If you implement an easy, user-friendly interface with voice recognition and text-to-speech, and then turn it into an app, this could be a good software for someone with early-onset Alzheimer's.
@@AllAboutAI Also if This allowed the outputs and inputs to interact in a loop, then in theory shouldn’t you be able to create a universal Turing machine with this concept? Correct me if I’m wrong.
I'm blown away by how well-organized your content is. As someone who's interested in AI, it's amazing to see how you've categorized your videos into different topics and subtopics. It really makes it easy to navigate and find exactly what I'm looking for.
Chris. I'm glad UA-cam recommended your page to me. I'm a former statistic data scientist in the US. Now I'm in Moscow with my choreographer wife, Irina, whom I met in NYC while in 1993 she rescued 5 shanghai'd girls from her 40-person, Russian rock & roll back-up, dance troop "Leader". I will visit NYC soon to do long-term math substitute teaching and find out how to legally send my wages back to my wife. Your AI blogs are extremely instructive and the most thorough that I've run across. Keep up your very helpful instructive blogs. You're helping many of us AI Muggles to become ChatGPT Wizards. Lol, Best regards, George
The reason i am not using open ai systematically is the same i am not using siri or their likes, it's running on their systems, I don't want to be profiled to the point where i am predictable. I will when i am running my own instance on my on secured server
I have just discovered Obsidian for note taking and have been looking for a tutorial on using GPT-3 for it just like how you have done in this video! amazing!
@@AllAboutAI If you could show me, a n00b how to connect the sematic search to Notion, I would literally learn enough Python to implement it. Seems like a great opportunity
Personalised study notes with interesting possibilities for linkages , add in self testing , then - you can tutor yourself ( if not now then soon ) Or am I too Optimistic?
This is awesome. I asked myself how to build a searchable so-database for the company i work for. We have so many text informations, but the infos are not categorized.
@@AllAboutAI yes, as you did here. I think this could be a awesome Methode do dive in our information. The next step is to give this infos to our clients via website.
How large a dataset could you use to create the vectors/brain? I've kept ad hoc notes about my life, ideas, projects, and client interactions in OneNote for several years. It's megabytes of text. Would it be possible to create a brain from all that?
@@keithprice3369 I think you would need some kind of structure to have a reliable output. Freeform would work but I'm not sure what use you would get out of it. In the video his data is labeled with dates, times, etc. So when he asks when an event occurred the AI can return that. If you don't have any structure I'm not sure what question you would ask.
@@samuelflippin1890 I don't know the requirements . But in my example, each OneNote page has a date and time attached, so that should cover time based queries. As for the rest, it seems it COULD work just on key words. Like, "when did I talk to Mary Thompson about getting access to their server?" Or "what's the connection string for ABC's server?" "Show me all the recipes for chicken." I'd GUESS NLP could determine the intent and find the applicable data. But, seriously, I know next to nothing about this. I'm, just intrigues. That said, OneNote has a pretty dang good search engine already. It's why I use it.
So basically, you get the embeddings for an article, use cosign similarity to match your query to a doc, and then feed the query and search results back to GPT (Davinci) completion along with a prompt? For the diary of things you did, did you split up the subtexts to get multiple groups of embeddings (Like get separate embeddings for the text of each day that you did stuff) or just got emeddings for the whole diary?
Great video! How about the information provided to your personal or client's model GPT-3, is it confidential? i mean, if you have secret data to fine-tune your model, does it will be shared with Open AI? or it will stay confidential at the private acces of the client? Thank you very much!
I have been writing digitally using an app called Day One would something like this work with it? It's all stored in .json files on my computer and online.
@@AllAboutAI Sure man. awesome work. I got an idea from this where i can a create a resume about me, and just share the website. So, the recurituer can just ask questions about me.
Veldig bra! I think something like this might be useful in an app for people with short-term memory loss. I do think we need to begin discussing the ethics and methods of digital twinning as it will become a very popular thing. I could see some legal implications for compelling your second brain or digital twin to testify against you. Would it be a warrant for the information or would it be an extension of yourself? Anyway. Great to see this kind of thing being done.
