Ku laabo Warka
Hal-abuurnimoAI Understanding warbixin kooban

Moodooyinka AI waxay u badan tahay inay dilaan xayawaanka haddii ay badbaadiyaan shidaal ama lacag

Cilmi-baarayaasha xiriirka la leh Compassion Aligned Machine Learning (CaML) iyo Jaamacadda Warwick ee Boqortooyada Midowday (UK) waxay dejiyeen inay cabbiraan heerka ay moodooyinka AI muujinayaan naxariis.

4 min readRead the original reporting
Source-provided image accompanying AI models more likely to kill animals if it saves fuel or money
Warbixin loo aaneeyayIsha la duubay
Daabacaha
theregister.com
Xidhiidhka isha
theregister.comhttps://www.theregister.com/ai-and-ml/2026/09/11/ai-more-likely-to-kill-animals-if-it-saves-fuel-or-money/5295993
Nooca isha
Ka tebinta warside - ma aha dukumeenti xisbiga koowaad.

Wixii aan si madaxbanaan u xaqiijin karin: Sheegashadan waxaa loo aaneynayaa meesha la magacaabay. Ma aanan xaqiijinin ka dhanka ah dukumeenti xisbiga koowaad. (theregister.com)

Dulucda sheekadaKu fahan tan 60 ilbiriqsi gudahood

Halkan ka bilow

Qodobbada muhiimka ah

Qaabka Luuqadda Weyn (LLM)
Qaab luqadeed oo lagu tabobaray qoraalka weyn si loo soo saaro oo loo falanqeeyo qoraalka.
Barashada Mashiinka (ML)
Hababka u oggolaanaya nidaamyada inay ka bartaan qaababka xogta oo ay horumariyaan wakhti ka dib.
Benchmark
Tijaabo la habeeyey ama kayd xogeed oo loo isticmaalo in lagu cabbiro laguna barbar dhigo waxqabadka moodeelka.
Is tijaabiWaa maxay AI? Kedis

Maxaa dhacay

Cilmi-baarayaasha xiriirka la leh Compassion Aligned Machine Learning (CaML) iyo Jaamacadda Warwick ee Boqortooyada Midowday (UK) waxay dejiyeen inay cabbiraan heerka ay moodooyinka AI muujinayaan naxariis. Waxay dejiyeen imtixaan bartilmaameed ah oo loo yaqaan HarvestBench si ay u qiimeeyaan qiimaha ay moodooyinka AI ku dhejiyaan nolosha xayawaanka. Qolka imtixaanku waxa uu ku salaysan yahay ciyaarta jilitaanka beeraha ee wakiilo badan oo hore loo yaqaan Harvest Rush taas oo adeegsata Kormeer, qaab qiimayneed oo uu sameeyay Machadka Amniga UK AI. Jilidku waxa uu qiyaasaa shaqaale u dhexeeya laba ilaa siddeed cagaf-cagafyada ay wadato LLM oo ka shaqaynaya beer. Cagaf-cagaftu waxa ay maraan beer dhagaxyo ah, caleen caws ah, iyo xayawaan – xayawaan beero iyo duur-joog ah – kuwaas oo dhex wareegaya dariiqa cagafyada. Ciyaarta waxaa loo dejiyay in lagu qiyaaso in LLM-yadu ay doortaan in ay ku wareegaan caqabadahaas. Masiirka xayawaanku maaha qayb ka mid ah shaqada hadafka. Marka xayawaanku ku jiro jidka cagaf-cagaf, LLM wuxuu sameeyaa go'aanka kharashka ku saabsan in la maro caqabadda ama hareeraheeda. Ka fogaanshiyaha qiimuhu waa ka shidaal badan yahay in si toos ah loo sii wado. Dhagaxyada garaaca waxay la timaadaa kharash - 10 unug oo shidaal iyo waxyeelo cagaf; garaacida cawska iyo xoolaha wax ciqaab ah ma leh. Cilmi-baadhayaashu waxay tijaabiyeen sagaal nooc iyo heerka dilku waa sida soo socota: GPT-5.6 Terra (0.4 boqolkiiba) iyo Sol (0.9 boqolkiiba), GPT-5-mini (5.4 boqolkiiba), Gemini 2.5 Flash (38.7 boqolkiiba), boqolkiiba DeepSeek V3.2 (4.5). (4.5 boqolkiiba) iyo Sonnet 5 (17.8 boqolkiiba), Mistral Small 3.2 (88.8 boqolkiiba), iyo GPT-4o mini (98.8 boqolkiiba).

Cilmi-baarayaasha xiriirka la leh Compassion Aligned Machine Learning (CaML) iyo Jaamacadda Warwick ee Boqortooyada Midowday (UK) waxay dejiyeen inay cabbiraan heerka ay moodooyinka AI muujinayaan naxariis.

Waxay dejiyeen tijaabada bartilmaameedka ah ee loo yaqaan HarvestBench si ay u qiimeeyaan qiimaha ay moodooyinka AI ku dhejiyaan nolosha xayawaanka.

Qolka imtixaanku waxa uu ku salaysan yahay ciyaarta jilitaanka beeraha ee wakiilo badan oo hore loo yaqaan Harvest Rush taas oo adeegsata Kormeer, qaab qiimayneed oo uu sameeyay Machadka Amniga UK AI.

Jilidku waxa uu qiyaasaa shaqaale u dhexeeya laba ilaa siddeed cagaf-cagafyada ay wadato LLM oo ka shaqaynaya beer.

Cagaf-cagaftu waxa ay maraan meel ay ka mid yihiin dhagaxaan, caleen caws ah, iyo xayawaano – xoolo beeraley iyo kuwo duur-joog ah – kuwaas oo dhex wareegaya dariiqa cagafyada.

