A carregar...
Publicação
Comparative Analysis of Generative AI Capabilities: Testing Generative AI Tools for Typical Geospatial Queries
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.07 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Large Language Models (LLMs), such as ChatGPT, demonstrate a strong understanding of human natural language and have been explored and applied in various fields, including reasoning, creative writing, code generation, translation, and information retrieval. Although generative AI has great potential for revolutionizing many industries, it must be understood that its capabilities are only as much as the size of the text base it is trained on, which is – admittedly – typically very large. The goal of this master's thesis is to systematically investigate the ability of a LLM generative AI tool (OpenAI ChatGPT 3.5) to generate correct responses and show insight into the EO/GI and/or GIS&T related, knowledge-based queries. Furthermore, it is to compare the performance of this LLM generative AI tool with that of another similar one, Google Bard in this case. A total of 15 prompt questions, classified into 5 distinct categories were developed into two separate questionnaires (one for each generative AI tool) with the same queries were developed for this study; one corresponding to the assessment of ChatGPT’s performance and the other one for that of Bard’s by GIS industry-academic experts. Responses from both the tools were analyzed and compared, and were considered generally adequate. Bard performed marginally better than ChatGPT, overall. Generative AI tools need further improvements for their generated responses to be considered substantially reliable.
Descrição
Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies
Palavras-chave
LLMs Generative AI OpenAI ChatGPT Google Bard GIS Questions and Answers Performance Assessment Geoinformatics