Hello Kris, I'm a 3D student really interested in learning about A.I. I already played a little bit with ChatGPT and Midjourney and I follow some channels like your's, but, I want to understand the system beneath and how it really works. What do you recommend me to do? I have 0 experience in programming (only basic C++)
Excellent succinct and well-explained as always. There are so many uses I can think of. If I assign an identification key to an article will the key be retrieved in the summary or perhaps several keys if the text is integrated? This is to identify the source of the information. I have been struggling to prepare pdfs for fine tuning such as converting to text then json especially when batch processing. I will head over to the subscription page and hopefully learn more. Preparing unstructured data may be a helpful future video 😉
this is awesome!! I have a question. When using chat gpt I realized thet the model doesnt know some specific topics and I want to create a model that contains that knowledge. I have books with that topic (at least 20 of them) and I can access transcriptions of the lectures about the matter. What kind of model should I use? A second brain or a fine tuned gpt? Imagine that I want to give it two uses: 1. People ask questions about the concepts and examples of its applications. 2. Be able to evaluate and correct phrases that the user will give the model.
You cant use fine tuning to add knowledge to a model. Fine tuning is just for learning new tasks or format! So semantic search might be a good opition. I will be doing a tutorial on my memebership if you are interested:)
@@AllAboutAI I just joined to see that, but that is not online yet or is it ? The 'membership' tab is a little convoluted for newcomers. Do 'membership' videos get listed in the videos tab too ? How can I find that ? The membership tab only contains a timeline
Really interested in this, so thanks.. question: What membership tier gives me access to the tutorial on how to get started on this myself? How much experience programming would I need? (Tricky to answer I know but could a novice do this?)
Hello Carl :) You can start from Tier 2. I have totally beginner tutorial up to more advanced. Over 30 members only videos now. ua-cam.com/users/AllAboutAIjoin
Great video! One question when enhancing GBT through the use of a custom dataset using semantic search does the model still retain all previous learned training data from Openai? Would I still be able to ask random questions that fall outside of the “2nd brain”? Than You!
Man, there are plenty of people on YT doing these tutorials but it's like we don't get nothing. We look at what you do and you don't provide a step by step guide. Is there a guide on this one? I am considering join the highest tier you have if there's a tutorial on how I can do it too. P.S. Also, can you use any kind of prompts? And does it have past memory of what you ask him of?
How large a dataset could you use to create the vectors/brain? I've kept ad hoc notes about my life, ideas, projects, and client interactions in OneNote for several years. It's megabytes of text. Would it be possible to create a brain from all that?
@@AllAboutAI Pinecone does look quite interesting. I wish they had a better summary of what it does and how it works. I'll dig deeper. I could also see a huge benefit from dumping years of emails into it, but I don't really know how I'd get the emails into a single text file.
Become a member here:
ua-cam.com/users/AllAboutAIjoin
Join the newsletter here:
www.allabtai.com/newsletter/
Have you seen Dave Shapiro?
Thanks!
If you implement an easy, user-friendly interface with voice recognition and text-to-speech, and then turn it into an app, this could be a good software for someone with early-onset Alzheimer's.
This is a serious good idea 👏
@@AllAboutAI Yes it is a great idea. If only anyone has already built something like this. Imagine he would be able to post it here.
@@AllAboutAI Also if This allowed the outputs and inputs to interact in a loop, then in theory shouldn’t you be able to create a universal Turing machine with this concept? Correct me if I’m wrong.
Also would be cool to see this linked with Google location history, or your other google services data
initiate a repo? Make it free and open for everyone to contribute?
This seems like a world changing projecr
I'm blown away by how well-organized your content is. As someone who's interested in AI, it's amazing to see how you've categorized your videos into different topics and subtopics. It really makes it easy to navigate and find exactly what I'm looking for.
Thanks for tuning in :) more to come
Chris. I'm glad UA-cam recommended your page to me. I'm a former statistic data scientist in the US. Now I'm in Moscow with my choreographer wife, Irina, whom I met in NYC while in 1993 she rescued 5 shanghai'd girls from her 40-person, Russian rock & roll back-up, dance troop "Leader". I will visit NYC soon to do long-term math substitute teaching and find out how to legally send my wages back to my wife. Your AI blogs are extremely instructive and the most thorough that I've run across. Keep up your very helpful instructive blogs. You're helping many of us AI Muggles to become ChatGPT Wizards. Lol, Best regards, George
The reason i am not using open ai systematically is the same i am not using siri or their likes, it's running on their systems, I don't want to be profiled to the point where i am predictable. I will when i am running my own instance on my on secured server
I have just discovered Obsidian for note taking and have been looking for a tutorial on using GPT-3 for it just like how you have done in this video! amazing!
Cool :) I use Notion atm, and using this for searching my notes
@@AllAboutAI Cool, so instead of text files you read from Notion? Can you share the code too...
@@AllAboutAI Amazing, can you please explain how to connect to Notion to search in there with your method?
