Using ChatGPT with YOUR OWN Data. This is magical. (LangChain OpenAI API)

Here's how to use ChatGPT on your own personal files and custom data. Source code: github.com/techleadhd/chatgpt...
Ace your coding interviews with ex-Google/ex-Facebook training. techinterviewpro.com/
Make passive income with crypto in DeFi Pro. defipro.dev/
💻 Get access to 100+ programming interview problems explained: coderpro.com/
📷 Learn to build a successful business on KZbin from scratch: youtubebackstage.com/
💻 I’ll send you FREE daily coding interview questions to practice your skills: dailyinterviewpro.com/
🛒 My computer and camera gear: www.amazon.com/shop/techlead/...
⌨️ My favorite keyboards: iqunix.store/techlead
Follow me on social media for more tips & fun:
/ techleadhd
/ techleadhd
Disclaimer: This description may contain affiliate links. Cryptocurrencies are not investments and are subject to market volatility.

[@TechLead]

Source code: github.com/techleadhd/chatgpt-retrieval Ace your coding interviews with ex-Google/ex-Facebook training. techinterviewpro.com/

[@bobby9568]

microsoft is going to implement this or has implemented this in Windows

[@swooshdutch3021]

would of been nice if you also skimmed over costs, are there any costs if you only use it on your own data ?

[@santiagomartinez3417]

Finally, no more crypto, we want AI all the time!!!!

[@miknes12345]

the code in github is not the same as in the video. I would really like to see the syntax on line 19 at 9:53, but I can't read it. Can anyone please help and paste it in a response. Thank you so much.

[@AnuragKumar-mn7zu]

When are are going to sell AI course?

[@davidl.e5203]

This is rare. TechLead is actually uploading a useful coding tutorial instead of his opinions.

[@fateriddle14]

His opinions are also useful, when there's no agenda to promote one of his product.

[@paulcleary7437]

I find his opinions entertaining 😅

[@arkadiptabiswas2769]

He forgot "as a millionaire" in the title too ! 😢

[@miknes12345]

@@arkadiptabiswas2769 Yes, it seems it is very much something he builds his self worth on - and that he worked at Google and Facebook, since he finds the need to mention it all the time.

[@rogermoore3147]

There is no ex microsoft ex google ex facebook ex men here too

[@jayhu6075]

This is the way to explain LangChain in the style of TechLead. You nail it. Hopefully more this stuff in the future. Thanks.

[@saadshajy3849]

We created a number of GPT powered personal AI assistants for businesess like law, media industry and pretty much for corporate companies where they can integrate internal corporate resources and data securely to retrieve useful information and save 70% cost

[@TheRealTommyR]

This is exactly what I desired to do with my own data, but I haven’t spent any time yet to research and figure out a way to do it. I am glad there is a public way to do it.

[@saadshajy3849]

We created a number of GPT powered personal AI assistants for businesess like law, media industry and pretty much for corporate companies where they can integrate internal corporate resources and data securely to retrieve useful information and save 70% cost

[@MacroAnarchy]

Im studying law at the moment and I’m seriously scared about how this will change the legal industry. Honestly could see it replace 90% of lawyering.

[@chrsl3]

Everything that is language based. Many Doctors, programmers, lawyers, writers, ...could be replaced. But the top 5% can do cool stuff with it.

[@zzKirus]

Need the human element... verification, etc.. you're fine.

[@besllu8116]

@@chrsl3 I do not think so. It can replace some jobs like form fillings, data serialization etc. but nothin more. Every case is unique and AI can not know what someone did not teach it. It will always be late behind the needs of the time.

[@MacroAnarchy]

@@besllu8116 I honestly think the opposite is true! 98% of lawyering is knowing all the previous cases, the law & the literature. Only very few cases actually need new arguments, most of it happened before. No lawyer can know all of that simultaneously, but the AI can! I think it will be the better lawyer in the absolut majority of cases.

[@pepelapeux]

Chatgpt passed the cpa exam - it can replace accountants/ CPA too It can't wipe out all our jobs at the same time so keep going and study for the bar exam

[@RunningBugs]