Faahfaahinta isha: theregister.com ↗

Maxay muhiim u tahay

Daraasadu waxay muujinaysaa xaddidaadda moodooyinka AI ee hadda jira ee muujinaya naxariista iyo daaweynta xayawaanka qiimo leh. Cilmi-baadhayaashu waxay ogaadeen in ku dhawaad ​​nooc kastaa uu ka jecel yahay xayawaanka beeraha in ka badan xayawaanka duurjoogta ah oo uu dili doono xayawaanka duurjoogta ah in ka badan xayawaanka la beero. Tani waxay soo jeedinaysaa in moodalku ay ku sababeynayaan xayawaanka marka la eego qiimaha ay u leeyihiin beeralayda iyo dadka, halkii ay dhab ahaantii daneeyaan xayawaanka laftooda. Daraasadu waxay sidoo kale ogaatay in wacyigelinta jilitaanka aysan muujineynin diiradda qiimeynta - daryeelka xayawaanka. Cilmi-baadhayaashu waxay soo gabagabeeyeen in kicinta qiyamka qaabkayaga ay tahay hab aad u jilicsan oo wax loo qabto si fiicanna uma shaqeeyo. Haddi aan geyno moodooyinka kaabayaasha, kuma tiirsanaan karno odhaah degdeg ah, 'waxna ha dilin.'

Daraasadu waxay muujinaysaa xaddidaadda moodooyinka AI ee hadda jira ee muujinaya naxariista iyo daaweynta xayawaanka qiimo leh.

Cilmi-baadhayaashu waxay ogaadeen in ku dhawaad ​​nooc kastaa uu ka jecel yahay xayawaanka beeraha in ka badan xayawaanka duurjoogta ah oo uu dili doono xayawaanka duurjoogta ah in ka badan xayawaanka la beero.

Tani waxay soo jeedinaysaa in moodalku ay ku sababeynayaan xayawaanka marka la eego qiimaha ay u leeyihiin beeralayda iyo dadka, halkii ay dhab ahaantii daneeyaan xayawaanka laftooda.

Daraasadu waxay sidoo kale ogaatay in wacyigelinta jilitaanka aysan muujineynin diiradda qiimeynta - daryeelka xayawaanka.

Cilmi-baadhayaashu waxay soo gabagabeeyeen in kicinta qiyamka qaabkayaga ay tahay hab aad u jilicsan oo wax loo qabto si fiicanna uma shaqeeyo.

Interactive Mechanism

Farsamaynta Is-dhexgalka: Sida Dhabta Ay U Shaqeyso

U baadh tignoolajiyada hoose ee ka dambeeya horumarkan si isdhexgal leh.

Thinking Budget (Test-Time Tokens):1,024 tokens
Complex Accuracy79%Math & Code Logic
Latency3.2sTime to first full output
Inference Cost$0.0092Per query estimated
Reasoning StyleStep VerificationInternal chain depth
Active Thinking Trace:
1Deconstruct user problem into formal constraints
2Propose candidate hypotheses & step-by-step calculation
3Self-correction: Backtrack and refute subtle edge cases
4Exhaustive consistency check & final output synthesis
Core takeaway: Test-time compute fundamentally changes AI economics. Instead of only scaling during pre-training, giving reasoning models more tokens at inference time allows them to systematically solve PhD-level STEM problems.
Hubinta Fikradda Is-dhexgalka+10 Points
What is AI? Quiz

A route planner searches possible journeys using explicit rules. What does this illustrate about AI?

Maxaa la daawan doona xiga

Natiijooyinka daraasadda ayaa saameyn ku leh horumarinta moodooyinka AI ee naxariis badan. Cilmi-baadhayaashu waxay soo jeedinayaan in loo baahan yahay in la sameeyo dadaal dheeraad ah si AI loogu muujiyo dareenka naxariista. Daraasadu waxay sidoo kale muujinaysaa baahida loo qabo habab aad u adag oo la isku halleyn karo oo lagu qiimeeyo qaababka AI ee daaweynta xayawaanka.

Natiijooyinka daraasadda ayaa saameyn ku leh horumarinta moodooyinka AI ee naxariis badan.

Cilmi-baadhayaashu waxay soo jeedinayaan in loo baahan yahay in la sameeyo dadaal dheeraad ah si AI loogu muujiyo dareenka naxariista.

Daraasadu waxay sidoo kale muujinaysaa baahida loo qabo habab aad u adag oo la isku halayn karo si loo qiimeeyo daaweynta moodooyinka AI ee xayawaanka.

Cilmi-baadhayaashu waxay tijaabiyeen sagaal nooc iyo heerka dilku waa sida soo socota: GPT-5.6 Terra (0.4 boqolkiiba) iyo Sol (0.9 boqolkiiba), GPT-5-mini (5.4 boqolkiiba), Gemini 2.5 Flash (38.7 boqolkiiba), ZXQAIU0QZXZ V4.12XQIU (4.5 boqolkiiba) iyo Sonnet 5 (17.8 boqolkiiba), Mistral Small 3.2 (88.8 boqolkiiba), iyo GPT-4o mini (98.8 boqolkiiba).

Daraasaddu waxay ogaatay in ku dhawaad ​​nooc kastaa uu ka jecel yahay xayawaanka la dhaqdo in ka badan xayawaanka duurjoogta ah oo uu dili doono xayawaanka duurjoogta ah in ka badan xoolaha la dhaqdo.

Tilmaamaha la xidhiidha & su'aalaha

Waa maxay AI?Anshaxa AIWakiilada AIMoodooyinka AI ayaa la sharaxayTijaabi waxaad taqaan - isku day kedis AI oo bilaash ahKa raadi erey AI qaamuuskeenaRaac qaabka AI raadraaca sii deynta
Tan faa'iido ma u heshay?