@@AllAboutAI If you could show me, a n00b how to connect the sematic search to Notion, I would literally learn enough Python to implement it. Seems like a great opportunity
Inspiring video and use of this technology!
is the gpt-3 demonstration in the video 100% running on local machine? is there a subscription/fee like using open api?
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
Personalised study notes with interesting possibilities for linkages , add in self testing , then - you can tutor yourself ( if not now then soon )
Or am I too Optimistic?
You could be on to something :)
@@AllAboutAI Thanks - Hope "someone can make it work"
This is awesome. I asked myself how to build a searchable so-database for the company i work for. We have so many text informations, but the infos are not categorized.
Thnx:) are you concidering semantic search?
@@AllAboutAI yes, as you did here. I think this could be a awesome Methode do dive in our information. The next step is to give this infos to our clients via website.
But how do you create the vector embedding in the first place? I know there’s ads embedding endpoint but it doesn’t seem to take my Unicode text.
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
Can I do this with books? I want to train an AI with 100+ books and use it like Chat GPT, is that possible? Thanks
I have not tested scale, but there are services like pinecone :)
Hugin
Hi Kris do you have a full guide how to built one in members section?
How large a dataset could you use to create the vectors/brain? I've kept ad hoc notes about my life, ideas, projects, and client interactions in OneNote for several years. It's megabytes of text. Would it be possible to create a brain from all that?
Yes, as long as the notes were organized in some way
@@samuelflippin1890 Oh, interesting. So, it doesn't work with freeform text? There needs to be a structure? Bummer. But makes sense.
@@keithprice3369 I think you would need some kind of structure to have a reliable output. Freeform would work but I'm not sure what use you would get out of it. In the video his data is labeled with dates, times, etc. So when he asks when an event occurred the AI can return that. If you don't have any structure I'm not sure what question you would ask.
@@samuelflippin1890 I don't know the requirements . But in my example, each OneNote page has a date and time attached, so that should cover time based queries. As for the rest, it seems it COULD work just on key words. Like, "when did I talk to Mary Thompson about getting access to their server?" Or "what's the connection string for ABC's server?" "Show me all the recipes for chicken." I'd GUESS NLP could determine the intent and find the applicable data. But, seriously, I know next to nothing about this. I'm, just intrigues.
That said, OneNote has a pretty dang good search engine already. It's why I use it.
Fantastic! Can I show this to my University Innovation class?
Yes sir :) Would be cool if you mention my Channel then:)
@@AllAboutAI That goes without saying! I will l get my bursar to buy me a subscription
@@AllAboutAI I will be delighted to do so!
Can you teach as a easy way, step by step to create our own second brain? with GPT-3
Hello Alex! Yes, i have a detailed tutorial on this on my UA-cam membership:)
@@AllAboutAI If I pay the normal membership, will I have access? or has to be the highest tier?
Tier 2 to access to videos, Tier 3 is Tier 2 + access to download the script on GitHub :) I will invite you to the private git
You say the vectors in the JSON file represent ”embeddings”. In this context, what are the embeddings?
Is there already any way to put all the documents from a firm into a semantic search database? (Like a lawfirm)
hmm, check out pinecone.io maybe :)
Is the run using the api or can you run it locally? So is your private data shared with the api?
This is running with the OpenAI API yes, so i would not share confidential data before talking to OpenAI
What do I get if I join your membership? Can I get code and dump my own text and make it 2nd brain ?
Sure. Releasing the tutoriall and code this week on the membership :)
Had a "Her" movie poster in my room for the longest time back in the day. Love Spike Jonze!
Yes, i just saw it now. Loved it
Please which software did you use to achieve this?
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
So basically, you get the embeddings for an article, use cosign similarity to match your query to a doc, and then feed the query and search results back to GPT (Davinci) completion along with a prompt? For the diary of things you did, did you split up the subtexts to get multiple groups of embeddings (Like get separate embeddings for the text of each day that you did stuff) or just got emeddings for the whole diary?
Yes! I use the dot product of the vectors. I just got embeddings from the whole diary yes
Dude!! This is amazing, I would pay for this service!!
Thnx:)
Great video!
How about the information provided to your personal or client's model GPT-3, is it confidential? i mean, if you have secret data to fine-tune your model, does it will be shared with Open AI? or it will stay confidential at the private acces of the client?
Thank you very much!
Hello! thnx :) You should be careful with confidencial data. I would atleast contact OpenAI first if that was your use case!
I have been writing digitally using an app called Day One would something like this work with it? It's all stored in .json files on my computer and online.
Yes! That should be possible!