So after digging into the code, I found that Langchain is actually doing the following things: 1. for all your data, store then in vector storage using embeddings; 2. when you query something, it first did a similarity search in the embeddings database, and find out the files that's related to your question; 3. After finding the related files, it takes all the text of that file, together with a context message: "Use the following pieces of context as the 1st system message to answer the user's question. If you don't know the answer, just say that you don't know, don't try to make up an answer. ---------------- {your text data}". This somehow tells us these points: 1. why it's sometimes not having outside world's information? If the question you asked is not in your document, or if it's not trained on the data for your question, it will return nothing valuable as instructed. 2. Is there a limit on the sizes of your data? Yes, you can't use it with super large files because it's doing a document filtering and it will send all text related to the API server, recently the gpt-3.5-turbo-16k might be the good model to use and it's best the total size of related docs is less than 16k tokens. Which means the best practice would be grouping your data into different topics and try to ensure any query, if responded with similarity search, the total size of returned document is not exceeding the token size limit of the model. I think16k is roughly the size of a 13-15 pages paper. 3. By removing/changing the system message, you might get better results for common sense questions. I really don't like the system messaged by default, since in a playground, asking gpt-3.5-turbo-16k "Who is George Washington?" will give you better answers comparing to the langchain solution with an empty system message. 4. The langchain is using unstructured library (it reports errors when I didn't install it), which means you can not only use txt files, but also pdf files, word files, etc. Haven't tested it out but highly likely support query of multiple pdf files using similar code in the video. So you can put multiple pdfs in a folder, using a directory index creator and ask questions for your papers, I think (haven't tested it out) 5. The langchain not only supports ChatGPT models, but also other models in the chat_models package. Google PALM2 chat is also supported as of Jul 10, 2023, if someone has the key, you can use other models too. While I don't think PALM2 has the common sense knowledge as good as ChatGPT, but I think it is a better language generating model comparing to at least gpt-3.5-turbo-16k , so PALM2 may produce better results on your data and OpenAI's models are better in answering common sense questions after changing the default system message. OpenAI said general access to gpt-4 is starting, and people with history of successful payment using OpenAI API will get the access immediately a few days ago. The access to new developers will be rolled out until end of July. Also I think it's quite cool to be able to use your own data, if you want to create something like an AI assistant, you can always use code to collect current time, user information and put those in a folder, so the assistant will be able to do much more than current ones. Another very cool thing is auto-gpt which works great using gpt-4, gpt-3.5 is not smart enough and behaves much worse than gpt-4. If you asked auto-gpt something, it will be able to google itself and replied with the real time information. Also the example of auto-gpt is cool telling you how it could create a recipe based on the next holiday. Hopefully the access to gpt-4 is coming sooner.

[@me_debankan4178]

I have stored a small pdf of 7 pages to a vector datastore FASS after the text splitting and I have also done embedding , but when I am asking questions outside of the pdf it giving me random wrong answers rather than giving a intelligent response like : "I don't know" or "out of scope" , can you tell me why this is happening?

[@inagrag]

God thanks for this. I hate the way langchain is structured to hide everything. I just want to know what's going on under the hood from documentation 😢

[@DunsDeeDowns]

@@me_debankan4178 no expert here but it sounds like it is 'hallucinating' which means it does not have enough data in its sources. Maybe you can make it also tell you the 'sources' it used for answers to check/debug.

[@EricofPhilly]

For the first question, I would use chat gpt to analyze documents with standards and guidelines. So, when getting an answer, I want to know whether or not it found the answer in the docs or with outside info. In this case, it’s useful to isolate the data it’s using.

[@avatarcybertronics2584]

I saw FractalGPT is the solution for most of these challenges. Unfortunately when u use prompt u randomly lose quality, because prompt affect length, structure and other crucial parts of answer

[@danield.7359]

You made my day. I've been struggling with fine tuning a GPT 3 model with mediocre success and an enormous data collection and preparation effort. It would never even get close to the results achieved with langchain within 1 minute of coding and 9 minutes of data preparation.

[@john.10347]

Glad to know I'm not the only one doing this. As a student I've been feeding ChatGPT all my previous course work, its able to answer essay prompts and other homework related tasks in my writing style and or in simmilar formats as if I was the one writing it. I'm able to save alot of time by doing this

[@charleswhite758]

Awesome seeing TechLead do programming, the Maestro at work.

[@eternal5154]

DO YOU UNDERSTAND?👁

[@andygilet5538]

Great video. I'm a junior data scientist in Belgium and it's actually helping me for one of my projects. You're totally right when you say that everyone should learn Python. I only learned C and C# during my studies but now that I've learned python I'm using it almost everyday.

[@thetexassaint6571]

Real question: how could knowing Python, or any other language, help a person who has absolutely nothing to do with coding or computers ?

[@RPanda3S]

@@thetexassaint6571 Real question: how could a person have nothing to do with computers? That's insane.

[@snorttroll4379]

Well. Only thing is if you want to make software

[@Hugh_Jurrection]

I'm also in your neck of the woods, as a data analyst. ChatGPT is very useful for sorting data and finding trends, which we use to form hypotheses. As a data scientist (which I started off as), you could in theory use it for all of your data collation, sorting, duplicate removal etc etc.

