Subscribe to never miss a video: www.youtube.com/@noduslabs?sub_confirmation=1 Timecodes: 0:00 Introduction 0:28 "Generate a new idea" app 1:20 How to process AI content 2:12 Using the knowledge graph to generate AI ideas 3:50 Make your own AI query 4:28 Elaborating an AI-generated idea 6:02 A big difference from using ChatGPT 7:23 "Human in the loop" AI 9:24 How to check if you're missing anything from your ideation process 10:30 Using the graph as a map to navigate your thinking process 15:42 Reveal non-obvious: slice off the knowns to get to the unknowns 18:11 Finding structural gaps in your thinking
Awesome app, I just wish I had the idea to use it for, it will come, now I've seen this! My first thought was this would be great for Biochemitry and specifically protein or genetics research, linking molecular structure to function: if we know a certain conformation has a specific action, then we can predict a novel molecular structure will have a similar function.
this app reminds me of collocation dictionary from Oxford. You should develop this app for learning languages. It can structure similar words, adjectives and antonyms. By clicking button you can create sentences with this words. Which is Amazing for language learners.
Amazing! Nothing compares to getting a visual representation of a complex problem/solution! Love how you can expand your thinking in each node and go deep or wide into a topic. It almost feels like a 3D representation of our thought process, or that’s what it is…
Thank you! Yes, I also love graphs. In fact, the reason why I started this project was first to just visualise the abstract structure of thought, so the first version didn't even have the node names. It was really interesting to see how different types of texts (e.g. poetry vs science) have a completely different structure (dispersed for poetry and highly connected for science). It's like poetry leaves the space for our mind to fill in (with emotion?) while something that is coherent and meaningful is about connections and structured relationships. Then it evolved into InfraNodus and I still kept some of those structure analysis indices inside the tool (if you switch to the Advanced Mode).
@@noduslabs that's awesome! Yes, it's amazing how the mind creates these buckets for us (or others) to fill. NLP is really good at teaching how to entice the creation (when used in a positive way) of buckets and to empower the person to fill them by themselves. It would be pretty cool to convert them into decision trees and assign probabilities of outcomes and other statistical/predictive tools. Imagine the day when you can connect some type of EEG in your brain, create a map, and determine areas of improvement, enhancement, and/or healing. Pretty cool! Would love to connect and chat.
This is the most beautiful and relevant thing I've seen on TH-cam. This is actually perfect for what I want to do. I am a Sales professional of 15 years and I am in the processes of developing what I call a silver bullet. There are a lot of variables that I need to consider which to my peers appear unrelated but I can feel that there is a correlation between them. Just can figure it out on my own. Usually what happens is that I get lost or buried or just plain bored after like 2 hours of work at a time, and when I go back, I have to start from scratch. This tool you made is going to be a big help. Thank you so much.
I'll wait for @a.bukhari to answer, but I guess it helps him see the patterns emerging in the relations between the different variables he has to consider?
Incredible! I am deeply grateful how Ai and people like you are creating tools that help us explore our creativity in ways were impossible before. Thank you!
This is fantastic - possibly my soon-to-be favorite tool! I’ve garnered so much insight through various networked information graphs. I’m so excited that you’ve made this tool and I look forward to diving in!
this is just a superhuman way of working , i always wanted obsidian to be able to do this one day. i like to look for relationships , this is great tbh
I am actually using Obsidian to try to replicate this sort of setup. That from my web of knowledge, I may inspire myself of new ideas from two or more seemingly unrelated ideas. I really like what I see here from this app! :D
Great, happy you find it interesting. There's so much more to using the graphs to explore what's missing. Not only the gaps, but also exploring the periphery, linking it to other graphs, etc. Working on all of those implementations now and will soon post more videos on the topic!
I love seeing some of the first limbs of the AI tree grown. I can’t imagine what turns it’ll take next, but this one’s amazing. Thank you so much great explanation and I look forward to more examples even all the way to B2B stuff.
Thank you! Yes, in fact for organizations it can be also very useful especially in marketing, patent study, research, and knowledge base development. I have more videos on this channel about these topics.
Amazing! Definitely unlocking the potential of LLMs right here. Best service on the market. Will be using this for my thesis workflow and idea mapping.
THIS IS EXACTLY WHAT I WANTED TO SEE. I'm such a noob and cant code for the life of me, but i always get ideas and solutions to things, then find it materialise in reality somewhere across the world.
Thanks! Well, I can totally imagine. That's why I preferred to just program the whole thing with the good old code. If language is the new programming language thanks to the AI models, then I prefer the good old Javascript for tasks like this :) Would be curious to see your experiments though. Did you make any customGPT with it?
You don't realize it it but you've helped me get that much closer to solving the hard problem of simulated consciousness, and self-awareness in my model.
Absolutely, my work is based hard coded survival base, python, C, multi-agent, on a few theories of mind and following the footsteps and expansion of Dr. Stephen Thaler, from Imagination Engines Inc., with local models and the use of large models intertwined with deep learning focused on panpsychism consciousness (in a super nutshell), will post once I have it up and running. @@noduslabs
I recommend you allow a trial use of this. Most, including me, will not pay for something they have not tested. I don't want to subscribe for a month just to find out I don't like it, which means I won't subscribe for a month to find out I love it either.
@@sdas683no it doesn't. Having to give your credit card and agreeing to a contract to auto charge you if you forget to cancel in two weeks, is not a free trial. It's a predatory sales tactic.
Congratulations for such an amazing tool! It is incredibly interesting how you thought on the gaps between ideas and the fact of elaborating on a concept with and without considering the entire context. Fascinating, I will try it out!
