What Is ChatGPT Doing ... and Why Does It Work?

Full of theory and you would love the statistics. Nothing about the chatgpt prompts here. Learn about neural nets or neural networks.

This was recommended to me by my colleague, who is working with me on a project involving AI and ChatGPT. I loved the simplistic explanations in the first part of the book and the comparisons between the (inaccurate, but plausible) responses provided by ChatGPT and Wolfram Alpha. That really grounds you in understanding how one should use ChatGPT for general purposes.

I purchased hardcopy thinking it'd be of better quality and have colored print. Highly disappointed ☹️.

The book is more of an attempt to show how much the author understands the subject, not really to teach the subject matter. The diagrams which are key to drive home the point are small, skewed and in black and white (even when color coded). It is just a hurried attempt to cash in on the hype. It does not even motivate the subject. It is more like an executive summary.

When I mentioned to my Machine Learning friend that I'm reading Wofram's book to understand ChatGPT, he kinda laughed because Wolfram is not a machine learning scientist. However, for such a short book the ROI of reading it seemed high. For me, this provided a nice bridge because I have a background in Math and Data Analytics, but don't have much hands-on experience with neural networks. This short and inexpensive book provided a good overview of LLM limitations and gave me something to think about in the future.

This is one of the most poorly written books I've ever read; for someone who works on language tools, Wolfram seems to have little to show for it. When you see that one paragraph in three begins with "OK, so", you already know that you'll have to brace yourself.The writing style is quite bad and confusing, with a lot of parentheses and a copious use of quotes for common words or obvious analogies. There is no clear structure, and the topic suddenly changes from one paragraph to the next without any proper transition. Conjunctions seem to be randomly used, and I won't even talk about the punctuation. It's almost like reading the transcript of a young student struggling to explain something newly learned to another.For a topic that isn't trivial, adding so much confusion is a fatal mistake. Fortunately, I know enough about neural networks, and I have some basic insight into LLMs, or I wouldn't have been able to read past the first chapter.The content is rather disappointing. When I ordered this book, I was hoping to get a better insight into the LLM-based engines and why they worked so well, perhaps also what they do well, how they could evolve, or how they shouldn't be used. Unfortunately, the author doesn't really answer the question in the title and doesn't attempt to address the other related topics. Most of the time, the answer is "we don't know why it works or why we do it like that", which might be partly true but makes me wonder if that's the right person to write such a book (or if it was worth writing in the first place).The author doesn't miss an opportunity to talk about Wolfram tools, though. Is that advertising in disguise, seizing the opportunity created by the current AI hype? Why, instead of presenting some obscure tool screenshot, hasn't he used simple diagrams or more commonly used tools? The charts, where the colours needed to interpret them have been greyed out, are also quite a letdown.I've seen relatively short videos that did a much better job of explaining GPT's inner workings and discussing their limitations or possible evolution. For example, search for videos made by Judie Burchell or Jeremy Howard. There's simply no upside to this book that I can think of.Don't bother.

Assez technique mais passionnant pour les aficionados !Avec en prime des réflexions sur ce que tout ça nous enseigne du fctnnmt du cerveau.

This book is written by a person who has been working on AI and has been somewhat surprised by the success of Chatgpt. It is certainly worth reading, but a certain level of expertise on computation theory is necessary to understand it fully. Interesting to see the limitations of Chatgpt and LLM models. I think it is very good, but the title is a little misleading (that is why I do not give 5 *)

Gained a better understanding of LLM’s potential and limitations.