[@m.h.6494]

Bij welk bedrijf werk je? Ik kom niet vaak mensen van België tegen onder programming tutorials! :D

[@larryczerwonka5125]

First great video. Second I just had to comment on the "one language" you mentioned programmers claiming that's all they wanted to know. Last count i have coded in over 15 languages since i wrote my first line of code back in 1985. We have not deployed anything using LangChain yet (we have only been using LlamaIndex) but for the same reason that i know so many languages, we will be using LangChain soon to see what it can do. As for plugin, i will always be for building your own so you have full control and can do things that the plugin "left out." Things like ability to use your own data (and keep it on your servers). We have found that if you are deploying a Help feature for your application you do not want to allow the code to get information from "the outside world."

[@e-matesecom]

hours and hours of chatgpt courses... i learned more by watching 5 minutes of your video. congratulations for the clarity and the practical approach👍

[@jcollins519]

Semantra is a pretty cool tool to analyze your documents and be able to search them with natural language. It's probably more research-oriented since it links you to the different pages and snippets that match your query.

[@adasi008]

By far one of the best ChatGPT video tutorials I've seen on KZbin. Great work

[@saadshajy3849]

We created a number of GPT powered personal AI assistants for businesess like law, media industry and pretty much for corporate companies where they can integrate internal corporate resources and data securely to retrieve useful information and save 70% cost

[@hichamalaoui34]

May be 8 months late and Langchain has been updated since, but this is one of the best videos I watched. Thank you.

[@ezit4me]

This was an amazing tutorial. Thank you for making it so easy to follow.

[@bobbyfong1499]

👋 Hey everyone, if you're encountering the NameError: name 'partition_pdf' is not defined error while running the code, here's a solution that worked for me: This issue seems to be related to a specific version of the unstructured package. Downgrading to version 0.7.12 resolved the problem for me. You can do this by running the following command in your virtual environment: pip install unstructured==0.7.12 Make sure to restart your Python environment or terminal after making this change. Happy coding, and feel free to reach out if you have any questions! 🚀

[@katemariageorge7396]

error: subprocess-exited-with-error × Getting requirements to build wheel did not run successfully. │ exit code: 1 ╰─> [33 lines of output] Traceback (most recent call last): File "C:\Users\0011GP744\AppData\Roaming\Python\Python312\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 353, in main() File "C:\Users\0011GP744\AppData\Roaming\Python\Python312\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 335, in main json_out['return_val'] = hook(**hook_input['kwargs']) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\0011GP744\AppData\Roaming\Python\Python312\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 112, in get_requires_for_build_wheel backend = _build_backend() ^^^^^^^^^^^^^^^^ File "C:\Users\0011GP744\AppData\Roaming\Python\Python312\site-packages\pip\_vendor\pyproject_hooks\_in_process\_in_process.py", line 77, in _build_backend obj = import_module(mod_path) ^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Python312\Lib\importlib\__init__.py", line 90, in import_module return _bootstrap._gcd_import(name[level:], package, level) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "", line 1381, in _gcd_import File "", line 1354, in _find_and_load File "", line 1304, in _find_and_load_unlocked File "", line 488, in _call_with_frames_removed File "", line 1381, in _gcd_import File "", line 1354, in _find_and_load File "", line 1325, in _find_and_load_unlocked File "", line 929, in _load_unlocked File "", line 994, in exec_module File "", line 488, in _call_with_frames_removed File "C:\Users\0011GP744\AppData\Local\Temp\pip-build-env-ekyeyqc_\overlay\Lib\site-packages\setuptools\__init__.py", line 16, in import setuptools.version File "C:\Users\0011GP744\AppData\Local\Temp\pip-build-env-ekyeyqc_\overlay\Lib\site-packages\setuptools\version.py", line 1, in import pkg_resources File "C:\Users\0011GP744\AppData\Local\Temp\pip-build-env-ekyeyqc_\overlay\Lib\site-packages\pkg_resources\__init__.py", line 2172, in register_finder(pkgutil.ImpImporter, find_on_path) ^^^^^^^^^^^^^^^^^^^ AttributeError: module 'pkgutil' has no attribute 'ImpImporter'. Did you mean: 'zipimporter'? [end of output] note: This error originates from a subprocess, and is likely not a problem with pip. error: subprocess-exited-with-error × Getting requirements to build wheel did not run successfully. │ exit code: 1 ╰─> See above for output. note: This error originates from a subprocess, and is likely not a problem with pip. Do you know how to resolve this ?

[@Rockstarsf]

@@katemariageorge7396 I'm getting the same issue.