Very cool product. I'm working on something very similar but it is being developed as an assistive technology Agent for students with learning-related disabilities. The user interface is crucial and should be nonlinear and multimodal to be effective because humans are multisensory and our brains and languages are non-linear. However, it's important to design in a way that avoids sensory overload.
Exactly, in my experience, it's kind of a tight rope between usability and flexibility. I like to keep it close to non-linear network view, but lots of users find it too complex especially at the first sight. But that's getting better as network analogy is getting more and more ubiquitous. What's your experience with your students?
@@noduslabs I will soon find out. I am still at an early stage of development but will be releasing the MVP before the end of October to a small cohort of college students. For me, the initial focus is on building student confidence by enabling them to discover as many aha moments as possible early in the learning journey.
Thanks for sharing There is one suggestion from my side. You can get more number of clicks on your video by changing Thumbnail . Use ChatGPT icon + blury question mark icon = Picture of Knowledge Graph Keep the background of thumbnail black.
This is an exemplary demonstration of InfraNodus's capabilities, leveraging GPT-4 for dynamic knowledge graph construction. I wonder whether the node metadata and GPT-4-generated content could be used for model fine-tuning. Specifically, by extension of your tool, could you backpropagate this structured data into the GPT-4 architecture to optimize its performance for specialized research domains like your HRV study? The objective being to enhance the model's ability to generate interconnected and domain-specific responses, thereby accelerating the discovery of novel correlations and insights within the knowledge graph. Possibly a little "out there" and potentially more expensive to get going, but thought I'd ask :)
This is a very good point and I definitely thought of it, so it could be very interesting to find somebody who also knows about those models and to collaborate on this topic, for sure!
@@noduslabsMost people's attempts to fine-tune models based on body's of text has been an abysmal failure (including mine). I think that this would be far more likely to succeed using your data to the train another model where the parameters/weights/etc are actually available to work with (like LLaMa). Either way, if there was access to the raw graph data, I would become a customer instantly- as there are plenty of other ML uses for these (and this could basically be used as a way to export topical semantic relationships that GPT-4 is able to generate).
@@sophiophile yes, you can access the raw graph data, but how would you import it into the model? what format would you need? feel free to respond here but we can also move it to the email if you send me a note via support.noduslabs.com
Love it. Thank you, this is a game changer. I can't believe I only discovered your channel now. Thank you again for the excellent presentation and for all your work.
i think this program is great especially when learning onto new topic or trying to find new concepts or connections between known knowledge. Maybe this tool can become the base of education tool for students and researchers in the future. Inventing this tool and bringing it into reality is wonderful, nice job bro, it was fantastic !!
i've been looking into this concept theoretically. and voila you've developed an app in a similar vector space. BTW, I'm just 36 seconds in. I don't know if this is what I'm thinking it is. But I'm excited to watch more.
Wow, I have been looking for ChatGPT visual support guides for information like this! I am so excited to learn more how I will incorporate this program in my learning and professional Career goals! Thank you so much for your technical contributions and sharing your thoughts and ideas! This think of thinking seems so revolutionary to a Millennial like myself.
You really presented this well and convincing. I did look at the pricing and once again another site that requires subscriptions and it is quite expensive. So you have to really justify consistent usage of this tool.
Well, yes, we have to charge for it because we're not venture backed so we can't offer free subscriptions as we have to pay the costs of developing and supporting this tool and all the infrastructure. So this is just realistic. If you prefer to pay a one-off fee that you find fair, that's fine with us too, you can make your offer and we can see if we can make it work!
Definitely showcase more use cases for end user because those videos will blow up and yeah... you basically can do anything knowledge based but users wants simplicity have 2 modes advanced and beginner.
This is Crazy innovative I have been planning to create something super identical to it for modelling my own knowledge related to AI as an application of Knowledge Graphs and LLM Seeing this just made me so much inspired!!!
Yes, a lot of it was coded by myself, but not everything. The current version is actually Sigma.Js for visualization and the good old jQuery with some React. The new version (available in the extension) uses D3 and React plus some custom libraries.
This was very impressive. Science was Birthed from the existence of ignorance. Math gives Vision to the Science, AI replies to query with very reasonable feedback. But this offers something much more powerful than them all. It appears to Stimulate informed Questions which can be exposing the Mind to some of the Most Powerful and Productive discoveries we may not have otherwise Known about for Decades. I would like to see a reporting abilty based on any of the selected bullets in a hierarchical format of Headers, Sub-Headers and bullet points. I wouldn't put it past you at this point that is already available in some way even if it is within the ChatGPT chat itself. Just sharing my thoughts as I enjoyed your presentation. Looking forward to using this tool. P.S. Does it explore Code Development in a simular way if you wanted to develop with it ?
Great, thank you for your wonderful feedback! We're working on integrating bullet-based syntax, so hopefully you can try it soon. Coding could also be very interesting, but I'm wondering what type of data would be interesting to represent as a graph? Functions and the relations between them, maybe? What do you think?
@@noduslabs I'm working on AI generated pictures for a project and i would love to see the metadata of the nodes confronted in a graph, it would be much easier to work on and probably you could be more creative being able to see many variables and connections. Awesome work btw
@@noduslabsfor software achitects like a really old banking system with tons of interindependent (mispell ik) software to simplify nodes ofcourse you would want simply to throw at it zip files with source code and it should graph every project and it's for example let's say endpoints gpt 4 code interpreter can do that halfway but you need to run python linux env to achieve what i've mentioned but hey there you will have banks money so idk. Just thoughts. Thanks for reading
Exactly. I also like using it to stimulate the creative process and for learning rather than having it do everything for me. In the latter case, the result is too generic and I kind of lose the joy of the process.
