Issuu

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 11 Issue: 02 | Feb 2024

p-ISSN: 2395-0072

www.irjet.net

Opportunities in Research for Generative Artificial Intelligence (GenAI), Challenges and Future Direction: A Study J. D. Chavan1, C. R. Mankar2, Dr. V. M. Patil3 1 Student, 2 Research Scholar, 3 Research Guide 1,2,3 Department of Computer Science, Shri Shivaji college, Akola

-----------------------------------------------------------***---------------------------------------------------Abstract: The use of Generative Artificial Intelligence as a tool is growing tremendously, as GenAI helps users to generate results and apply them in various areas like research, new images, etc. GenAI has received attendance from users working in different areas like artists, researchers, and academicians where they found a tool that gives them the desired result to achieve their goal, during their work they came across a point where they found no way to go ahead. Here GenAI helped by guiding them to complete the work. The objective of AIGC is to expedite the generation of high-quality content by streamlining and simplifying the process of content creation. AIGC is accomplished by taking human instructions and deriving intent information from them, then using that information to generate content based on its expertise. The paper organization includes the introduction to GenAI, history, types of GenAI, its application and challenges, revie of literatures and future direction. Keywords: GenAI, GAN, VAE, Machine Learning, Deep Learning.

Introduction:

The GenAI is not new for all but the term was available publicly in 2022 as a chatbot called ChatGPT. The working ChatGPT is based on the learning AI model which uses a provided dataset from which the designed algorithm works, understands, classifies, and then from the appropriate learning gives the result. The GenAI is modelled such that it adopts the feature of self-learning from the dataset, also nowadays various deep learning models like VAEs, and GANs are seen to be used. GenAI gives creativity, innovation, and the ability to solve problems with new content. ●

II.

What is GenAI: AIGC is accomplished by taking human instructions, deriving meaning from them, and using that aimed information to create content depending on its knowledge and understanding. Large-scale models have grown in significance in AIGC in recent years because they enable superior intent extraction and thus, better generation outcomes. The distribution that the model can learn becomes more extensive and true to reality as data and model sizes increase, resulting in the creation of higher-quality and more realistic content. This survey offers a thorough analysis of the development of generative models over time, as well as an overview of their fundamental elements and current developments in AIGC from unimodal to multimodal interaction. We provide the generation tasks and related text and image models from the standpoint of unimodality. History of AI and Generative: Generative AI, also known as generative modelling, is a branch of artificial intelligence (AI) focused on creating models capable of generating new data that is similar to a given dataset. This field has a rich history spanning several decades, with significant advancements made in recent years due to developments in deep learning and neural networks. Below is a detailed overview of the history of generative AI: The 2010s witnessed significant breakthroughs in generative AI, largely driven by advancements in deep learning. Variational autoencoders (VAEs), introduced by Kingma and Welling in 2013, provided a probabilistic framework for learning latent data representations. Generative adversarial networks (GANs), proposed by Ian Goodfellow et al. in 2014, introduced a novel approach to generative modelling based on adversarial training. GANs consist of two neural networks, a generator and a discriminator, trained simultaneously in a minimax game framework, where the generator learns to generate realistic data while the discriminator learns to distinguish between real and generated data. GANs have demonstrated remarkable success in generating high-quality images, audio, text, and other types of data, leading to widespread applications in art generation, image synthesis, and data augmentation

Impact Factor value: 8.226

ISO 9001:2008 Certified Journal

Page 446