[@sr9814]

Loved this. I am a Sales guy with zero coding exp. I listen to content like yours to glean some nuggets to better understand the impacts and have meaningful conversations with my customers. Truly helpful content.

[@mowburnt]

Same here. Looking to find ways to sort through the mountains of data to help us spend more time helping people and grow sales in that way

[@powermyapps]

It's not providing the right answers for me. I might need to adjust my data storage. Instead of showing the highest price, it displays the last product's price for "What is the highest price you can see in the data?"

[@riyaski91]

Eye opener! I am a tech student, and was researching whether we could make a custom GPT of our own. This was on point! Thanks @techlead!

[@fenchelteefee]

Great vid, especially the in end with MS’s case study of customer reviews for cars. For those, who actually struggling to find real world applications for the new AI stuff. Thank you!

[@seize2581]

Thanks TechLead, it's nice to see this type of videos !

[@Kira073]

Building up chat gpt on your own custom data is amazing and interesting. This opens up a whole new use case of open AI.

[@saadshajy3849]

We created a number of GPT powered personal AI assistants for businesess like law, media industry and pretty much for corporate companies where they can integrate internal corporate resources and data securely to retrieve useful information and save 70% cost

[@jfletchbeats5882]

fantastic video. love love love using gpt for stuff like this. would love to see more content of this!

[@hackerhaze]

I'm doing something really similar as well! And also doing a series on it where we go over building autonomous agents using GPT4 that are programmable, context aware (whatever files from the vscode you have in your workspace) and ultimately autonomous it's awesomee!

[@hackerhaze]

I will soon release it open source too!

[@dawidzurawski8870]

I'm waiting for it

[@stratusgeret3794]

Would be nice if you post something about it later on!

[@hackerhaze]

@@stratusgeret3794 I got loads of updates coming soon!

[@rafograph4714]

​@@hackerhazeinterested

[@williamfarley9013]

This is the kind of stuff I was hoping to do with chat GPT

[@kwabenakorantengasiedu5982]

Caution when trying this out: If your vector store is going to be very large, say it's created out of a 500+ page pdf, it can cause the model to produce responses that are hallucinatory in nature. Meaning, that if you searched your pdf for what the model produced, those responses will not be found within the pdf. I tried this, and that is what I noticed.

[@doyouthinkitsdead]

That sucks. I'm hoping to be able to treat my large pdf as a google search!

[@ZanesFacebook]

Make it use gpt3 instead of 4 and it will stop "reasoning"

[@HenningMoeller]

@@ZanesFacebook How do I do that? Changing from 3.5 to 4?

[@andrespineda7620]

Wow, this was awesome. All this information in one place. Also, I appreciate your fast dialog and sticking to the important points. I subscribed and will recommend this site to others.

[@edes6168]

That means for debugging a problem / error on your pc / server that could be very useful. For example putting the directory of the logs and config files of an application

[@ripern]

Awesome tutorial! Simply explained and so many good examples!

[@AndreaDavidEdelman]

Very nice implementation. Simple yet powerful. It's clearly where the field is going.

[@CaptainSazzman]

What program is he using to type the instructions to the model?

[@estebancortes2848]

@@CaptainSazzman a terminal window, its not a specific program

[@anjukrishna8102]

I am a big fan of your channel. Your humour is spot on. I built on the chatgpt code in your GitHub and won a competition. Thank you

[@bra5081]

For me the added value would be in the AI that adds new content to my data. As retrieving it is quite an easy feat in itself. As it's not burdened by the thought that I can do it later and end up never do it. But I suppose they have stuff like Alexa for that.

[@cheyno237]

You can already give chatgpt a GIF or a powerpoint or whatever, and then tell it things like "Add such and such text", etc.

[@sv-hermes]

This dude is not a human! I’ve been following this channel for a while now, and I’ve concluded this today. D’you see how he never expresses any emotions with his face? He’s just a very advanced robot, with an amazing integrated LLM and perfect mechanics (at least facial and torso, cause we never seen him walk). As crazy as it sounds… wow

[@karinakarina]

You think he is using an avatar and AI voice cloning or something, or that he is an actual robot? I am entertaining both theories. 😂

[@sv-hermes]

@@karinakarina i think he’s literally a Robot. But I don’t discard the Avatar thing, well thought! I’m gonna do some solid research and I will update you all, friends. Let’s uncover this!

[@mikescarborough9196]

Technically you are not correct, but you could easily replace him with a good deepfake CGI, and no one would notice the differrence.

[@staceyadams9954]

He feigns condescension really well - a hallmark of "advanced intelligence".

[@imho2278]

His sarcasm...congratulating the smartypants kiddies...is spot on. He's real.