Just a reminder: infranodus.com - 14 day trial, so you can subscribe and then cancel and not pay anything and try it for 2 weeks. Here, we're giving out a hack on our own subscription system :)
@@noduslabs yes, he was! Please serve yourself. :) First time I learned about him, was on a numberphile video like 12y ago, he talked about the responsability of math teachers into inspiring love for mathematics as a world of discovery and creation for students, he's a great guy
You are a Genius. God, thank you for your work. Merging AI intelligence with human thinking will greatly impact society. Especially visual learning has a great bandwidth.
Thank you. I am interested in several matters. Let's start with the fact that I am a beginner in AI programming and I see its potential. I am in the phase of learning about AI's capabilities and limitations. My questions are: 1. Can I use Infranodus to train AI for different purposes - an RPG dialogue system? 2. How can I make my AI trained with Infranodus public - a website where a user could log in and use the trained AI? Or integrate it with MSG and Facebook? 3. What do I need to know? 4. How much does using Infranodus cost?
Pretty cool project. I think if I was to signup I would want to be able to use my own openai api key rather than paying a price that seems to include queries.
Yes, we will look into this for sure! But the cost is also justified by the fact that we're not a venture-backed company, so we're just charging what it costs to develop.
I've watched 5 minutes of it, not going to lie - this is exactly what I have imagined “supposed to be available” as a tool, when I have played games in aughties. It still is super confusing to me why people don't develop things that are cool, but rather that are...
So good had to leave another comment. We want this for exploring blockchain regulation and writing new regulatory theories for more efficient policies. Finance regulation is a quagmire of rhetoric and precedents. Sometimes the precedents don’t line up between the states and across the world. We know every time the US finance regulators sneeze the world catches the cold. Plus so many opinion letters, and updates in the law since 2013. Some states get it right others don’t, would love to plot all the laws for blockchain tech on the books so far across all 50 states and take all the potential finances laws related to article 9 of the UCC and SEC act among others, to make new theories. Thanks for breaking down how to synthesize complex explorations and identity correlations.
Great idea! I think where InfraNodus can be helpful is in identifying the main topics that emerge in regulation and then finding connections between them as well as the gaps (possible legal gaps that could be both exploited but also bridged depending on your objective). It can’t yet visualize a lot of data at once (the limit is our ability to get meaningful insights from large graphs as humans) but we could try to do that through an API. You would still use a standard RAG LLM to search for specific parts of the document.
I am working on a crowd simulator and what was and probably still is the problem for me - is the fact that my original and even few next understandings of the problem was far away from the actual complexity. And GPT didn't helped me by spammin my already melted brain with more information that yet needed a very god restructuring. So this is an interesting tool as it firstly shows you the scope of things that you will have to deal with filtering away the contradicting details.
I like how the visualization is, and works in your program. I would say the only thing that's missing is a text to speech workflow and graph level of beginner moderate and advanced. I also feel hearing the words that could be highlighted or prompted to say could help guide creative thinking more in rhythm and patterns. Tailoring it to your own thinking patterns vs thinking how to tailor it should be the objective, overall i really like it !
🎯 Key Takeaways for quick navigation: 00:00 🧠 The video demonstrates how to combine data visualization and GPT-4 AI to develop ideas iteratively. 00:26 💡 InfraNodus is introduced as the tool for this demonstration, allowing you to input your topic of interest. 01:19 📚 You can gather information from the AI, save interesting points, and start building your own thoughts around the topic. 04:26 👩🔬 GPT-4 AI can provide scientific explanations, and users can request elaboration to deepen their understanding. 06:12 🗂️ The knowledge graph allows you to visualize and organize your ideas, making it easier to identify connections. 09:11 🕵️ Analyzing high-level ideas in the graph can help you identify the core themes of your discourse. 14:54 🌐 You can explore seemingly unrelated topics by finding gaps in the graph and generating questions to connect them. 16:16 🧩 Highlighting gaps in the graph can prompt the AI to generate ideas that bridge these gaps. 18:32 🤔 Crafting well-phrased questions can lead to more interesting outcomes when connecting unrelated topics. 20:16 🔄 Using the current context to generate questions can help maintain coherence in your discourse.
thank you for the transcription. you guys do a service that is literally invaluable. Not even Chat GPT is at that level of understanding and summarizing.
@@noduslabs ironically, neither, as a first pass. I would like to just be able associate relevant nodal points with supporting imagery (like you’ve done with Notes), so that the groupings and milestone nodes could be become richer and more valuable. Example, diving deep into HRV could be accompanied by a found graph of data - decline of HRV over time, or anatomical images of the cardiovascular in good health and bad. After a while this becomes a pictorial doorway into the hypergraph where tapping on an image of interest takes you to the relevant vertex in the model. From there, image interpretation could become useful, as you express in your presentation, to present previously unexpected topics to then unwrap and explore.
@@noduslabsindeed. And that is not in any way a criticism of what you’ve created here. I think it’s just incredible. I just could see the potential of image supplementation in my minds eye.
This sounds like a cool idea for forming a film, adding more to it than a lineral story, but character background, location explanations, etc. If I understand this correctly I really want to do a film that lets you choce at a couple key points, have a feeling this might help figure where and then help in those branches and subsequent endings
Do you use the graph in Obsidian? To me it's pretty, but I couldn't get any use from it. I'm surprised they didn't go further with it, especially that InfraNodus was available since 2014, so it's not like what we're doing is rocket science (although it is years of work and experimentation and research and building upon the work of others, notably - Gephi).
With ur knowledge of HRV, you were able to identify the factual in accuracy what is the user does not have the same level of background knowledge you do in order to be able to identify these inaccuracies?