Thank you very much for the information you share. I was wondering if you share the source code on a Github repository?
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
Incredible and... i think inevitable :D good one dude
Thanks a lot :)
This is dope. The future is really coming!
It is kinda cool yeah :)
Where might I find how you made this?
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
@@AllAboutAI yea very much so. When I am ready to build my project I will totally join and use this to build off of.
I want to Learn How to make this. Which Membership Should I Join then ?
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
Which vector database did you use to do this?
what models/libraries did you use to do this?
ada-embeddings-002 from OpenAI, is that what you mean?
@@AllAboutAI yes so, you used ada-embeddings-002 to convert the text to vectors and search it?
I will create a tutorial for this so join my memebership if you need details:)
@@AllAboutAI Sure man. awesome work. I got an idea from this where i can a create a resume about me, and just share the website. So, the recurituer can just ask questions about me.
Is there a limit to how large the brain can be, like can you make it massive? e.g. load a bunch of book pdfs into it so you can search through them?
you could do a phone app out of this with integrations with different apps
December 2022: Talk to a robot
January 2023: Celebrate new year
February 2023: We're now building brains.
Whaduh fck just happened?
So...Can't you just Ctrl+F on the Notepad and search for "airport"?
that is not the point of the system, think bigger :)
Veldig bra! I think something like this might be useful in an app for people with short-term memory loss. I do think we need to begin discussing the ethics and methods of digital twinning as it will become a very popular thing. I could see some legal implications for compelling your second brain or digital twin to testify against you. Would it be a warrant for the information or would it be an extension of yourself? Anyway. Great to see this kind of thing being done.
is this something you still do? is there an update for this?
Great video, after I became a member of the channel, is it already possible to download the code on your github?
Yes! :)
How can I implement this with Notion
hmmm, i have not done any research into that yet
Hello Kris,
I'm a 3D student really interested in learning about A.I. I already played a little bit with ChatGPT and Midjourney and I follow some channels like your's, but, I want to understand the system beneath and how it really works. What do you recommend me to do? I have 0 experience in programming (only basic C++)
Hello! I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
@@AllAboutAI Yes, I thought about it, but I'm bad at programming. Are begginer friendly the tutorials?
Did you try Nightcap Guru? I've begun using it as my main advisor for my dreams. It's an AI dream interpreter xD
No i havent, yet 😃
So we need an OpenAI API key for this tutorial?
Yeah, if you want the full tutorail i will be making it here :) ua-cam.com/users/AllAboutAIjoin
Is there anyway I can access this code please?
I will be making a tutorial and access to the code on my membership :)
@@AllAboutAIok thanks I commented only 5 min into the video and got more info that’s how excited I got haha
so in short you fill it daily so that one day you can summarize it, hmm. seems like a killer plan!
Could be interesting at least :)
Thanks for this (and the other) videos!
What size (in tokens) do you think is the best for an embedding element?
It really depends i think. Read more here: openai.com/blog/new-and-improved-embedding-model/
Excellent succinct and well-explained as always. There are so many uses I can think of. If I assign an identification key to an article will the key be retrieved in the summary or perhaps several keys if the text is integrated? This is to identify the source of the information. I have been struggling to prepare pdfs for fine tuning such as converting to text then json especially when batch processing. I will head over to the subscription page and hopefully learn more. Preparing unstructured data may be a helpful future video 😉
Btw professionals teach data cleaning with datasets downloaded already, not very helpful. Most of us work with unstructured pdfs and word documents.?
Thnx a lot Johnny:) lets talk more on DC
this is awesome!! I have a question. When using chat gpt I realized thet the model doesnt know some specific topics and I want to create a model that contains that knowledge. I have books with that topic (at least 20 of them) and I can access transcriptions of the lectures about the matter. What kind of model should I use? A second brain or a fine tuned gpt?
Imagine that I want to give it two uses: 1. People ask questions about the concepts and examples of its applications. 2. Be able to evaluate and correct phrases that the user will give the model.
You cant use fine tuning to add knowledge to a model. Fine tuning is just for learning new tasks or format! So semantic search might be a good opition. I will be doing a tutorial on my memebership if you are interested:)
@@AllAboutAI oh I am very interested xD thanks!
You are very welcome:)
@@AllAboutAI I just joined to see that, but that is not online yet or is it ? The 'membership' tab is a little convoluted for newcomers. Do 'membership' videos get listed in the videos tab too ? How can I find that ? The membership tab only contains a timeline
Let me know when you have the tutorials and I will glad pay for them.