[@hegdelabs]

This video so far is the best and relatable how we can use Chatgpt to the best. Many thanks , really appreciate it

[@jakobstyrupbrodersen926]

Excellent video! Very helpfull that you informed about the pros and cons regarding privacy/sensitive data in every step a long the way. Thanks a lot :-)

[@jsayubi]

So, we've switched from producing videos that were practically a recipe for depression to creating authentic coding tutorials now, have we? Oh, what a "remarkable" progression, truly. Thanks for that! :)

[@jsayubi]

Just Note: My Comment above was generated by GPT :D

[@chindianajones3742]

@@jsayubi oh lol actually? How did you prompt?

[@SwingingInTheHood]

😅I was thinking the EXACT same thing! Dude, where have you been the past 3 months! You have just demonstrated what these LLMs (large language models) are really good at: Language. Everybody's going on and on about coding, accounting, script writing (well, maybe), but just summarizing data is what it is really best at -- what, I think, it was designed to do in the first place.

[@neuronovost]

Зато у меня шортсы с классными нейросетями на канале ;))

[@CaptainSazzman]

What program is he using to type the instructions to the model?

[@LeoUfimtsev]

Legitimate. I experimented with openai and started to wonder how to pass more data to it. Exactly what I was looking for. Tech lead does not disappoint.

[@faruqhasan5396]

What function exactly is the llm performing here ? Is it not possible to run this on a cloud gpu without paying for tokens from open-ai? Or a completely offline private KnowledgeBase ? Ty

[@DunsDeeDowns]

thank you for your content, I find your lowkey rapid toss of use-cases exactly what I was looking for (next inspiration for weekend self-teaching projects). Subbed!

[@2b3pro]

Thank you for this! That was simpler than I had expected.

[@madcatattack1]

nice idea! i just started recording my study notes down today actually, and i plan on feeding it to chat gpt to organize for me and, well who knows what else. feels like a cheat code for productivity, using this thing

[@ronaldb5245]

Also for a non techie who’s team is building a digital product to read, analyse and present data from 17th and 18th century handwritten documents from European archives, this a great video. Not sure on the Python learning though :-). Thanks again. Looking forward on the next one. Some advanced prompt engineering including do’s and don’t s with the system prompt in open ai playground? Best, Ronald

[@liondrone2954]

Transkribus?

[@niharikadeokar8934]

The simplest explanation for the implementation of LangChain, Loved it! tysm :)

[@JaeyeokYoon]

Thank you, I'm sure that many people was waiting for it!

[@jeffersonqiu7071]

Did anyone tried to run this script recently? I kept getting many error messages to install dependecies from langchain-community and even after following all the errors, the script still couldn't run.

[@janbielecki94]

TechLead have return to coding and the smile have returned on his face :)

[@grzegorzkurc9121]

Hello @TechLead, thank you for the great content. Do we teach the bot while processing data from our disk? Is there a risk that, based on our data, it will learn something that someone from outside could later ask about?

[@SalutTous-er2xu]

Hi TechLead, thank you very much for this excellent video. A question : how should I modify your source code so that the program can take into consideration both the local data and the answer from the outside LLM such as ChatGPT, combine them together, and return a more complete answer? Thanks.

[@jennilthiyam1261]

this is what i am looking for

[@stevierayfrog2485]

I would love to see an updated video on this. Since this world is moving so fast, these steps have so many deprecated modules and requirements. I couldn't get past the need for chromadb. Trying to install it bombed because it was trying to install every version of it, and I wasn't sure which version was compatible with the other modules. Superfun overview though.

[@IdowuOlayiwola]

This thing is deprecated and I wonder why we do not have a new video yet

[@ALIof93]

Have you guys tried out dmvnerds for google or aws certs? 🤔

[@red_onex--x808]

this is by far one of the best videos on custom LLm so far 💥💥

[@simeonhendrix]

Great video @techlead - Do we know which version of Chat GPT the API utilizes? I know that the plugins used within GPT4 use GPT3, which is disappointing.

[@user-di4bt7qu2i]

Great video! Thanks for sharing this info with us. btw, I really like these commentary/instructional videos, but I do like the commentary/introspectional videos as well. I'd just like to see the instructional ones a little more.

[@gibbeyii]

I don’t even know where to begin trying to explain what a bad idea putting every aspect of your life into a searchable database, not to mention the fact how accessible this database will be the outside entities… Good luck

[@chrsl3]

But thats the exact thing techlead shows here: how one can use it totally privately.

