A Research On The New Generation Artificial Intelligence Technology Generative Pretraining Transformer 3


Aydin N., Erdem O. A.

3rd International Informatics and Software Engineering Conference, IISEC 2022, Ankara, Türkiye, 15 - 16 Aralık 2022 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/iisec56263.2022.9998298
  • Basıldığı Şehir: Ankara
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: Artificial Intelligence, Deep Learning, GPT-3, Natural Language Processing, Transformer
  • Gazi Üniversitesi Adresli: Evet

Özet

In the digitalizing world, Artificial Intelligence (AI) paves the way for the automation of routine work done by humans and makes life easier. Recently, there is no area where AI and its applications are not used in daily life, from health to education, from transportation to energy, and from agriculture to tourism. AI applications are making rapid progress in the direction of important and current developments, especially in Natural Language Processing (NLP) and Deep Learning (DL). A concrete example of progress in these areas is the GPT-3 (Generative Pre-trained Transformer 3) language model. AI-assisted GPT-3 technology is the DL model that is effectively used in many NLP fields, which can produce long and consistent content similar to the texts written by people using pre-trained algorithms. The GPT-3 architecture has reached a level that can compete with people in many areas by producing optimum solutions for all kinds of inputs by using the Transformer-based language model, which is an attention-based deep learning technique. This article aims to convey the efficiency, structure and potential of the Transformer-assisted GPT-3 model, which is one of the most up-to-date NLP technologies, to the reader. Since the number of Turkish studies in the field of GPT-3 AI is quite limited, it is considered that this study will contribute to the literature in terms of both quantity and quality. In addition, the performance parameters of the model were examined by making a customized fine-tuned sample application in the beta version of the GPT-3 model.