Utilizing Large Language Models as No-code Interface in a Software Development Toolkit
Keisala, Jukka (2023)
Keisala, Jukka
2023
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2023051410612
https://urn.fi/URN:NBN:fi:amk-2023051410612
Tiivistelmä
Large language models and their applications, such as ChatGPT and Bing AI, have gained much hype and visibility in media in past months (winter 2022–2023). Big companies like Microsoft, OpenAI, Meta and now Google have invested billions of dollars trying to keep themselves in pole position in the technology regarding artificial intelligence and machine learning.
The development of AI has been going on for years, but now the results and applications are available to almost everyone. The usage of large language models is free, with certain restrictions, and models can be used for different use-cases. OpenAI was the first company to monetize their models also for companies that are ready to pay extra to get faster responses and to utilize models in their applications and use-cases.
The assignment comes from the client company that asked to investigate the utilization of large language models to generate code that could be used in the analytics add-on of their IoT-based product.
In the thesis, the theory and methodology is related to large language models are presented, model catalog is presented and models are investigated, dataset is created, proper models are selected, test bench is created, fine-tuning is done, quantitative results are analyzed and conclusions are presented. The test framework, i.e., code-basis and dataset structure, was written during thesis writing and was designed in such way it could be used in further development of current project, but also in similar future projects.
Results achieved were encouraging and will be used in the product, at least as an experimental feature.
The development of AI has been going on for years, but now the results and applications are available to almost everyone. The usage of large language models is free, with certain restrictions, and models can be used for different use-cases. OpenAI was the first company to monetize their models also for companies that are ready to pay extra to get faster responses and to utilize models in their applications and use-cases.
The assignment comes from the client company that asked to investigate the utilization of large language models to generate code that could be used in the analytics add-on of their IoT-based product.
In the thesis, the theory and methodology is related to large language models are presented, model catalog is presented and models are investigated, dataset is created, proper models are selected, test bench is created, fine-tuning is done, quantitative results are analyzed and conclusions are presented. The test framework, i.e., code-basis and dataset structure, was written during thesis writing and was designed in such way it could be used in further development of current project, but also in similar future projects.
Results achieved were encouraging and will be used in the product, at least as an experimental feature.