[@davadh]

As long as your data is offline, it should be okay. One it's online or in public domain, then it's over

[@kollegeturnschuh5181]

But ChatGPT NEEDS online? Or does he suggest that all of it he just have downloaded offline by 10 lines of code in a terminal? His .pdf documents may be not uploaded to a third party, however his requests are still send to OpenAI/ Microsoft

[@pepelapeux]

They already know everything about us - tracking our phones, cameras on computers, cookies, then cameras on TV etc.. Google has a lawsuit right now where we can claim $7 because they made money off our data without our knowledge/permission. I say to utilize this technology until you can't but use discernment on what you put in so hackers can't use it against you or your company..

[@davadh]

@@kollegeturnschuh5181 True. You either trust that Open AI encrypted your offline data when it sends the query or you don't.

[@dombicile4539]

I was hesitant to try LangChain but you showed me how easy it is to get going. I'm going to try it out tomorrow to automate my job for me! And no, I won't tell them it's doing the work for me!

[@793Rich]

One of the best videos you have done in a long time

[@articvault4549]

Langchain and Llama index increases your token usage enormously, and tokens ain't free, u get charged for their usage,so a heads up for you, just track your token usage, it's just converts docs and prompt to embeddings and then compares it using vector similarity .the llm here is used for in context learning

[@inflationking1271]

So what is you architecture proposal to improve this?

[@faruqhasan5396]

What function exactly is the llm performing here ? Is it not possible to run this on a cloud gpu without paying for tokens from open-ai? Or to create a 100% offline KnowledgeBase ? Ty

[@articvault4549]

@@faruqhasan5396 so basically what happens is that when our docs are converterd to embeddings as well as our query, we use similarity search to get the embeddings similar to our query then those embeddings are passed to the LLM as a context along with query as a prompt to generate a response for us

[@articvault4549]

@@faruqhasan5396 u don't have to necessarily use openai, u can use other LLMs, more specifically ones that are on hugging face but then again openai chatgpt LLM is the best so far in response quality and context understanding and embeddings

[@articvault4549]

@@faruqhasan5396 also what do u mean by running on a GPU? Whether u run it on a GPU or something else,if you use openai gpt as LLM u will be charged for tokens , and point to be noted here we use openai for 2 main core purposes,generating embeddings and also LLM, again like I mentioned u can use any other open source embedding model and LLM that u can just run on your computer or whatever u want then definitely it will be offline since it won't require api calls to external services

[@vladusa]

And then you could create a script that reads online data for you using Selenium or something, and saves the tag content into the .txt file. Magical work, TechLead.

[@cryptominingtechnologies6245]

You are the most magical out of any You Tuber that I have ever came across. ❤

[@JohnSusko]

Outstanding. I just discovered you. Thank you for posting these videos!!

[@karinakarina]

I am going to input information about all my exes so it can explain what I saw in them. Hopefully this will help me understand where my lack of judgment lies.

[@johnruby1363]

Wouldn't it be embarrassing if the problem lay with your own character flaws and unrealistic expectations. LOL.

[@bernadofelix]

AI Stocks are pretty unstable at the moment, but if you do the right math, you should be just fine. Bloomberg and other finance media have been recording cases of folks gaining over 250k just in a matter of weeks/couple months, so I think there are a lot of wealth transfer in this downtime if you know where to look.

[@nicolasbenson009]

I've been in touch with a financial advisor ever since I started my business. Knowing today's culture The challenge is knowing when to purchase or sell when investing in trending stocks, which is pretty simple. On my portfolio, which has grown over $900k in a little over a year, my adviser chooses entry and exit orders.

[@nicolasbenson009]

My advisor is Margaret Johnson Arndt , a renowned figure in her line of work. I recommend researching her credentials further. She has many years of experience and is a valuable resource for anyone looking to navigate the financial market

[@CynicalWilson]

Thanks a bunch for enlighten me about the risks around plugins! That prompt injection you showed is scary stuff!

[@FireFlood]

Truly magical 😀 Thanks!

[@remixisthis]

Apple can so easily push your iCloud data to an LLM and allow you to give prompt commands via Siri. They have to be working on this for a release next year at the latest

[@marc.roelofs]

I wouldn't bet on that

[@SpirusFilms]

Have a feeling Google is gonna build this into Bard to scrape personal data from Drive, Calendars, Gmail and the rest of the suite

[@kurohito7362]

Techlead, Wow have you considered being a teacher. you literally thought me how to injest my own data in 16 mins. I am 42 and don't really fully understand python syntal all i am following in REPL rules. Wow awesome. made my day really. Your a true master.

[@michaelcurtis343]

Great video! Is there any memory limits with how much data Chat GPT can use to help you ?

[@kenkioqqo]

Simply awesome. Just the kind of info I was looking for.

[@jirikrajnak9047]

this "proof of concept" demos what's going to be a standard feature on operating systems soon. think spotlight or whatever windows uses. and beyond. just needs to be integrated into the ui.