Good point! I think at this point we always need to verify it somewhere else, but I think very soon we'll have the tools that will allow us to do it automatically.
Can you explain the graph a bit more. Some questions I have: 1. What's the size of the dots indicate? 2. What do the colors indicate? 3. How'd you determine which nodes are connected to each other in the word cloud?
1. The size of the nodes (words) indicates their relative influence in the discourse, calculated using betweenness centrality measure from network science. 2. The colors indicate that the nodes (words) belong to the same topical cluster of words (concepts that tend to be used together in this text). Calculated using community detection algorithm based on modularity measures of the text network. 3. The connections are identified based on the words co-occurrence using n-grams (4 word windows).
I had a similar idea combining chatGPT and graphs from the Obsidian, but you seem to have something more nuanced. I'm a Computer Science student with 8 classes left. Can I work for you in this project? It is something I am extremely passionate about! It would change my life for the better.
Hello and thank you for your offer! Sure, you can send us some of your ideas and code samples via support.noduslabs.com and we'll be happy to see if there's a space for collaboration.
A fascinating tool. I wonder, however, if it would be possible to add references GPT-4 uses when returning a prompt? In my field it gets things wrongs and i find myself having to verify. I actually prefer it that way but having the references makes it a bit easier
GPT-4 doesn't really return references and even when it does it usually makes them up. That's why I prefer using it to stimulate imagination but not to produce hard facts. For facts we have other modes in InfraNodus, such as using Google Scholar or scientific libraries as the source of truth.
![[1hr Talk] Intro to Large Language Models](http://i.ytimg.com/vi/zjkBMFhNj_g/mqdefault.jpg)
Subscribe to never miss a video: www.youtube.com/@noduslabs?sub_confirmation=1
Timecodes:
0:00 Introduction
0:28 "Generate a new idea" app
1:20 How to process AI content
2:12 Using the knowledge graph to generate AI ideas
3:50 Make your own AI query
4:28 Elaborating an AI-generated idea
6:02 A big difference from using ChatGPT
7:23 "Human in the loop" AI
9:24 How to check if you're missing anything from your ideation process
10:30 Using the graph as a map to navigate your thinking process
15:42 Reveal non-obvious: slice off the knowns to get to the unknowns
18:11 Finding structural gaps in your thinking
Looking forward to this video
That looks awesome, would love to use it! I wonder though what's privacy and data security like with this? Sorry, I'm completely uninformed.
Awesome app, I just wish I had the idea to use it for, it will come, now I've seen this! My first thought was this would be great for Biochemitry and specifically protein or genetics research, linking molecular structure to function: if we know a certain conformation has a specific action, then we can predict a novel molecular structure will have a similar function.
This is how you do marketing correctly. I never felt like I was watching an ad. You can only pull this off when you have a good product to start with
this app reminds me of collocation dictionary from Oxford. You should develop this app for learning languages. It can structure similar words, adjectives and antonyms. By clicking button you can create sentences with this words. Which is Amazing for language learners.
Amazing! Nothing compares to getting a visual representation of a complex problem/solution! Love how you can expand your thinking in each node and go deep or wide into a topic. It almost feels like a 3D representation of our thought process, or that’s what it is…
Thank you! Yes, I also love graphs. In fact, the reason why I started this project was first to just visualise the abstract structure of thought, so the first version didn't even have the node names. It was really interesting to see how different types of texts (e.g. poetry vs science) have a completely different structure (dispersed for poetry and highly connected for science). It's like poetry leaves the space for our mind to fill in (with emotion?) while something that is coherent and meaningful is about connections and structured relationships. Then it evolved into InfraNodus and I still kept some of those structure analysis indices inside the tool (if you switch to the Advanced Mode).
@@noduslabs that's awesome! Yes, it's amazing how the mind creates these buckets for us (or others) to fill. NLP is really good at teaching how to entice the creation (when used in a positive way) of buckets and to empower the person to fill them by themselves. It would be pretty cool to convert them into decision trees and assign probabilities of outcomes and other statistical/predictive tools. Imagine the day when you can connect some type of EEG in your brain, create a map, and determine areas of improvement, enhancement, and/or healing. Pretty cool! Would love to connect and chat.
@@noduslabs is it easy and possible to take a text and analyze it such that it is translated into concept maps that can then be further analyzed?
@@talibalhidayasure, that's exactly what the tool does!
This is the most beautiful and relevant thing I've seen on TH-cam. This is actually perfect for what I want to do. I am a Sales professional of 15 years and I am in the processes of developing what I call a silver bullet. There are a lot of variables that I need to consider which to my peers appear unrelated but I can feel that there is a correlation between them. Just can figure it out on my own. Usually what happens is that I get lost or buried or just plain bored after like 2 hours of work at a time, and when I go back, I have to start from scratch. This tool you made is going to be a big help. Thank you so much.
Hey, just curious, how does it help you for sales?
I'll wait for @a.bukhari to answer, but I guess it helps him see the patterns emerging in the relations between the different variables he has to consider?
yessssssssssss
Please answer the question: how would it help sales?
@@CiscoZero in many ways!
Incredible! I am deeply grateful how Ai and people like you are creating tools that help us explore our creativity in ways were impossible before. Thank you!
This is fantastic - possibly my soon-to-be favorite tool! I’ve garnered so much insight through various networked information graphs. I’m so excited that you’ve made this tool and I look forward to diving in!
Thank you!
this is just a superhuman way of working , i always wanted obsidian to be able to do this one day. i like to look for relationships , this is great tbh
I am actually using Obsidian to try to replicate this sort of setup. That from my web of knowledge, I may inspire myself of new ideas from two or more seemingly unrelated ideas.
