What Is LangChain? - LangChain + ChatGPT Overview
ฝัง
- เผยแพร่เมื่อ 30 ก.ย. 2024
- Twitter: / gregkamradt
Newsletter: mail.gregkamra...
Overview about why the LangChain library is so cool
In this video we're going to look at the problem with vanilla ChatGPT and how LangChain comes to the rescue.
Follow me on Twitter
Personal: / gregkamradt
My first Agent will be called Agent Smith - what could possibly go wrong?
One of my next videos is how you spin up multiple agents, what could go wrong?
Requested HD videos, max resolution is 720 here for video
Stick w/ the playlist! They get better, I invested in my setup to help make better content
By the way, thanks for this, it's very well presented, coherent and easy to follow 😀
Glad it was helpful!
Can you do a tutorial on using locally running Alpaca with Lanchain? I need to use a LLM that can be used for sensitive information that we index and load into the LLM as context. Your videos are great man -- keep it up, I appreciate what you're doing!
This is really interesting. My boss told me to look into langchain and i think this playlist will give me some valuable insight to go along with the official documentation!
earned a new sub
Nice! Sounds great, I would recommended checking out the core concepts + use cases video as a good next step
This is exactly what I've been looking for and thanks for the really good intro!
Glad you like it!
Let's go Gators, swore I knew a dude with your exact name
Save yourself time, video starts at 1:41
Thank you!
Now that OpenAI announced ChatGPT plugins, how relevant is LangChain? What usecases does it have that ChatGPT doesn't? Are there any obvious way these two could work together to complement each other?
Check out my latest video on the 7 core concepts of LangChain.
Plugin's are 1 aspect of LangChain and they are already working with them in a great way
@@DataIndependent I literally just swallowed your entire playlist in a couple of hours, thank you 🤣
@@jakobkristensen2390 totally agree! spitters are quitters! LOL
Excellent series for beginners, thank you Greg.
Isn't OpenAi building direct integrations to the tools we are using every day by rolling out Plugins? If so, langchain is losing the bet?
so can langchain then connect chatgpt intelligence to our mdm database and therefore provide insight into the data in the form of either chat bots or other ?
LOL. Well they did it
Yep - super cool to see what's happening
Yeah doesn’t plugins compete with langchain?
@@nickle485 point is langchain can work with other free llm
This video didn't age well AT ALL
Imagine watching this video and thinking plugins for ChatGPT is the same thing 🤦♂️
Thank you for this, subscribing and diving into the other videos. One question - what app did you use to record the video? Thanks!
In the beginning I used loom, but now I use OBS and edit with davinci resolve
Thanks a bunch! I spent the long weekend here in Thailand catching up on your videos and following your coding instructions. This video on LangChain was super helpful [What Is LangChain? - LangChain + ChatGPT Overview]! I just finished this one and I'm excited to keep going through your playlist. Thanks again!
Great series, Thanks for breaking it down to smaller chunks
Just subscribed, it's clear to the point🙂
Nice! Thank you!
Thank you for putting this together. Much appreciated.
Nice Intro! Looks to me, this is the flow orchestration system that specializes in LLMs tasks & components. This category of products exists already for other data automation solution, but now, LangChain is focusing on their LLM ecosystem building. That is my guess.
Check out my latest video here for a better overview of the material
th-cam.com/video/2xxziIWmaSA/w-d-xo.html
Curious to know about data privacy. Any idea if langchain sends our data over to OpenAPI and then OpenAPI can use that data?
OpenAI (or any LLM you use) will only have access to the pieces you send over to them. For example we load up 5-10 essays to LangChain. But to answer a questions about those docs we only send 4 smaller chucks over to OpenAI.
So if you're worried about OpenAI seeing any of your data, then yes there are privacy concerns. They no doubt are using your data to train more models, beyond that I'm not sure what they are doing with it.
@@DataIndependent Thats what I thought. As a use case, folks need to be careful about implementing summarizations or q&a chatbots with their personal data. For businesses, this is an issue and thus not likely a good solution.
There was a recent change to the openai TOS where they state that your prompts won’t be used to train the model moving forward. They do keep the data for 30 days for abuse detection.
@@anandsheth5490 This is good to know, thank you! Is there a potential workaround to this? (Using the model but without sending the data over)
@@koganin9 well they changed the terms of the API where they will not store the data
I'm a teacher and I really like your style! and info!
Nice Intro into LangChain. Sub'd and 👍Liked!
Look forward in viewing the other 4 to begin with. I saw you had Vid on Webscraping. I'd like to request a LangChain process / project? in that direction.
Keep it Going 💯!
Nice thank you. Yes please request away.
What problem are you trying to solve?
Subscribed! Amazing intro to langchain!
Awesome glad that worked out - What else would you like to see?
Can I starded with LangChain? Is it recommended for a beginner?
I have a idea about an project to do
How does this compare to Haystack which has been around for years?
Now as we know that OpenAI introduced plugins does it make Langchain obsolete?! 🤔
Not at all.
