Imodeli ye-AI yesizukulwane esilandelayo se-Anthropic u-Claude 3 Opus uthathe isikhundla esiphezulu ebhodini labaphambili le-Chatbot Arena, ephusha i-OpenAI's GPT-4 kweyesibili ehamba phambili.

Selokhu yethulwa ngonyaka odlule, kungokokuqala ukuthi imodeli ye-Claude 3 Opus ibe phezulu ohlwini lwe-Chatbot Arena, olunazo zonke izinhlobo ezintathu ze-Claud 3 ezikleliswe kweziyishumi eziphezulu.

Amamodeli kaClaude 3 enza uphawu

I-LMSYS Chatbot Arena amazinga abonisa ukuthi uClaude 3 Sonnet uthathe isikhundla sesine ngokuhlanganyela neGemini Pro kuyilapho uClaude 3 Haiku, eyethulwe kulo nyaka ikleliswe endaweni yesithupha kanye nenguqulo yangaphambili ye-GPT-4.

Nakuba Claude 3 Haiku ingase ingahlakaniphi njenge-Sonnet noma i-Opus, imodeli iyashesha futhi ishibhile ngokuphawulekayo, nokho “ifana namamodeli amakhudlwana ekuhlolweni okuyimpumputhe,” njengoba imiphumela yenkundla yembula.

“UClaude 3 Haiku ubahlabe umxhwele bonke, waze wafinyelela ezingeni le-GPT-4 ngokuthanda kwethu! Ijubane layo, amandla kanye nobude bengqikithi akufaniswe manje emakethe,” kuchaza i-LMSYS.

Ngokusho kukaTom's Guide, okwenza i-Haiku ihlabe umxhwele kakhulu ukuthi “iyimodeli yosayizi wendawo eqhathaniswa ne-Gemini Nano.” Ingakwazi funda futhi ucubungule ucwaningo oluxube ulwazi amaphepha ngaphansi kwemizuzwana emithathu.

Imodeli ithola imiphumela emihle ngisho nangaphandle kwethriliyoni nesilinganiso sepharamitha ye-Opus noma imaphi amamodeli ekilasi le-GPT-4.

[Isibuyekezo senkundla]

70K+ amavoti amasha e-Arena🗳️ asengenile!

U-Claude-3 Haiku ubahlabe umxhwele bonke, waze wafinyelela ezingeni le-GPT-4 ngokuthanda kwethu! Isivinini sayo, amandla nobude bomongo abufaniswe manje emakethe🔥

Siyakuhalalisela @AnthropicAI ekwethulweni okumangalisayo kwe-Claude-3!

Okujabulisa kakhulu… pic.twitter.com/p1Guuf0B3K

— lmsys.org (@lmsysorg) Mashi 26, 2024

Ingabe lokhu kungaba impumelelo yesikhashana?

Yize iphushelwe esikhundleni sesibili, izinguqulo ze-OpenAI's GPT-4 zisabusa eziyishumi eziphezulu ohlwini ngezinguqulo ezine.

Ngokuvumelana ne Umhlahlandlela kaTom, Izinguqulo ze-OpenAI's GPT-4 ngezindlela zazo ezihlukene zibambe indawo ephezulu “isikhathi eside kangangokuthi noma iyiphi enye imodeli esondela kuma-benchmarks yayo yaziwa njengemodeli ye-GPT-4-class.”

Nge-GPT-5 "ehluke ngokuphawulekayo" elindelwe isikhathi esithile kulo nyaka, i-Anthropic ingase ingasibambeli leso sikhundla isikhathi eside, njengoba igebe kumaphuzu phakathi kwe-Claude 3 Opus ne-GPT-4 lincane.

Yize i-OpenAI ihlale iqinile ekukhishweni kwayo kwangempela GPT-5, imakethe ikulindele kakhulu ukwethulwa kwayo. Kubikwa ukuthi lo modeli ubhekene nokunye “ukuhlolwa kokuphepha okuqinile” nokuhlasela okulingisayo okubalulekile ngaphambi kokukhululwa.

I-LMSYS Chatbot Arena

Lesi simo sincike kumavoti abantu, ngokungafani nezinye izinhlobo zokulinganisa zamamodeli e-AI. Ngalesi, abantu balinganisa okukhiphayo kwamamodeli amabili ahlukene ekwazisweni okufanayo.

I-Chatbot Arena iqhutshwa yi-LMSYS futhi ihlanganisa inqwaba yamamodeli ezilimi ezinkulu (LLMs) alwa nayo “ezimpini ezingahleliwe ezingaziwa.”

Yethulwe okokuqala ngoMeyi odlule futhi iqoqe amavoti angaphezu kuka-400,000 kubasebenzisi abanamamodeli e-AI avela ku-Google, Anthropic kanye I-OpenAI.

“I-LMSYS Chatbot Arena iyinkundla evulekile enemithombo eminingi yama-evals e-LLM. Siqoqe amavoti athandwayo angaphezu kuka-400,000 ukuze siklelise ama-LLM ngohlelo lokukleliswa kwe-Elo,” kusho i-LMSYS.

Isistimu ye-Elo isetshenziswa kakhulu kumageyimu afana ne-chess ukuhlola ikhono elihlobene lomdlali. Kodwa kulokhu, izinga lisetshenziswa ku-chatbot futhi "hhayi umuntu osebenzisa imodeli."

Futhi funda: I-Microsoft Yembula Ama-PC 'Okuqala' Aphezulu Ngenkinobho Ye-Copilot AI

Amaphutha

Izinga le-Chatbot Arena alishodi ngamaphutha. Ngokusho komhlahlandlela ka-Tom, awubandakanyi wonke amamodeli noma izinguqulo zamamodeli afakiwe ngenkathi abasebenzisi ngezinye izikhathi behlangabezana nembi nge-GPT-4 ehluleka ukulayisha. Ingase futhi ithande amanye amamodeli anokufinyelela ku-inthanethi bukhoma, isibonelo i-Google Gemini Pro.

Ngenkathi amanye amamodeli afana nalawo avela ekuqaleni kwe-French AI I-Mistral kanye namafemu aseShayina afana ne-Alibaba muva nje asanda kungena ezindaweni eziphezulu enkundleni ngaphezu kwamamodeli anemithombo evulekile, inkundla isawakhumbula amamodeli asezingeni eliphezulu. Isibonelo, ishoda ngamamodeli afana neGoogle's Gemini Pro 1.5

• I-SEO Powered Content & PR Distribution. Khuliswa Namuhla.
• I-PlatoData.Network Vertical Generative Ai. Zinike Amandla. Finyelela Lapha.
• I-PlatoAiStream. I-Web3 Intelligence. Ulwazi Lukhulisiwe. Finyelela Lapha.
• I-PlatoESG. Ikhabhoni, I-CleanTech, Amandla, Environment, Ilanga, Ukuphathwa Kwemfucuza. Finyelela Lapha.
• I-PlatoHealth. I-Biotech kanye ne-Clinical Trials Intelligence. Finyelela Lapha.
• Source: https://metanews.com/claude-3-opus-takes-top-spot-on-chatbot-rankings/