Viewing a single comment thread. View all comments

josep-panadero t1_j3h4m2w wrote

At this point ChatGPT seems to be quite good at getting answers to this kind of questions. What it does not provide is a personal assessment like "this book is inspiring" or "it was fun to read".

ChatGPT:

>There isn't a single reference that covers the entire history of neural networks in great detail, but there are a number of good books that you might find helpful. Here are a few that you might want to check out:
>
>"Deep Learning" by Ian Goodfellow, Yoshua Bengio, and Aaron Courville: This book provides a comprehensive introduction to deep learning, including a historical overview of the field. It covers the key ideas that have driven the development of neural networks from the early days up to the current deep learning era, and provides information on their main contributions and inspirations.
>
>"Neural Networks: A Comprehensive Foundation" by Simon S. Haykin: This book is considered a classic reference in the field of neural networks. It provides a detailed historical overview of the field, starting with the early perceptron models and moving through the development of more sophisticated architectures such as convolutional neural networks and recurrent neural networks.
>
>"The Deep Learning Revolution" by Terrence J. Sejnowski: This book provides a broad overview of the history of neural networks and deep learning, and discusses how these technologies have been applied to a wide range of applications, including image and speech recognition, natural language processing, and more. It also provides a glimpse into the future of deep learning and how it is likely to continue to impact a variety of fields.
>
>I hope these references are helpful! Let me know if you have any other questions.

So, I asked: Is "The Deep Learning Revolution" fun to read?

ChatGPT:

>"The Deep Learning Revolution" by Terrence J. Sejnowski is an engaging and informative read for anyone interested in the history and future of deep learning. It provides a broad overview of the field and discusses how deep learning has been applied to a wide range of applications, including image and speech recognition, natural language processing, and more. Sejnowski writes in an accessible style and does a good job of explaining complex concepts in a way that is easy to understand. Overall, I think most people would find it an enjoyable and educational read.

I still think that ChatGPT does not have a proper opinion, as it could think that a book is easy to read but be inapprehensible for a human. But the answer may be useful nonetheless. It, at least, can be used as a starting point to further investigate.

ChatGPT can be used to find answers just like Google is used to find an answer in StackOverflow.

I foresee that new rules may be needed in Reddit to what to do with ChatGPT copy/paste, like Stackoverflow already did.

−7

fredlafrite t1_j3hit01 wrote

Here it is accurate, but for more technical questions chatGPT very often invents references!

6

peter201943 t1_j3hfad0 wrote

I'll make a counter argument. If you are going to compare ChatGPT to Google, then why do you feel it is appropriate to directly paste the results of ChatGPT on Reddit?

I ask because Google can have factually incorrect search results that must then be evaluated by whoever is performing the search. By posting the raw output, without applying any of your own critical perception onto the suggestions, you are merely dilluting the level of intelligence in the conversation.

So on Reddit, a forum for humans, I might remind you, why would we post the results of an automatic query of whatever kind, that then requires effort to determine if the results of the query are actually useable themselves?

I think it's ok for you or anyone else to use ChatGPT to inform your answer, the same way its ok to use Google or Wikipedia. And yes, mentioning that your answer came from ChatGPT is nicer than just pasting from GPT directly.

Another suggestion, if you have read enough ChatGPT responses, you'll notice a pattern in them, that they are full of filler or have a hard time just getting to the point. This filler is not desirable for a quick response (such as on a forum, or in this case, getting a list of references). Compare the length of your comment to the other comments here. Do you see how much longer your comment is?

Lastly, I know that there are a lot of easily Google-able questions on Reddit, but for something subjective, such as this forum thread in particular, then the kind of information being sought is not objective or measurable. The OP is asking for opinions. Let's assume they've already Googled "Books on History of Artificial Neural Networks". Since ChatGPT is trained on the same dataset as what Google accesses (the Internet), there is no benefit in copy-pasting the output of ChatGPT. It does not have novel opinions, it simply aggregates known existing ones.

I look forward to your introspection.

4

gbfar OP t1_j3hjdpa wrote

Complementing your answer...

I've actually used ChatGPT before posting this thread, and the answers were all unsatisfying, just like in the comment above yours. Actually looking into the resources recommended by ChatGPT will quickly show that most of them simply do not meet the criteria I specified in my post. The only useful reference is the book by Haykin, but it looks like the historical references are somewhat scattered throughout the book, which makes it not so easy to read.

Also, I've not come here after just opinions. I've come here hoping for informed opinions from experienced researchers, who may have already read and evaluated many of the references on NN history that I'm asking for. This is something ChatGPT is very far from being able to provide.

5