I really like what I see here from this app! :D
Great, happy you find it interesting. There's so much more to using the graphs to explore what's missing. Not only the gaps, but also exploring the periphery, linking it to other graphs, etc. Working on all of those implementations now and will soon post more videos on the topic!
@@jasmeetbrar8609God bless you in that endeavour.
I love seeing some of the first limbs of the AI tree grown. I can’t imagine what turns it’ll take next, but this one’s amazing. Thank you so much great explanation and I look forward to more examples even all the way to B2B stuff.
Thank you! Yes, in fact for organizations it can be also very useful especially in marketing, patent study, research, and knowledge base development. I have more videos on this channel about these topics.
Amazing! Definitely unlocking the potential of LLMs right here. Best service on the market. Will be using this for my thesis workflow and idea mapping.
THIS IS EXACTLY WHAT I WANTED TO SEE. I'm such a noob and cant code for the life of me, but i always get ideas and solutions to things, then find it materialise in reality somewhere across the world.
but i want this method to be implemented into everything, from training to memory recall.
@@abdullahsulaymaan9085I guess that would work! like you mean using it with your notes, for instance, right?
@@noduslabs yeah, totally. it could also work with audio too right? and images. u can group them into location taken, content shown/recorded etc etc.
@@abdullahsulaymaan9085 Look into Obsidian as a note taking tool! very similar in scope
YOU ARE A GENIUS!!!
This is incredible mind mapping tool!
I mean just the visualizer is a work of art!! the rest is just pure magic!! Thanks
This is so cool! You have no idea how many plugins and prompt engineering hours I spent on ChatGPT to produce a resultant knowledge graph like this!
Thanks! Well, I can totally imagine. That's why I preferred to just program the whole thing with the good old code. If language is the new programming language thanks to the AI models, then I prefer the good old Javascript for tasks like this :) Would be curious to see your experiments though. Did you make any customGPT with it?
TH-cam recommendations obviously taking me for someone smart.
So true 😅
It’s been months since being impressed with LLM integrations. This is refreshing.
You don't realize it it but you've helped me get that much closer to solving the hard problem of simulated consciousness, and self-awareness in my model.
Great! Would be curious to hear more!
Absolutely, my work is based hard coded survival base, python, C, multi-agent, on a few theories of mind and following the footsteps and expansion of Dr. Stephen Thaler, from Imagination Engines Inc., with local models and the use of large models intertwined with deep learning focused on panpsychism consciousness (in a super nutshell), will post once I have it up and running. @@noduslabs
do tell
I signed up for this tool like 7 years ago and loved it but it never stuck. Awesome to see this latest evolution and excited to dive in!
Great! You must be one of the oldest subscribers then!
I recommend you allow a trial use of this. Most, including me, will not pay for something they have not tested. I don't want to subscribe for a month just to find out I don't like it, which means I won't subscribe for a month to find out I love it either.
It comes with a 14 Day free trial 😉
But one have to give your payment details..
@@sdas683no it doesn't. Having to give your credit card and agreeing to a contract to auto charge you if you forget to cancel in two weeks, is not a free trial. It's a predatory sales tactic.
If you don't know if you can afford chat gpt a month you have other things to worry about...
@JohannFritz78 wow crazy cause that's standard practice so you don't keep making new accounts. Grow up and cancel before your trial ends.
Congratulations for such an amazing tool! It is incredibly interesting how you thought on the gaps between ideas and the fact of elaborating on a concept with and without considering the entire context. Fascinating, I will try it out!
Very cool product. I'm working on something very similar but it is being developed as an assistive technology Agent for students with learning-related disabilities. The user interface is crucial and should be nonlinear and multimodal to be effective because humans are multisensory and our brains and languages are non-linear. However, it's important to design in a way that avoids sensory overload.
Exactly, in my experience, it's kind of a tight rope between usability and flexibility. I like to keep it close to non-linear network view, but lots of users find it too complex especially at the first sight. But that's getting better as network analogy is getting more and more ubiquitous. What's your experience with your students?
@@noduslabs I will soon find out. I am still at an early stage of development but will be releasing the MVP before the end of October to a small cohort of college students. For me, the initial focus is on building student confidence by enabling them to discover as many aha moments as possible early in the learning journey.
@@Aryankingz Great! Will be happy if you'd like to share your prototype for some feedback.
can you share your github link?
Thanks for sharing
There is one suggestion from my side.
You can get more number of clicks on your video by changing Thumbnail .
Use ChatGPT icon + blury question mark icon = Picture of Knowledge Graph
Keep the background of thumbnail black.
Thank you for the idea! So much space for improvement but I'm too busy coding the tool and researching, so didn't yet nail the blogging game fully...
@@noduslabsYe do that first finish the product (1year) then start marketing it.
This is an exemplary demonstration of InfraNodus's capabilities, leveraging GPT-4 for dynamic knowledge graph construction. I wonder whether the node metadata and GPT-4-generated content could be used for model fine-tuning. Specifically, by extension of your tool, could you backpropagate this structured data into the GPT-4 architecture to optimize its performance for specialized research domains like your HRV study? The objective being to enhance the model's ability to generate interconnected and domain-specific responses, thereby accelerating the discovery of novel correlations and insights within the knowledge graph. Possibly a little "out there" and potentially more expensive to get going, but thought I'd ask :)
This is a very good point and I definitely thought of it, so it could be very interesting to find somebody who also knows about those models and to collaborate on this topic, for sure!
