I may be in the wrong area to ask this question, so i apologize in advance. But i would like to build an assistant to help me in writing a book. I have zero programing knowledge. Additionally i have a large volume on research papers I would like to incorporate. These would be files in PDF, PPTx, and audio/video data of me presenting and talking to the topic for the book.
Can anyone assist me with lamens guidance on how to build this assistant using OpenAI?
I don’t know for sure, but it sounds like what you want to do is exactly what custom GPTs were created to do.
Just click the + sign indicating you want to create a new GPT. You will be guided through the process. You describe what you want the assistant to do, and you will be able to upload the files you need.
Hi and welcome to the community!
The advice to start with building a custom GPT is solid in it’s own right, especially if you have no coding experience.
But I think we should also moderate expectations a little bit. Especially when you mention that lots of research papers will be involved in writing your book. This tells me that your expectations regarding the final product are considerably high.
Custom GPTs are very young, standardized tools and it’s really likely that standard results may not fit your individual use case. Ultimately you will end up building and developing features that expand on the standard functionality.
I suggest you start small with distinct work packages and implement those partially to learn about the results and compare those to what you will need later at full scale. And you set aside some time to deepen your understanding and learn about these tools.
You are in a trade-off situation where AI will help you convert and retrieve knowledge and also support you with writing but it’s the development know-how that will actually leverage these new tools to get the results you are looking for.
I hope this helps in assessing the current situation.
Hi, I am familiar with OpenAI api integration and I can help you out.
Can we discuss more about your project?
I been working on that with GPT for months. It’s doesn’t work. Doesn’t matter how many different promts you try to use you will hit OpenAI policy biases anyway.
I learned it hard way. After thousand attempts I can spot OpenAI characteristics on each output if you give GPT space for any creativity.
If you give theme, instructions on style, link to knowledge base on that theme and subthemes, give examples, concepts and needed message to deliver, method’s and so on… - you could get output of decent text with some kind of OpenAI biased characteristics. But it’s basically you already writing book yourself and by creating such profiles on each chapter or subchapter you will spend more time then you can write it yourself. And if you an expert on the theme, you will do better then GPT.
If you want to make expert GPT , firstly you will use much more time - I mean much more time - and still most likely you won’t succeed completely, secondary it will still be biased by OpenAI characteristics.
You can feed GPT data, use GPT to hint you , give new ideas and help you become expert. But it will never do better on complex tasks, not with current version and it’s capabilities is far from that. You can’t build GPT to be expert in any complex area, not yet.
Where GPT is very useful is to edit you text of your book. Essentially you can rephrase it in any style GPT knows (with in OpenAI policies). You can also ask GPT to elaborate on any part of your text to hint you ideas on what you can write more.
As OpenAI is only making their policies more strictly applied it’s even getting worse to train GPT to be an expert as it’s more sticking to it’s “frame” . No matter how good instructions the output will always have ChatGPT “identity” .
I’d be happy to discuss. I have multiple asset types, like PDF files, powerpoints with visuals, along with podcast recordings (with transcription) that I want to feed into the model. Then use this data to create the chapters and context for the book. I’d like it to write in my style from the data inputed.
What’s the best way to connect?
thank you for this. I generally agree. Hopefully we can get there soon.