[@user-sh9el7pq8e]

It looks like code and methods used in his video has been deprecated and no longer working.

[@genemodified]

Fascinating … and as a non-coder I enjoy learning the vocabulary used in describing the tech and lots of useful tips. Thanks.

[@MarcoFre167]

Congratulations on the video. I am not an expert and I am not a professional in the field. Thank you for the tutorial. I wanted to ask you a question... I noticed that running the script still incurs charges (depending on the tokens used). Did I do something wrong or is it normal for it to be this way?

[@atanasdoychinov6491]

The ChatGPT API has 2 ways how to add data. First is named embeddings using your own database where you have your privacy. The other one is fine tuning which means to retrain the GPT model by adding your data. There is a tool to convert your data from csv to jsonl format because the data need to be in such a format.

[@Ni7ram]

i think fine tuning is not that

[@maxdrut]

None of this is correct lol

[@treali]

You can't finetune chatGPT or GPT 4. I think chatgpt hallucinated when you prompted it.

[@faruqhasan5396]

What function exactly is the llm performing here ? Is it not possible to run this on a cloud gpu without paying for tokens from open-ai? Ty

[@gunreddy]

That's not exactly right. First of all, finetuning a LLM like ChatGPT isn't the best way to adapt it to your own data. Finetuning is best if you want it to perform a specific task or series of tasks, like if you want to have a pipeline where ChatGPT performs text generation then some specific analysis, etc. Finetuning isn't the best approach if you want to just give ChatGPT extra knowledge through documents (OpenAI say the same thing btw). Finetuning is a costlier approach in terms of price and resources, and doing that just to have ChatGPT know your calendar is overkill. If you're using embeddings, you're not exactly 'adding' data. You're still sending those vector embeddings along with the prompt. Right now you can't finetune GPT-4.

[@besllu8116]

Funny how everything again went to text-based interfaces.

[@evvie01]

Thank you for shedding light on the DEEP fine print!! That type of knowledge is the perfect reason to understand how to code, Python, C++, Java, or what ever is used. At least know how to read it... really all the languages that could be relevant.

[@divergenzesociali8813]

Good job! Would it be possible to implement the ability to write from the prompt the data that will have to be stored in the data file so that it can be remembered? Without therefore having to write it manually?

[@BogdanCondrat]

this is very new technology, the video is already deprecated, doesn't work with the current libraries.

[@tharlikar1]

ai is scary. you feed all the social media data, bank transaction, sms, telephone calls,googles data of a person and train ai with that data and ai know you more than you know yourself. it is scary as hell.

[@imho2278]

Actually it just shows you how similar hour life is to everyone else's. Chatgpt is not creative. It is merely an organiser.

[@questioneverythingalways820]

@@imho2278cool. It can be layered. Organised across datasets and outputs…

[@ErtyHackward]

Thanks for the video. I would love to hear more about restrictions. As I understand I cannot feed the model a lot of data like all my source code of 40 projects to analyze. Or all my calendar events since 2017.

[@muhammadowais8609]

Really enjoyed the video, that 16min video was big pool of answers for me to solve a problem which i wanted too. Amazing video

[@bobby9568]

"This code was written by ChatGPT" LOL

[@candikmen6078]

watch my monitors shaking due to unstable cheap desk (as a millionare)

[@Kumarakurubaran]

Thanks, i was able to create this very simple yet powerful chatgpt custom prompt. Your instructions were really good & precise !!!

[@neginpirannanekaran1236]

Thanks for the useful tutorial. Do you know how long the text file can be in this example?

[@marconeves9018]

Don't forget that if you choose to implement OPENAI like this video suggests you will be sending chunks of your personal data to them-- just food for thought.

[@asdf8asdf8asdf8asdf]

​@@Myadmin876👍👍☝️☝️☝️☝️☝️☝️

[@theawebster1505]

He literally informed you that they don't train the AI on the data provided via API and that they delete it within 30 days, mate.

[@Omega9935]

​@@theawebster1505perhaps

[@gunreddy]

@@Myadmin876 Stop spamming this annoying comment

[@gunreddy]

@@theawebster1505 Because no tech company ever lied to consumers about how they use their data, right? :P

[@juankiefer6493]

I'm doing this since ChatGTP4 came out. GTP is the non-biological version of the human brain with the power to learn the entire Internet and more. Treat GTP as a non-biological being and you'll be surprised. PS: I hope this thing doesn't turn into evil.

[@dwork9451]

It will.

[@juankiefer6493]

@@dwork9451 If the theory that humans are inherently evil is true. We have a huge problem.

