“Developing Apps with GPT-4 and ChatGPT” book review

It seems that ChatGPT is everywhere today. Sometimes I am afraid to even open up my fridge. However, I feel like there’s still way more hype and worries, than actual knowledge and reasonable use cases. “Developing apps with GPT-4 and ChatGPT” seriously changed my view of the whole LLM landscape.

Disclaimer before we start

For the sake of the readability, I am using words like GPT/LLM/ChatGPT as synonyms. I realize the difference between them (and the same claim the authors make in the text), but as this is just a review, I don’t want to go into much detail. Let’s assume that all those words mean “AI solution that can be used by the users”.

A book about ChatGPT? Seriously?

I want to start my review with a bold statement. Huge KUDOS to O’reilly, for actually publishing a book about such a fast-paced area, which is GPT and LLMs. What is more – this is a second edition already! I realize that we have ebooks today, but still, trying to publish an up-to-date book about ChatGPT and all that jazz – huge KUDOS once more.

However, I will be honest. Sometimes during a lecture, stumbling upon phrases like “at the time of the writing” or “that may be a subject to change” wasn’t that nice. After a while it stops bothering the reader (at least it stopped bothering me), and it seems that it is just how things are. OpenAI solution is just changing all the time, and there is no way to avoid a certain level of “work in progress” vibe there.

At the beginning there was a model

The book starts with a short introduction to the history of LLMs. All the other chapters are filled with heavy technical stuff, therefore starting the book with some history was a nice touch by the authors. I tend to look at LLMs as some kind of rabbit-of-of-the-hat solution that appeared out of nowhere. That’s not true for sure, and there is a long line of predecessors to LLMs as we know them today. I found reading about evolution and the process that led to the current state of things very pleasant. It is not purely historical though – we get technical details of the models too, but without snippets of code. Just an overall view.

Gimme the code!

What I like about this book is its focus on the actual problems that GPT can solve. It is also filled with the actual examples of the prompts/outputs/configs that the authors tested. Upon reading, I got a feeling that authors actually know their stuff, and their hints and good practices are the result of the actual work experience.

They start with the description of the pure OpenAI solution, how to ask questions using GPT-4 or its predecessors. Code samples are given in Python, using the official SDK from OpenAI. We got through all the possibilities for configuring, and tweaking the pure model, in order to serve our purposes. This topic is expanded in much more detail in the following chapter. However, this time we leave the depths of GPT-4 usage, and we get practical information about security, handling of tokens/secrets, cost efficiency and lots of other operational stuff.

To be honest – at that moment, the book could end. The amount of information given is huge, and the reader should have all the tools at hand, in order to use GPT-4 in her application. But we’re not done yet! We go back to the GPT-4 specifics, and we take another deep dive into the prompt engineering, step-by-step reasoning and RAGs. Everything is filled with samples of the code, and actual responses given by the OpenAI product. At the end of the book, the authors provide us with the additional chapter, about the solutions and tools that are building upon the OpenAI product. The reader is left not only with the ChatGPT knowledge, but also with the clear view of the ecosystem that spans from it.

Am I still afraid of opening my fridge?

Maybe not that much. Truth to be told – I am not a fan of using GPT in my everyday work. I mean – using it as a copilot/coding support – sure, why not. However, if somebody asks me if I would like to create the applications using OpenAI tools, I would politely decline. Why? Because the authors of “Developing apps with GPT-4 and ChatGPT” did such a great job in showing me the ropes of working with LLMs. I think that speaks best about my feelings towards this book. I would recommend reading it to every developer out there.