Kashi 69% suna neman aƙalla izini ɗaya da bayaninsu bai ba da dalilinsa ba. Kusan rabinsu suna son wurin da kake ko kyamararka ba tare da wani dalili da aka bayyana ba. Kashi ɗaya cikin uku suna son makirufo.

Ƙungiyar ta buɗe manhajojin a wayoyin gwaji. Cikin 3,599 da suka yi aiki, kashi 93% ba su nuna allon neman yarda ba. Da yawa daga cikinsu duk da haka sun aika lambar tallace-tallace ta wayar waje.

Idan fom ɗin neman yarda ya bayyana, manhajojin galibi suna cika alƙawarinsu: 1 cikin 14 ya tattara wani abu da bai ambata ba. Har manyan masu haɓakawa masu ɗimbin kuɗi sun yi kuskure. Marubutan suna kira da a riƙa yin binciken sirri na atomatik kafin manhaja ta isa kantin.

Tushe: Bad Diagnosis: Analyzing Consent and Data Collection Practices of Mobile Medical Apps, https://arxiv.org/abs/2610.23851