Timestamps: Part 1: 0:00 Summary & introduction 0:55 Importing content & knowledge graph representation 2:06 💡Slicing off the top layers of ideas 4:03 💡 Finding blind spots in a discourse 6:38 ChatGPT o1 model use with InfraNodus 7:44 LLM chain for ChatGPT for finding hidden topics and gaps 8:23 ❗An investment advice from InfraNodus! 9:37 💡Using this approach without any apps or tools Part 2: 10:37 📙Using this approach in science and paper analysis 12:08 ❗Topical diversity: optimal modularity measure! 13:24 Iterative approach to “slicing off” ideas from the graph 15:07 Making notes and writing down ideas 16:12 Removing the whole topic 17:22 💡 Using ChatGPT o1 model to advance your scientific research 19:15 Slicing off the surface ideas in InfraNodus extension 20:08 Summary and advantages of this approach
Hi DMitry, great function I didn't know about to completely delete a cluster with a single click. And perfect, too, this new feature for requesting GPT o1 for more scientific questions, agentic, and exploiting the extension on top of it in an iterative way.
Thank you - as always, Dmitry Paranyushkin! Every of your posts starts a journey .... But please, can you post in the European morning, not in the night? Your post forces me :-) now to stay away from bed ...
Thank you for the compliment, Johannes :) My statistical analysis of our channel tells me most of the eyes are watching the videos around 9pm CET, so I tried to go with its advice. But I'll be happy to post another video in the morning to help my European viewers go to bed early. By the way, I have a short tutorial on using InfraNodus to improve deep sleep, which is available here: noduslabs.com/featured/how-to-increase-the-duration-of-deep-sleep-with-daily-activities/
For me, I actually like that they hallucinate because I use them to generate new ideas that don't exist yet based on research and existing content. So I hope there will be an option to keep them hallucinating :)
Thanks for the quick update. I was trying to use the tool to digest the RUclips content such as this video, but so far, it keeps reporting errors ..is a still RUclips link enough for the resource input ?
@@vitamin7179 Ok, i mean you could take the Reddit's RSS and analyze it using InfraNodus. It would work the same: you choose a Reddit, analyze it, find patterns within. Pretty useful stuff. What would be your particular use case so I can recommend you the best workflow?
I am also wondering how to generate insight from a university library database search ..appearly the gap from an individual paper is not enough to support the rationale of a new research...
@@hongmeixie409 yes another approach is go beyond the periphery of the graph. I’ll record another video on it. How is your university database search results available? Which format?
@noduslabs Thanks for a quick response. The results are not much difficult if you are familiar with Web of Science, Scopus etc.. But another way to generate a graph is from the citation manager tool such as Zotero, which normally contains a lot detailed information including a full text PDF
This software would be a heck of a lot cooler if it was local. I'd love to use it on queries against my vectordb, but can't upload that data to your service for privacy reasons.
@@manslaughterinc.9135 why not just use the private graph function where nothing is saved in the db but you just process the requests to get the graph?
@@noduslabs what are you smoking, it’s a 720p video and youtube can go for 4k. Even at a given resolution you can have different levels of quality like via bit rates etc. Go and evaluate this in your program
@@Sheeshening what makes you think that it’s ok to use this tone of voice with somebody you don’t know? Would you talk like this to me in person? Why is that ok to do that online?
@@noduslabs @Sheeshening That tone of voice is also not suited to be used with someone you do know. That being said, he is correct about the resolution, and maybe his tone of voice came as a response to the phrase "The video itself is as high res as you can get." ..which is not accurate and also sounds over confident.
@@johannesdeboeck well that’s a very good point :) all I can say is that the video you are seeing is indeed not the biggest possible resolution you can theoretically get these days, but it should be sufficient to follow the workflow and understand the ideas I’m trying to convey. At the same time, there’s always room to improve.
Thanks for the quick update. I was trying to use the tool to digest the RUclips content such as this video, but so far, it keeps reporting errors ..is a still RUclips link enough for the resource input ?
@@hongmeixie409 you can still do RUclips keyword research but the subtitles are sometimes glitching. We are working to improve this solution and I expect to have results by tomorrow.
Timestamps:
Part 1:
0:00 Summary & introduction
0:55 Importing content & knowledge graph representation
2:06 💡Slicing off the top layers of ideas
4:03 💡 Finding blind spots in a discourse
6:38 ChatGPT o1 model use with InfraNodus
7:44 LLM chain for ChatGPT for finding hidden topics and gaps
8:23 ❗An investment advice from InfraNodus!
9:37 💡Using this approach without any apps or tools
Part 2:
10:37 📙Using this approach in science and paper analysis
12:08 ❗Topical diversity: optimal modularity measure!