@@noduslabsMost people's attempts to fine-tune models based on body's of text has been an abysmal failure (including mine). I think that this would be far more likely to succeed using your data to the train another model where the parameters/weights/etc are actually available to work with (like LLaMa). Either way, if there was access to the raw graph data, I would become a customer instantly- as there are plenty of other ML uses for these (and this could basically be used as a way to export topical semantic relationships that GPT-4 is able to generate).
@@sophiophileI've also thought about this!
@@sophiophile yes, you can access the raw graph data, but how would you import it into the model? what format would you need? feel free to respond here but we can also move it to the email if you send me a note via support.noduslabs.com
I'm really happy to see this amazing project and the philosophy behind it - I truly love it.
Love it. Thank you, this is a game changer. I can't believe I only discovered your channel now. Thank you again for the excellent presentation and for all your work.
Thank you for your kind words!
i think this program is great especially when learning onto new topic or trying to find new concepts or connections between known knowledge.
Maybe this tool can become the base of education tool for students and researchers in the future.
Inventing this tool and bringing it into reality is wonderful, nice job bro, it was fantastic !!
In my 2nd year of Software Engineering at the moment and this is really inspiring to see. Thank you for this!
Great to hear that!
i've been looking into this concept theoretically. and voila you've developed an app in a similar vector space. BTW, I'm just 36 seconds in. I don't know if this is what I'm thinking it is. But I'm excited to watch more.
This will revolutionise brainstorming in different industries
Wow, I have been looking for ChatGPT visual support guides for information like this! I am so excited to learn more how I will incorporate this program in my learning and professional Career goals!
Thank you so much for your technical contributions and sharing your thoughts and ideas!
This think of thinking seems so revolutionary to a Millennial like myself.
Mind blowing, have to dig deeper into this tool and approach
This is exactly how my brain feels inside and trying this will help me build out my thoughts, I appreciate you!
Same here!
You really presented this well and convincing. I did look at the pricing and once again another site that requires subscriptions and it is quite expensive. So you have to really justify consistent usage of this tool.
Well, yes, we have to charge for it because we're not venture backed so we can't offer free subscriptions as we have to pay the costs of developing and supporting this tool and all the infrastructure. So this is just realistic. If you prefer to pay a one-off fee that you find fair, that's fine with us too, you can make your offer and we can see if we can make it work!
This is awesome just signed up and found great youtube content again for the first time! this fixes alot of issues!
An open-source version of this would be really interesting in terms of developers being able to progress your great work.
We do have an API.
Definitely showcase more use cases for end user because those videos will blow up and yeah... you basically can do anything knowledge based but users wants simplicity have 2 modes advanced and beginner.
We do have two modes. Have you tried the app?
Ye I did but from the phone😅sorry@@noduslabs
Wow the future is here! This is awesome I'm making this part of my process immediately! :) Thank you!
This is Crazy innovative
I have been planning to create something super identical to it for modelling my own knowledge related to AI as an application of Knowledge Graphs and LLM
Seeing this just made me so much inspired!!!
You made me subscribe in 1 minute 😂
I immediately noticed that this channel is great
Are you still subscribed after 2 months? :)
Did you code this yourself ? I’m guessing three js maybe for the actual view ? Cool stuff man
Yes, a lot of it was coded by myself, but not everything. The current version is actually Sigma.Js for visualization and the good old jQuery with some React. The new version (available in the extension) uses D3 and React plus some custom libraries.
Great to see the progress on this! Well done.
Astonishing content Man 💯💯 Thanks to share your knowledge with us 🥳 Thumbs Up from Brazil !!! 👍👍👍 (plus +1 subs)
This was very impressive. Science was Birthed from the existence of ignorance.
Math gives Vision to the Science, AI replies to query with very reasonable feedback.
But this offers something much more powerful than them all. It appears to Stimulate informed Questions which can be exposing the Mind to some of the Most Powerful and Productive discoveries we may not have otherwise Known about for Decades.
I would like to see a reporting abilty based on any of the selected bullets in a hierarchical format of Headers, Sub-Headers and bullet points. I wouldn't put it past you at this point that is already available in some way even if it is within the ChatGPT chat itself. Just sharing my thoughts as I enjoyed your presentation. Looking forward to using this tool.
P.S.
Does it explore Code Development in a simular way if you wanted to develop with it ?
Great, thank you for your wonderful feedback! We're working on integrating bullet-based syntax, so hopefully you can try it soon. Coding could also be very interesting, but I'm wondering what type of data would be interesting to represent as a graph? Functions and the relations between them, maybe? What do you think?
@@noduslabs I'm working on AI generated pictures for a project and i would love to see the metadata of the nodes confronted in a graph, it would be much easier to work on and probably you could be more creative being able to see many variables and connections. Awesome work btw
@@noduslabsfor software achitects like a really old banking system with tons of interindependent (mispell ik) software to simplify nodes ofcourse you would want simply to throw at it zip files with source code and it should graph every project and it's for example let's say endpoints gpt 4 code interpreter can do that halfway but you need to run python linux env to achieve what i've mentioned but hey there you will have banks money so idk. Just thoughts. Thanks for reading
Yes this tool is certainly cool, it has a lot of potential. Imagine being able to compare ideas 😦
This is quite incredible, and I love the concept.
WOW, Just wow. this gonna be HUGE!
My mind is blown by this ! Thank you !
This is brilliant. I'm going into it RIGHT now! wow....
absolutely amazing, props for giving us such a powerful tool for interactive mind maps!!🧠🔥
Exactly what i need! WOW. Im sooo blown away!
Thanks!
Chat GPT with NODE edit.. love it ❤
Great job fellow scientist! 🎉 Enjoyed the video!
what a fantastic demonstration!
This is facinating. Looking forward to trying it.
очень крутая идея. я поначалу не хотел смотреть до конца, но меня сильно заинтриговала концепция, поэтому определенно лайк)
Спасибо!