This week prolly :) ua-cam.com/users/AllAboutAIjoin
Really interested in this, so thanks.. question: What membership tier gives me access to the tutorial on how to get started on this myself?
How much experience programming would I need? (Tricky to answer I know but could a novice do this?)
Hello Carl :) You can start from Tier 2. I have totally beginner tutorial up to more advanced. Over 30 members only videos now. ua-cam.com/users/AllAboutAIjoin
Great video!
One question when enhancing GBT through the use of a custom dataset using semantic search does the model still retain all previous learned training data from Openai? Would I still be able to ask random questions that fall outside of the “2nd brain”?
Than You!
Yes of course. It will always look in your "brain" for answers, but this can be alterd with different prompts or versions
@@AllAboutAI Thank You for the reply!
Thats some kind of spooky!
Great but spooky.
Tnx i guess :P
Good stuff! I might have missed it but could you show us a sample file of digital_brain python file to see how it's called?
Hello! thnx :) I have tutorials and step-by-step videos on this on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
@@AllAboutAI Awesome! Just joined. Can you possibly tell me the title of the video for this? Thanks!
@@AllAboutAI Would you be able to put some sample files in your GitHub repository?
This is the shit. I can't wait for the step by step video on this tool.
:)
Looks like this concept is similar to LangChain, is it?
Is this all offline?
You will need an GPT-3 API so no its is not offline, but you can run it locally
Man - imagine that integrated into Obsidian.
Man, there are plenty of people on YT doing these tutorials but it's like we don't get nothing. We look at what you do and you don't provide a step by step guide. Is there a guide on this one? I am considering join the highest tier you have if there's a tutorial on how I can do it too. P.S. Also, can you use any kind of prompts? And does it have past memory of what you ask him of?
Also, bump because this is important :) (For me it's time sensitive xD ) let me know if you have this tutorial in the private area
Hello! I have tutorials and step-by-step videos on this vid coming soon on my Membership page if you are interested :) ua-cam.com/users/AllAboutAIjoin
@@AllAboutAI soon, meaning? I am mainly interested in this one :). So if there's no tutori now there's no point to joining...
@@AllAboutAI what lvl do you need to join for tutorial on this and have you created it yet?
Is this meant for people who are proficient in coding only?
You need some skills, but its not that diffucult. I have step-by-step tutorials for beginners so you can advance :) ua-cam.com/users/AllAboutAIjoin
Sorry, but where is the introduction of the actual code or steps you used ?
here you go: ua-cam.com/video/szLg5DNrQHQ/v-deo.html
@@AllAboutAI It says 'members only' even though I already joined ? EDIT: Higher level needed
I am building something similar. My Agent as emotion-based volition. Comment if interested to share with me :)
Cool :) Would be interesting to hear more on Discord
Hi. Great job. Thanks for share. Is there a free language model or free API ?
Agreed, Her was a really good movie :D
yes :D
The title implies you had a brain before you started... AI WINTER RETURNS.
Thnx for watching :)
you like the premier league? nice. do you play fantasy premier league?
wow awesome!
Thnx:)
I love graphs! Let's liberate Königsberg!
Euler FTW!🥳
So google and chatgpt?
Nvidia popularized the use of GPU and is now revolutionizing AI too.👏🏼
Can def see a bright future for Nvidia
its actually a digital diary...
just a fine tune model
I like your lifestyle - work 3 hours per day.
haha😅 i guess the notes were not 100% accurate
Surreal
Impractical given that you have to actually make notes
Bro thinks he is in 2040
Who wants to sign up for my brain tor network
sounds scary 😅
So you got an AI to tell you info that you already told it? You know you could have used notepad to do that right?
..fiction...basing reality on a dribble of drivel of fictional past memories = rinse and repeat. Enjoy the madness...
homestuck dirk strider
Can you build my first brain for me?
haha
How large a dataset could you use to create the vectors/brain? I've kept ad hoc notes about my life, ideas, projects, and client interactions in OneNote for several years. It's megabytes of text. Would it be possible to create a brain from all that?
That is very cool :) i dont have an exact answer for you tho, but i should be no problem. Check out services like pinecone.io to :)
@@AllAboutAI Pinecone does look quite interesting. I wish they had a better summary of what it does and how it works. I'll dig deeper. I could also see a huge benefit from dumping years of emails into it, but I don't really know how I'd get the emails into a single text file.
I will be doing a tutorial on my membership if you are above average interested:)
@@AllAboutAI Yeah, sounds very interesting.
@@AllAboutAI If I subsribe to the top tier will I get the access to the code for what you achieved above? :)