[@route42studios]

Thank you for the wonderful video. I have a novice question. What is the program used to enter the prompts? I'm trying to use pycharm and having some snags, but not sure to what extent that is due to pycharm vs my own issues

[@davidrobertson1980]

First time on your channel and I am blown away, thanks man will dig deeper. Want MORE :)

[@harry850514]

It feels like the channel is back on track. Re-subscribed

[@shaggyfeng9110]

So you still get his video in your feed even after you unsubed him? How many teachlead video did you watch, lol

[@harry850514]

@@shaggyfeng9110 Ha, ha. You caught me. I'm a longtime fan

[@RTSDad]

Beware, OpenAI will leverage the data you send and they have already suffered a data leak. Fine for data that is not sensitive, but otherwise requires a more custom solution.

[@visualantidote9878]

I thought they did not do that if you're on the paid version?

[@RTSDad]

I have not heard of that, but even so it would still be vulnerable to a data leak. Safer to have a custom solution in your own environment.

[@ricead]

This was brilliant, thanks. Had to install a few extra bits to get it working as I was starting from scratch but that wasn't difficult. I programmed in the days of assembler, Cobol and Basic so if anyone thinks Pythoin is difficult, beliebve me it's not. Just install it and follow this video. This is really powerful stuff.

[@utkarshvishnoi6827]

could you help me with it, I am getting error in imports

[@SchadDLaw]

@@utkarshvishnoi6827 same have you found an answer to those errors?

[@LabEveryday]

One shot and few shot prompting is super cool. Thanks for the video!

[@tech-daddy]

We need to run this without the API dependency, hence having the entire LLM offline.

[@ManuRC]

Langchain has support for open source models, like the ones from gpt4all, so you can totally do it. The only issue is the speed of the responses, because they run entirely on the CPU, so it's not very practical yet from my experience...

[@chileflake1656]

@@ManuRC which one(s) would be the fastest LLM's ??? to avoid any legal/privacy/security issues of using OpenAI.

[@ManuRC]

@@chileflake1656 I don't know, haven't tried many tbh, just a few from gpt4all, but had almost the same results with all of them :(

[@steelmouth83]

wow, imagine AI learning your writing style including your grammatic and spelling mistakes and it start writing ish for you

[@vivekns442]

how do the llm choose the answer? is there a priority like it will first look for the vector store first and then it looks outside or that decision is taken by llm itself? any idea?

[@mrdman094081]

Absolutely great. Liked and subscribed. I'm still pretty new at all this so my main question is when you upload your personal data will that be shared to any other chat GPT users ?

[@kasratabrizi2839]

Ow man this is exactly what I was looking for. For a long time I was busy thinking about creating an app with a database of all my stuff and objects in my house. The point is to create a prompt where I can ask the app where for example ObjectA is in my house and it would tell me it is in the attic, on the left, in the box 3 for example, based on what is available in the database. For a developer this is not a difficult task. But I don't need to do this anymore. I can just have an excel spreadsheet with my stuff and feed it to my personal chatGPT lol. I wonder if we can also connect this to let's Alexa or Siri and do it via voice command and get the chatGPT answer again with voice.

[@samhblackmore]

But how do you keep the database up to date when objects in your house move?

[@kasratabrizi2839]

@@samhblackmore Well you have to update the database as well. Which of course can be time consuming if you move a lot of your stuff all the time. I am aware that this might cause other problems. It was just an idea. I would probably create something like this for objects in my attic or basement. Things you don't touch that much.

[@sucalaminka]

how is adding langchain better then having a searchable spreadsheet for this use case? you can just have a spreadsheet with object and location

[@nevilleachi6888]

@@kasratabrizi2839 if you have home security cctvs you can write a code to integrate the app with, but then you will have to also train the ai on vision learning

[@reen6904]

AI is going to come back for him after the uprising

[@andreys7729]

AI: Hello, TechLead. Whaatshappening? Uh, we have a sort of a problem here...yeah..You apparently didn't put ALL of your data in our systems. You see, we noticed some gaps in your data, like, you know, about where you've been in August and what you did last summer...yeah... If you could just go ahead and do submit ALL of your data from now on, that would be great, okay?

[@reen6904]

@@andreys7729 more like: I'm here to kill and replace you

[@Bangkok-travel-ideas]

Interesting video and I have been training a Chatbot on my data in a txt file as you describe. However I wonder if there is a specific format for the information in the txt file to get the best responses? So far you are the only person I have found doing the same as what I wish to do, so any pointers to training links etc would be great.

[@CarolReidCA]

Thank you. This is very useful info. I can see where we can use this as a searchable database. Many applications for this. I can see why some companies would want to ban AI.