I also like to use AI to dig deeper into topics. This method add a better layer of understanding and cross-connecting things.
Exactly. I also like using it to stimulate the creative process and for learning rather than having it do everything for me. In the latter case, the result is too generic and I kind of lose the joy of the process.
wow this just blew my mind! AMAZING WORK !
Thank you!
amazing what've you done so far. keep it up! Congratulations
Wow, this is awesome! Great stuff - keep it up :)
I really want this. Bookmarked for now. I have way to many subscriptions. Incredible app; very impressive.
Just a reminder: infranodus.com - 14 day trial, so you can subscribe and then cancel and not pay anything and try it for 2 weeks. Here, we're giving out a hack on our own subscription system :)
Your accent reminds me to Dr. Edward Frenkel. Great discussion topic and very well presented. Keep'em coming!
hahaha didn't know about him but will explore. saw he has a lex interview, good place to start maybe!
@@noduslabs yes, he was! Please serve yourself. :) First time I learned about him, was on a numberphile video like 12y ago, he talked about the responsability of math teachers into inspiring love for mathematics as a world of discovery and creation for students, he's a great guy
This is a great idea with such a wide potential. Reminded me of Konig's. Do you need an intern?
Yes, we actually do! Please, send us an email to info@noduslabs.com Thanks!
Wow, love it! Thanks for the Knowledge Graph tool.
You are a Genius. God, thank you for your work. Merging AI intelligence with human thinking will greatly impact society. Especially visual learning has a great bandwidth.
Great idea! This is perfect for visual thinkers and for going deeper in a topic.
Thank you!
Thank you. I am interested in several matters. Let's start with the fact that I am a beginner in AI programming and I see its potential. I am in the phase of learning about AI's capabilities and limitations. My questions are:
1. Can I use Infranodus to train AI for different purposes - an RPG dialogue system?
2. How can I make my AI trained with Infranodus public - a website where a user could log in and use the trained AI? Or integrate it with MSG and Facebook?
3. What do I need to know?
4. How much does using Infranodus cost?
Pretty cool project. I think if I was to signup I would want to be able to use my own openai api key rather than paying a price that seems to include queries.
Yes, we will look into this for sure! But the cost is also justified by the fact that we're not a venture-backed company, so we're just charging what it costs to develop.
wow this is so cool! i had a similar idea and tried implementing it using D3.js but i did not get anywhere close to what you have accomplished
me too, maybe we can start a community of like minds and work together because there is always room for improvement!
I think we'll soon shift to D3 :) Thank you for the compliments! It took me about 9 years of work and research, so I'm glad to hear it resonates.
@@Aryankingzor join forces and work on it all together? Please, drop me a line to info@noduslabs.com - would be curious to see if we can collaborate.
I've watched 5 minutes of it, not going to lie - this is exactly what I have imagined “supposed to be available” as a tool, when I have played games in aughties. It still is super confusing to me why people don't develop things that are cool, but rather that are...
Haha well because it takes time to develop and you're really not sure if it's gonna pick up or not... In my case I was too fascinated with the topic.
thank you for such a great content you brought to life here😍
So good had to leave another comment. We want this for exploring blockchain regulation and writing new regulatory theories for more efficient policies. Finance regulation is a quagmire of rhetoric and precedents. Sometimes the precedents don’t line up between the states and across the world. We know every time the US finance regulators sneeze the world catches the cold. Plus so many opinion letters, and updates in the law since 2013. Some states get it right others don’t, would love to plot all the laws for blockchain tech on the books so far across all 50 states and take all the potential finances laws related to article 9 of the UCC and SEC act among others, to make new theories. Thanks for breaking down how to synthesize complex explorations and identity correlations.
Great idea! I think where InfraNodus can be helpful is in identifying the main topics that emerge in regulation and then finding connections between them as well as the gaps (possible legal gaps that could be both exploited but also bridged depending on your objective). It can’t yet visualize a lot of data at once (the limit is our ability to get meaningful insights from large graphs as humans) but we could try to do that through an API. You would still use a standard RAG LLM to search for specific parts of the document.
Fuck regulations, fuck the government
I am working on a crowd simulator and what was and probably still is the problem for me - is the fact that my original and even few next understandings of the problem was far away from the actual complexity. And GPT didn't helped me by spammin my already melted brain with more information that yet needed a very god restructuring. So this is an interesting tool as it firstly shows you the scope of things that you will have to deal with filtering away the contradicting details.
I like how the visualization is, and works in your program. I would say the only thing that's missing is a text to speech workflow and graph level of beginner moderate and advanced. I also feel hearing the words that could be highlighted or prompted to say could help guide creative thinking more in rhythm and patterns. Tailoring it to your own thinking patterns vs thinking how to tailor it should be the objective, overall i really like it !
Both are in there. If you try the software you’ll see it has both speech to text and advanced / simple modes.
🎯 Key Takeaways for quick navigation:
00:00 🧠 The video demonstrates how to combine data visualization and GPT-4 AI to develop ideas iteratively.
00:26 💡 InfraNodus is introduced as the tool for this demonstration, allowing you to input your topic of interest.
01:19 📚 You can gather information from the AI, save interesting points, and start building your own thoughts around the topic.
04:26 👩🔬 GPT-4 AI can provide scientific explanations, and users can request elaboration to deepen their understanding.
06:12 🗂️ The knowledge graph allows you to visualize and organize your ideas, making it easier to identify connections.
09:11 🕵️ Analyzing high-level ideas in the graph can help you identify the core themes of your discourse.
14:54 🌐 You can explore seemingly unrelated topics by finding gaps in the graph and generating questions to connect them.
