Waa Maxay Artificial Intelligence ? Baro AI Af Soomaali | AI #1

Kani waa muuqaalkii ugu horreeyay ee taxanaha AI ee Dugsiiye, waxaana loogu talagalay qof kasta oo doonaya inuu aasaaska AI-ga si sax ah u fahmo isagoo aan hore wax badan u aqoon.

Waxaan ka billaabayaa qeexidda: artificial intelligence waa caqli macmal ah oo lagu ekaysiiyay caqliga bini'aadamka. Kadib waxaan kala saarayaa saddexda heer: Narrow AI oo ah meesha hadda aan joogno, AGI, iyo super intelligence oo aanan aaminsanayn inay dhici doonto. Waxaan idinla marayaa taariikhda AI-ga marxalad marxalad. Rule-book era-dii 1950-kii oo wax walba xeer ku qornaa oo edge cases-ka aan la soo koobi karin. Expert system era-dii 1980-kii oo aqoonta khubarada mishiinka la geliyay, sida MYCIN iyo XCON. Machine learning revolution-kii 2000-ka oo spam-ka email-ka lagu bartay pattern-ka. Iyo deep learning-kii 2010-ka oo sawirrada layer layer loogu jajabiyo edges, shapes ilaa aqoonsi.

Kadib waxaan gelayaa Generative AI iyo Large Language Models: farqiga u dhexeeya model gaar ah oo hal shaqo qabta iyo hal model oo shaqooyin badan qabta. Waxaan sharraxayaa attention mechanism-ka anigoo tusaale u soo qaadanaya miis toban qof ku fadhiyaan. Ugu dambayn waxaan qaadayaa supervised, unsupervised iyo reinforcement learning, iyo limitations-ka hadda jira sida hallucinations, context limit-ka iyo training cut-off-ka.

The first video in Dugsiiye's Somali-language (Af-Soomaali) AI series, a 21-minute foundation lesson for anyone who wants to understand artificial intelligence properly. It defines AI, separates narrow AI from AGI and super intelligence, and then walks the whole history: the 1950s rule-book era and why hand-written rules could not handle real-world edge cases, the 1980s expert systems such as MYCIN and XCON, the machine learning revolution of the 2000s explained through email spam detection, and the deep learning era where images are broken into edges, shapes and features. It then covers generative AI and large language models, the shift from one model per task to one model for many tasks, and the attention mechanism. It closes with supervised, unsupervised and reinforcement learning and today's limits: hallucinations, context limits and training cut-offs. Free on YouTube.