Multimodalni jezični modeli kao asistent u nastavi informatike

master's thesis

• Title (english): Multimodal language models as assistant in computer science education
• Author: Toni Stipaničev
• Mentor: Davor Horvatić (mentor)
• Committee member: Davor Horvatić (predsjednik povjerenstva)
• Committee member: Karolina Matejak Cvenić (član povjerenstva)
• Committee member: Maro Cvitan (član povjerenstva)
• Committee member: Katarina Jeličić (član povjerenstva)
• Committee member: Petar Žugec (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2024-09-27, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

Neuronske mreže omogućuju računalima da uče iz podataka, prepoznaju obrasce, predviđaju ishode i čak kreiraju sadržaje. One su ključni alat u razvoju dubokog učenja, grane umjetne inteligencije koja nastoji simulirati složenost ljudskog razmišljanja, otvarajući time mogućnosti u mnogim disciplinama. U ovom radu istražujemo kako multimodalni jezični modeli, poput ChatGPT-a i Claudea, mogu simulirati rad nastavnika i pružiti podršku učenicima kroz interaktivno vodeno učenje. Pokazali smo da ovi modeli mogu pomoći u prevladavanju čestih poteškoća koje djeca susreću u procesu učenja. Koristeći ChatGPT i Claude kao primjere, utvrdili smo da oba modela mogu biti adekvatan alat u ponavljanju gradiva i vježbanju. Međutim, važno je naglasiti da ovi modeli, iako korisni, ne mogu i ne smiju zamijeniti ulogu nastavnika. Nastavnici osim prenošenja stručnog znanja, pružaju pedagošku i društvenu podršku koju nijedan AI model ne može replicirati. Stoga, ključno je da se ne zaboravi niti zanemari nezamjenjiva uloga nastavnika u obrazovnom procesu.

Abstract (english)

Neural networks enable computers to learn from data, recognize patterns, predict outcomes, and even generate content. They are a crucial tool in the development of deep learning, a branch of artificial intelligence that aims to simulate the complexity of human thinking, thus opening up possibilities in numerous disciplines. In this paper, we explore how multimodal language models, such as ChatGPT and Claude, can simulate the work of teachers and provide support to students through interactive guided learning. We have demonstrated that these models can assist in overcoming common difficulties children encounter in the learning process. Using ChatGPT and Claude as examples, we have determined that both models can serve as adequate tools for revision and practice. However, it is crucial to emphasize that these models, although useful, cannot and should not replace the role of teachers. In addition to imparting expert knowledge, teachers provide pedagogical and social support that no AI model can replicate. Therefore, it is essential not to forget or neglect the irreplaceable role of teachers in the educational process.

• Keywords: duboko učenje; neuronska mreža; povratna propagacija; vođeno učenje
• Keywords (english): deep learning; neural network; backpropagation; supervised learning
• Language: croatian
• URN:NBN: urn:nbn:hr:217:718724
• Study programme: Title: The Integrated University Undergraduate and Graduate Programme in Physics and Computer Science Education; specializations in: Physics and informatics; Course: Physics and informatics; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: sveučilišni magistar edukacije fizike i informatike (sveučilišni magistar edukacije fizike i informatike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2025-01-30 13:24:47