Developing Apps with GPT-4 and ChatGPT: Build Intelligent Chatbots, Content Generators, and More (Development of applications based on GPT-4 ChatGPT) (AI column)

On the second day of vacation, I read a short book. 2023 year on creating applications using GPT (Generative pre-trained transformers)which is great for beginners or those who have missed the last couple of years of hype relative to LLM (large language models). This book is as simple and practical as possible, there are only five chapters in it.

1) GPT review--4 ChatGPT Here the authors talk in a few words about NLP. (natural language processing)The appearance of transformers after the key article.Attention Is All You Need" from the guys at Google, and then about the development of this idea from OpenAI and the emergence of a series of models GPT-X, where X is from one to four. Next, the authors show the use cases of LLM (from Khan Academy, Duolingo, Morgan Stanley, Waymark). And the chapter ends with a story about AI hallucinations and the possibility of their tuning for tasks. (fine tuning) 2) Deep dive into GPT--4 ChatGPT API - here the authors talk about the API for using OpenAI and it is about obtaining an access key, using a playground or library. Then they go on to tell us what the models are. (It's obsolete.)They talk about the API format and show how to use it from python code, and then they talk about the pricing scheme and the security of models. 3) Creating applications with GPT-4 ChatGPT This chapter is about application design and the authors talk a little bit about the principles where we should analyze input and output to protect against attacks of the species. prompt InjectionAnd then there is a demonstration of simple applications.

• News generator. Short Sammari video from YouTube based on transcripts A video game assistant who can answer questions about lore and game directories Voice control with a simple set of functions 4) Advanced GPT technologies4 ChatGPT - here the authors talk about prompt engineering, or rather about how to write effective prompts, how to ask LLM to move step by step along the process, implement few-shot learning with examples in the prompt, as well as how it is even more difficult to design a prompt to solve the problem. The second half of the chapter is devoted to fine tunning OpenAI models, which allow you to finish the model on your data, but greatly raises the cost of using the API. Demonstration of fine tuning model is made on the example of creating an application for generating emails for email marketing:) 5) Expanding LLM capabilities with LangChaing Framework and plugins This chapter is perhaps the most interesting and dedicated to the framework. LangChainIt allows you to create powerful applications. Interestingly, the same chapter describes plugins that were later replaced by GPTs, custom versions of chatGPT that anyone can make for specific tasks. And in the beginning 2024 year appeared GPTs Store for these models. But LangChain has since pumped even harder:)

In general, this book provides basic insights into building applications using LLM, but the field itself is evolving so fast that a year-old book already seems an anachronism:)

#AI #ML #DataScience #Data #Engineering