Ku lunaku 24 olw’October eyayita, ku k邀ibwa okuva mu Korean Association for Industrial Intelligence (KOIIA), twetabye mu ‘2025 Academy y’Okukuliza Abakozi Abakulu mu Makampuni aga Wakati, Semina y’Enkola y’Okukola ku Makolero g’Omulembe Ogugenda’, era Ssenkulu Lee Hwan-ho owa CLEVI n’ayogera ebigambo ebikulu.
Twetabye mu sseŋŋano eyitibwa ‘Emisango gy’Okukozesa AI ne Enkola z’Emirimu eza DX’ nga twogera ku mulamwa ogugamba nti ‘Emisango gy’Okukozesa Physical AI, Ensonga ezireeta obutakola bulungi mu kuteeka mu nkola RAG agents, n’Enkola z’Okulongoosa’.
Olw’okuba abantu beeyongera okufaayo ku AI agents gye buvuddeko, ne kampuni ezingi zeeyongera okulowooza ku kuteeka mu nkola AI. Kiteeberezebwa nti eb संस्था n’amakampuni mangi gamaze okugezaako okukozesa AI mu ngeri ey’ekiseera era nga gakola nayo.
Naye eky’ennaku kiri nti okunoonyereza kulaga nti mu makampuni n’ebitongole bingi, ROI gye baali basuubira tefunikiddwa, oba emirimu egyaakolanga mu PoC tegikola bulungi oba tegivaamu n’ebyanukulwa ebituufu mu pulojekiti enkulu oba ku kifo ky’emirimu.
Olw’amawulire gano, amakampuni mangi geegendereza okuteeka AI mu nkola mu bwangu.
Mu kwogera kuno, Ssenkulu Lee Hwan-ho owa CLEVI yannyonnyola mu bujjuvu obuzibu bwa RAG systems, obumu ku nsonga enkulu ezireetera okuteeka AI mu nkola mu bwangu obutakola, awamu ne open-source models ne small LLMs (SLMs), era n’annyonyola lwaki emisango egy’okuteeka AI mu nkola egyasooka gyalemererwa.
Mu nkomerero, ensonga lwaki tukozesa AI eri nti tusuubira okukola emirimu gy’abantu mu kifo kyabwe, oba okulaga omutindo ogusinga ogw’abantu.
Abakozi abakola emirimu mu makampuni agaateeka AI mu nkola kati tebakyasobola kwesiga mutindo gwa byanukulwa bya AI. N’olwekyo, balina okukebera buli kimu AI ky’eyanukula, ekivuddeko okwemulugunya nti emirimu gyabwe gyongeerekeddeko mu kifo ky’okukendeera.
Ensonga esinga obungi eri nti amakampuni mangi gakozesa open-source SLM models nga solution oluvannyuma lw’okuzitendeka nate mu ngeri ya fine-tuning.
Mu ndowooza eyangu, oyinza okulowooza nti AI ejja kufuuka ey’amagezi singa eyiga ebipya, naye bwe oyigiriza AI okumanya okupya, efaayo okufiirwa okwijukira okw’edda, kwe kugamba, ebizitowa bya parameters ebitegekeddwa mu ngeri ey’obukodyo.
Ekizibu ekisinga obukulu kiri nti kati tewali amanyi mawulire oba bumanyirivu ki obubuze.
Olw’okuba okuleeta AI si mulimu ogwetaagisa ob संसimbi obutono, muba mwakola okukebera kungi okuva ku mutendera gwa PoC, si bwe kiri?
Lwaki obuzibu obwo tebwalabika mu kiseera ekyo? Kya lwatu nti kyali olw’okuddamu okutendeka model nga bakozesa amawulire ag’ekigendererwa.
AI eyali emanyi ennamba zokka yamanyisibwa ennukuta, era oluvannyuma n’ebuzibwa ebibuuzo ebikwata ku nnukuta zokka; kale kya bulijjo okuddamu obulungi. Naye mu kifo ky’okukoleramu, twetaaga model ekuguse mu mulimu ogw’enjawulo guno gwokka?
Buli lunaku, ebifo bye tukoleramu bituukirira embeera empya n’amawulire amapya. Naye AI eyatendekebwa ng’ezikiriza okumanya okwakadde mu ngeri ey’ekifuulannenge tesobola kukakasa kukola mu mbeera ez’enjawulo, era n’obutuufu tebusobola kukakasibwa.
Kye kiva kyetaagisa okuba ne From-scratch Foundation large language model, oba tekinologiya asobozesa okukola model ng’eyo.
Kubanga model zino zirina okumanya kumpi kwonna okw’ensi, ekivaamu omutindo gw’okuddamu n’obutuufu obw’ekigero ekya waggulu.
Mu domains ez’enjawulo, osobola okuvvuunuka obuzibu bwa SLM etuned nga okozesa tekinologiya wa SFT(Supervised Fine-Tuning), oba ku data entono okozesa enkola ya RAG oba tekinologiya wa Semantic Database, enkola ya CLEVI eyennyini eyakolebwa ey’obukugu mu database y’obwongo.
Olw’okuba gavumenti nayo emanyi obukulu bwa foundation models zino, ekola kyonna ekisoboka okuwa obuyambi obungi.
Ekyo ye National Representative AI Project. Ekigendererwa ekisinga obukulu mu pulojekiti eno kwe kufuna obwannannyini bwa digito.
CLEVI nayo, olw’okuba yamanya obuzibu obw’amaanyi obuyinza okuva mu kukozesa open-source models, yatendeka models za CLEVI ng’ekola buli kimu ku bwannannyini, okuva ku kukungaanya data okutuuka ku kukola architecture.
Mu kiseera kino, models za CLEVI zirina obusobozi obutali mabega na katono wadde bwe zigerageranyizibwa ne AI ezisinga obulungi mu nsi yonna.
Naddala olw’okuba tulina model zaffe ku bwannannyini, CLEVI tekwatibwako kizibu kya kwekwata ku model z’ebweru ekiyinza okubaawo nga zikozesebwa.
Okuleeta AI kati kufuuka kwa tteeka so si kusalawo kwokka. Naye tekiba kyangu kufuna bivudde mu mulimu bye twagala nga tuleese tekinologiya omupya gwokka.
Obuyiiya obw’amazima butandika ddala nga tumaze okutegeera obulungi ekkomo lya tekinologiya era ne tukola enteekateeka esinga obulungi etuukagana n’embeera n’ebigendererwa by’ekibiina kyaffe.
Nga eyambisa obukugu obw’enjawulo nga large language models zaayo ku bwannannyini ne Semantic Database, CLEVI ekendeeza ku nsobi eziddibwamu n’obulabe amakampuni bye gayitamu mu nkola y’okuleeta AI.
Tugenda kusigala nga tuwuliriza amaloboozi agava ku kisaawe, era tufuba okubeera omukwanaganya wa AI eyeesigika eri amakampuni amangi. Tusaba mukolere wamu ne CLEVI okutuukiriza obuwanguzi obw’amazima mu kussa AI mu nkola.