16:16 🧩 Highlighting gaps in the graph can prompt the AI to generate ideas that bridge these gaps.
18:32 🤔 Crafting well-phrased questions can lead to more interesting outcomes when connecting unrelated topics.
20:16 🔄 Using the current context to generate questions can help maintain coherence in your discourse.
thank you for the transcription. you guys do a service that is literally invaluable. Not even Chat GPT is at that level of understanding and summarizing.
O M G! This will make generating LLM training datasets so easy! I'm for sure gonna check this out this week!
Curious how you would approach this! Let us know!
awesome use of attentional weighting... Congratulations!
Thank you! I didn't know about this term, found some interesting research on that!
congratulations! awesome idea!
Impressive Tutorial. Loved it
This is really neat. Anyone given this a test drive?
You’re a visionary genius. I signed up immediately. Will you be adding image content?
Thank you! Yes, absolutely! Want to add both image generation and image interpretation. What would you want first?
@@noduslabs ironically, neither, as a first pass. I would like to just be able associate relevant nodal points with supporting imagery (like you’ve done with Notes), so that the groupings and milestone nodes could be become richer and more valuable. Example, diving deep into HRV could be accompanied by a found graph of data - decline of HRV over time, or anatomical images of the cardiovascular in good health and bad. After a while this becomes a pictorial doorway into the hypergraph where tapping on an image of interest takes you to the relevant vertex in the model.
From there, image interpretation could become useful, as you express in your presentation, to present previously unexpected topics to then unwrap and explore.
@@fiveshorts Right, so basically just using images to supplement the textual data and to be able to dive deeper into a topic visually, right?
@@noduslabsindeed. And that is not in any way a criticism of what you’ve created here. I think it’s just incredible. I just could see the potential of image supplementation in my minds eye.
😍😍 Thoughts are formed in front of my eyes, and I look at them like Narcissus
You should enjoy this abstract visualization of text made using InfraNodus then: th-cam.com/video/jOj2ghVex9U/w-d-xo.html
Excellent. Most similar idea of taken advantage of AI. AI helps to connect and organise and human put the creativity to discover.
Congratulations!
This sounds like a cool idea for forming a film, adding more to it than a lineral story, but character background, location explanations, etc. If I understand this correctly I really want to do a film that lets you choce at a couple key points, have a feeling this might help figure where and then help in those branches and subsequent endings
Sounds great! I'd be very curious to work on this use case! Do you have some data that we could play around with?
Thank you for showing this to us.
Good stuff. Not feeling like moving on from obsidian which works fine as my 2. brain but this looks fun to learn with.
Lets integrate this into Obsidian, it needs to keep up with AI haha
Do you use the graph in Obsidian? To me it's pretty, but I couldn't get any use from it. I'm surprised they didn't go further with it, especially that InfraNodus was available since 2014, so it's not like what we're doing is rocket science (although it is years of work and experimentation and research and building upon the work of others, notably - Gephi).
How is this 5 months old??? This is what I’ve been looking for. Wow.
Omg this combines chat gpt and obsidian it’s PERFECT for investigation purposes tysm
Great video I threw up some concepts on cosmo knowledge I’d love some feedback
Sure, where can I find it?
@@noduslabs on their community tab under black hole physics
With ur knowledge of HRV, you were able to identify the factual in accuracy what is the user does not have the same level of background knowledge you do in order to be able to identify these inaccuracies?
Good point! I think at this point we always need to verify it somewhere else, but I think very soon we'll have the tools that will allow us to do it automatically.
Awesome video and an awesome app!!! Keep it going
Iv been looking for something like this.
Can you explain the graph a bit more. Some questions I have:
1. What's the size of the dots indicate?
2. What do the colors indicate?
3. How'd you determine which nodes are connected to each other in the word cloud?
1. The size of the nodes (words) indicates their relative influence in the discourse, calculated using betweenness centrality measure from network science.
2. The colors indicate that the nodes (words) belong to the same topical cluster of words (concepts that tend to be used together in this text). Calculated using community detection algorithm based on modularity measures of the text network.
3. The connections are identified based on the words co-occurrence using n-grams (4 word windows).
Thank You. You just made my life far easier.
beautiful visualization
I had a similar idea combining chatGPT and graphs from the Obsidian, but you seem to have something more nuanced. I'm a Computer Science student with 8 classes left. Can I work for you in this project? It is something I am extremely passionate about! It would change my life for the better.
8 classes left? Congrats!!
Hello and thank you for your offer! Sure, you can send us some of your ideas and code samples via support.noduslabs.com and we'll be happy to see if there's a space for collaboration.
This is amazing. Thanks for this video.
Very interesting! Subscribed!
Very valuable stuff, thanks a lot!
Very good Idea which can be improved with time and can help a lot to achive more in shorter time.
Very nice, Now you just need to unbundle this graph into a building blocks and dependency graph, and boom you got the Backlog for your App :)
Hm... curious what you mean?
A fascinating tool. I wonder, however, if it would be possible to add references GPT-4 uses when returning a prompt? In my field it gets things wrongs and i find myself having to verify. I actually prefer it that way but having the references makes it a bit easier
GPT-4 doesn't really return references and even when it does it usually makes them up. That's why I prefer using it to stimulate imagination but not to produce hard facts. For facts we have other modes in InfraNodus, such as using Google Scholar or scientific libraries as the source of truth.
a way of using GPT that i have been dreaming for!
Nice!
I wonder if you could use this for psychology to diagnose someone.
I think definitely to get them to reflect on the connections and gaps between their ideas.
@@noduslabs Beautiful piece of tech. I wonder where it will go!
@@J.LucysCrib me too!