13:24 Iterative approach to “slicing off” ideas from the graph
15:07 Making notes and writing down ideas
16:12 Removing the whole topic
17:22 💡 Using ChatGPT o1 model to advance your scientific research
19:15 Slicing off the surface ideas in InfraNodus extension
20:08 Summary and advantages of this approach
Hi DMitry, great function I didn't know about to completely delete a cluster with a single click. And perfect, too, this new feature for requesting GPT o1 for more scientific questions, agentic, and exploiting the extension on top of it in an iterative way.
@@Mind-mapping-decision great to hear you like it! Yes I’m also adding an ability to select and filter out only the visible topic soon!
Brilliant, Dmitry!
@@kennethtaylor5225 thank you! Hope you can use it in your workflow!
At it again I see! Great work
Fire 🔥🔥🔥
welldone
Thank you - as always, Dmitry Paranyushkin! Every of your posts starts a journey .... But please, can you post in the European morning, not in the night? Your post forces me :-) now to stay away from bed ...
Thank you for the compliment, Johannes :) My statistical analysis of our channel tells me most of the eyes are watching the videos around 9pm CET, so I tried to go with its advice. But I'll be happy to post another video in the morning to help my European viewers go to bed early. By the way, I have a short tutorial on using InfraNodus to improve deep sleep, which is available here: noduslabs.com/featured/how-to-increase-the-duration-of-deep-sleep-with-daily-activities/
I hope this was said in jest. Because the entitlement is shocking otherwise.
when those models hallucination index reach 0 this tool will be awesome!
For me, I actually like that they hallucinate because I use them to generate new ideas that don't exist yet based on research and existing content. So I hope there will be an option to keep them hallucinating :)
liked and subscribed.
@@losbones thank you for your support!
@@noduslabs if anyone ever criticized your content please know that for me this was 100% A++. Articulate and informative knowledge sharing.
@@losbones thank you! it's great to hear!
Thanks for the quick update. I was trying to use the tool to digest the RUclips content such as this video, but so far, it keeps reporting errors ..is a still RUclips link enough for the resource input ?
Good news: we fixed the RUclips subtitles import, you can now use it!
Thank you for the tips!🖖 Is it possible to analyse the whole Reddit website with it?
@@vitamin7179 why would you want to analyze the whole Reddit?
@@noduslabs Something similar like gummysearch does but in a bit different perspective
@@vitamin7179 Ok, i mean you could take the Reddit's RSS and analyze it using InfraNodus. It would work the same: you choose a Reddit, analyze it, find patterns within. Pretty useful stuff.
What would be your particular use case so I can recommend you the best workflow?
I am also wondering how to generate insight from a university library database search ..appearly the gap from an individual paper is not enough to support the rationale of a new research...
@@hongmeixie409 yes another approach is go beyond the periphery of the graph. I’ll record another video on it. How is your university database search results available? Which format?
@noduslabs Thanks for a quick response. The results are not much difficult if you are familiar with Web of Science, Scopus etc.. But another way to generate a graph is from the citation manager tool such as Zotero, which normally contains a lot detailed information including a full text PDF
This software would be a heck of a lot cooler if it was local. I'd love to use it on queries against my vectordb, but can't upload that data to your service for privacy reasons.
@@manslaughterinc.9135 why not just use the private graph function where nothing is saved in the db but you just process the requests to get the graph?
Typo on the URL
Thank you! Fixed! 🙏🏼
low res video
@@Sheeshening just change your RUclips setting to 1080 at the bottom left. The video itself is as high res as you can get.
@@noduslabs what are you smoking, it’s a 720p video and youtube can go for 4k. Even at a given resolution you can have different levels of quality like via bit rates etc.
Go and evaluate this in your program
@@Sheeshening what makes you think that it’s ok to use this tone of voice with somebody you don’t know? Would you talk like this to me in person? Why is that ok to do that online?
@@noduslabs @Sheeshening
That tone of voice is also not suited to be used with someone you do know.
That being said, he is correct about the resolution, and maybe his tone of voice came as a response to the phrase "The video itself is as high res as you can get." ..which is not accurate and also sounds over confident.
@@johannesdeboeck well that’s a very good point :) all I can say is that the video you are seeing is indeed not the biggest possible resolution you can theoretically get these days, but it should be sufficient to follow the workflow and understand the ideas I’m trying to convey. At the same time, there’s always room to improve.
Taylor George Lee Dorothy Harris James
Jack Mark Mary Anne Ahmed
Thanks for the quick update. I was trying to use the tool to digest the RUclips content such as this video, but so far, it keeps reporting errors ..is a still RUclips link enough for the resource input ?
@@hongmeixie409 you can still do RUclips keyword research but the subtitles are sometimes glitching. We are working to improve this solution and I expect to have results by tomorrow.
Good news: we fixed the RUclips import, now you can use it.