Generative AI (GenAI) uses machine learning (ML) algorithms to create seemingly new content that can be in the shape of text, images, audio, video, or programming code. GenAI is a revolutionary technology within the field of artificial intelligence that empowers machines to world of GenAI, starting from its definition, historical timeline, and game-changing applications in different domains. First, it goes into the fundamental principles of ML, probabilistic modeling, and the difference between generative and discriminative models. Looking through some of the leading generative models—Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), diffusion models, and autoregressive models—reveals the variety of techniques that have been harnessed to fuel the advances. The generation of realistic examples by generative models is one of the major achievements in the field of AI (Dash et al. in IEEE Access 12:18330–18357, 2024 [1]; https://www.sciencedirect.com/science/article/pii/S209526352400147X#bbib41 [2]; https://medium.com/@nnabwani/full-fine-tuning-peft-prompt-engineering-or-rag-69689ba58a52 [3]). One type of GenAI is found in large language models (LLMs), which use a deep neural network architecture known as the Transformer to capture patterns and information from web-scale training data in a self-supervised manner. These models have achieved human-level quality in generating text, answering questions, and language translation. These abilities have led to a surge of interest in the field and new services. As GenAI continues to evolve and mature, such models will likely become increasingly integrated into people’s daily lives (Yadav et al. in Int J Multidiscip Res, 2024 [4]). The chapter goes on to highlight applications of GenAI across diverse sectors—from content creation in media and entertainment to data augmentation and innovation in scientific research. We discuss the challenges (e.g., ethical issues, computational requirements, evaluation difficulties) while emphasizing the strengths and weaknesses of the technology. The market for GenAI models, including LLMs, is growing rapidly with over 100 commercial and open-source offerings with different licensing models, leading to hype but also indicating quick obsolescence of current implementations. Legal and commercial risks can be mitigated by selecting the right models and ensuring the right contractual terms and their application (Kılınç and Keçecioğlu in Acad Platform J Eng Smart Syst, 2024 [5]).

Fundamentals of Generative AI

章节
发表日期：

2025年

查看全文

Krishna Kumar

摘要

Generative AI (GenAI) uses machine learning (ML) algorithms to create seemingly new content that can be in the shape of text, images, audio, video, or programming code. GenAI is a revolutionary technology within the field of artificial intelligence that empowers machines to world of GenAI, starting from its definition, historical timeline, and game-changing applications in different domains. First, it goes into the fundamental principles of ML, probabilistic modeling, and the difference between generative and discriminative models. Looking through some of the leading generative models—Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), diffusion models, and autoregressive models—reveals the variety of techniques that have been harnessed to fuel the advances. The generation of realistic examples by generative models is one of the major achievements in the field of AI (Dash et al. in IEEE Access 12:18330–18357, 2024 [1]; https://www.sciencedirect.com/science/article/pii/S209526352400147X#bbib41 [2]; https://medium.com/@nnabwani/full-fine-tuning-peft-prompt-engineering-or-rag-69689ba58a52 [3]). One type of GenAI is found in large language models (LLMs), which use a deep neural network architecture known as the Transformer to capture patterns and information from web-scale training data in a self-supervised manner. These models have achieved human-level quality in generating text, answering questions, and language translation. These abilities have led to a surge of interest in the field and new services. As GenAI continues to evolve and mature, such models will likely become increasingly integrated into people’s daily lives (Yadav et al. in Int J Multidiscip Res, 2024 [4]). The chapter goes on to highlight applications of GenAI across diverse sectors—from content creation in media and entertainment to data augmentation and innovation in scientific research. We discuss the challenges (e.g., ethical issues, computational requirements, evaluation difficulties) while emphasizing the strengths and weaknesses of the technology. The market for GenAI models, including LLMs, is growing rapidly with over 100 commercial and open-source offerings with different licensing models, leading to hype but also indicating quick obsolescence of current implementations. Legal and commercial risks can be mitigated by selecting the right models and ensuring the right contractual terms and their application (Kılınç and Keçecioğlu in Acad Platform J Eng Smart Syst, 2024 [5]).