Live data from Hacker News

ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

scholarturbo.com

51–60 of 70 posts

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#51

How far do we think this product is from something that use uses langchain to parse pdfs, create & store chunks / embeddings, and feed them to ChatGPT (or a local model)? Do we think there's any secret sauce?

This is probably it +/- implementation details

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#52

How far do we think this product is from something that use uses langchain to parse pdfs, create & store chunks / embeddings, and feed them to ChatGPT (or a local model)? Do we think there's any secret sauce?

I can say as someone who made something similar (Docalysis.com) that there's a lot of "secret sauce" in actually getting the outputs to be higher quality, especially with processing larger files. Just today I got another email from a user saying the quality was better than other AI products they tried.

I also don't use Langchain.

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#53
This is similar to wordpress websites or android apps. Iphone store had a high toll bar, $99/year .

There will be a lot these quick to build interesting utilities to show off. There is room to feed the demand in wordpress sites and ad-word based websites.

I think OpenAI will keep the platform cheap long enough to collect the small toll and make the innovation go wild.

There is perhaps room for one more OpenAI style provider, like what Android did to iOS to feed this demand. It is a pity if its not Google.

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#54
I want to try this, but it's stuck at processing, showing 100% but not usable.

I‘ve tried a lot of chat-with-pdf apps, including ChatDOC, ChatPDF, Humata and others. For serious paper reading, I must say that ChatDOC(https://chatdoc.com/) is the best option.

The accuracy of ChatDOC’s understanding of tables, data, and texts is significantly higher than other products. Also you can upload a folder of files and chat with the collection, easily navigating through each one.

For example: https://chatdoc.com/chatdoc/#/share/ub8Z7aarcXJHY_SzGwl3LFVw...

I uploaded a paper about GPT and asked ChatDOC to summarize the main content, explain the experimental data, and find relevant data. It completed all these tasks very well. Amazing.

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#56

How do we know ChatGPT isn't hallucinating parts of its answers? I think this is an interesting service but what are the guardrails so we can trust the answer?

That's the problem, we don't. All the tools on top of GPT that I tried has the same problem.

That's a big part of the reason I built the open source tool Semantra[1]. I only really trust the semantic search part of this problem as it is constrained by context, so I tried to make the user experience good on top of simpler semantic search without any chatbot capabilities.

[1] https://github.com/freedmand/semantra

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#57

I want to try this, but it's stuck at processing, showing 100% but not usable. I‘ve tried a lot of chat-with-pdf apps, including ChatDOC, ChatPDF, Humata and others. For serious paper reading, I must say that ChatDOC( https://chatdoc.com/ ) is the best option. The accuracy of ChatDOC’s understanding of tables, data, and texts is significantly higher than other products. Also you can upload a folder of files and chat…

ChatDOC’s document processing ability is excellent. I can preview the document online and find the corresponding original content for the answer, but the other app does not have this feature yet

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#58

This is similar to wordpress websites or android apps. Iphone store had a high toll bar, $99/year . There will be a lot these quick to build interesting utilities to show off. There is room to feed the demand in wordpress sites and ad-word based websites. I think OpenAI will keep the platform cheap long enough to collect the small toll and make the innovation go wild. There is perhaps room for one more OpenAI style p…

True. I can't believe how simple OpenAI made their API to be. They really made it trivial to start using their API. Comparing that to Google's, words fail me to even describe the frustration with GCP just to start testing it's text-bison.

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#59

I want to try this, but it's stuck at processing, showing 100% but not usable. I‘ve tried a lot of chat-with-pdf apps, including ChatDOC, ChatPDF, Humata and others. For serious paper reading, I must say that ChatDOC( https://chatdoc.com/ ) is the best option. The accuracy of ChatDOC’s understanding of tables, data, and texts is significantly higher than other products. Also you can upload a folder of files and chat…

same problem with this... can't upload my PDF... Thanks for the recommendation! I Just tried ChatDOC and I really like that the answers list all the page numbers of the cited sources. Cool! I'll continue to try out the multi-file chat feature you mentioned. Hope it can save me from a pile of files!

Re: ScholarTurbo: Use ChatGPT to chat with PDFs (supports GPT-4)

#60

In answer to file security: "We utilize Amazon Web Services (AWS) for uploading and storing files, ensuring robust security measures, including data encryption and access control." Well, sure, but that's the file at rest. What about the security / privacy of the data as it's fed to ChatGPT? More and more of these "ask your doc / data a question" apps are popping up. Isn't this about equivalent to putting the document…

I look at these tools to see how they communicate these data processing/security details to their users.

There appears to be a lot of apathy/total disregard to giving users any clear clue.

It might not matter to me or this audience - we probably know the direction our data is heading - but it matters to less techie people, to ordinary employees and to civil servants and government people at the bottom and the top, who have no notion where their data is processed and who they are trusting with it.

Post reply on HN