Check out my latest video on the 7 core concepts of LangChain which show why it's still extremely important
This series is awesome, building these bots is so easy!
How is this technology not threatened by OpenAI's ChatGPT plugins?
Check out my video on the 7 core concepts which should help clear it up
Hey Greg, can you please do a tutorial on how to use langchain for figma. Langchain have docs for that but probably I am not getting it wrong. Hopes on you...
very good, could you share the ppt?
Found your playlist on the Langchain site. Planning to go through the entire playlist. I'm running a macbookpro with an Intel chip. Do you think there will be issues running Langchain on my machine? Many thanks in advance!!
You should probably invest in building a pc or buy a gaming laptop as you’re going to be ask to run code etc. You might be able to do some things on the macbook pro tho
How does the introduction of ChatGPT plugins change the scene?
Check out my latest video on the 7 core concepts of LangChain to see why it's important
the best explanation on the internet! Thanks youu
Anyone do this please improve your chunk process from data randomly cut from middle data any one can create this kind of pipeline
I would love to see a note on how you think this should be done. It's a super cool concept
What tool is he using for recording the video?
Anyone Please help!
Old videos I used regular loom and my webcam (Logitech brio)
For new videos I use OBS to record and a Sony camera with lens
Hello can you tell me the prerequisites for this series?
Thanks
Hey! Just an eager to learn and a bit of python, not much!
Nice explanation.
Thanks
with release of plugin functionality, is langchain going to be relevant in future
Yes - check out my latest video on the 7 core concepts of LangChain and why its super important
Does the release of plug-ins for open ai mean langchain isn't needed?
Not at all - check out my latest video on the 7 core concepts to see everything else it does
You had me at RoaringKitty. Lol our sweet, wholesome boy!
Right?? I like throwing fun references in there. That was a wild time with RK
This is the missing playlist I've been searching the past two weeks for xD
Nice!! Glad to hear it
Awesome content. Subscribed !!!
Awesome thank you Sachin
Really exciting video. My request to you, please record videos in 1048p.
This was before I switched over to obs. New videos are better.
@@DataIndependent I completely understand. Really appreciate your work and love it.
Great video. What is Pinecone? Langchain in the cloud?
Pineone is a vectorstore which is a fancy way of saying a database that can hold embeddings or vectors.
LangChain is a python (and js) library to help you manage LLMs
Yes we are toast
How do you think this story will play out?
Are plugins similar to langchain?
Yes, both are ways to connect LLMs to the outside world. OpenAI has some tools and LangChain connects them as well. You'd use LangChain when you want more control
It is cool
Nice!
Hey man! Thanks for taking the time to make this video. I really appreciate it 🎉 Peace bro
So basically Langhchain is for people who don't know how to program in Python?
What Pandas is to data analysis, LangChain is to create LLM apps.
Do you do raw python when analyzing data too?
@@DataIndependent No, but I do build my own LLM implementations because the OpenAI is already highly abstracted. I found little value in Langchain
Good video, but why is the video quality so low? =/
I went scrappy to start and this was on Loom to get content out. I’ve moved to OBS and have other upgrades coming soon too.
Thanks for the comment!
Really nice Nico!!! Really well explained!! It'll be really aweson see much more , and one end to end app with js
please increase the video quality
You got it!
Excellent Video, However you should make a new one, as it seems ChatGPT will have extensions.
Nice catch by suggesting it was a bet that Chat GPT would want to stay on the API side, seems like its a platform now.
Ya, you're right. Technically OpenAI didn't create tools, they allowed others to integrate into them.
@@DataIndependent Yes, it seems the future is yet to be seen exactly!
great video thank you so much for making it! doing some cool things with this knowledge! maybe in the future, we can collaborate! :D
Just what I'm looking for, the practical aspect. Thanks
Glad it was helpful! Be sure to check out the '7 core concepts' video for more in depth explanation
I'm trying to understand how this works vs fine-tuning. Ultimately, I don't care about the difficulty of setting it up. I just need the best result.
I did a test with langchain and the results are quite good, but not nearly good enough. It's ok for most businesses I guess. But not for anything that involves complex or sensitive data.
I have access to thousands of saved tech-support and user-education conversations regarding our knowledge base. But I doubt that simply putting those into a fine-tuning format is going to eliminate hallucination.
So there's probably a huge amount of human work that would need to be put in preparing the data:
1. To remove anything that might potentially be ambiguous or confusing tot he LLM
2. To remove outdated information
3. To (somehow?) make sure that the LLM understands the difference between the information given/explained and the examples used to do so
- Which means that it would need a separate database of explicitly defined variables to ensure it can correctly interpret/understand the explanations given in the training data (Which, in my use case, would be thousands of variables. Though these can probably just be tucked in arrays, so that shouldn't be too difficult)
4. Figure out how the LLM interprets training data which contains prolonged conversations with people who had trouble understanding things.
- Will it take the confusing/frustration/struggle-to-understand from users and end up copying that into its own behavior?
- Do I need to weed out conversations that involved discussion about multiple topics, or is the LLM able to discern that?
And so on, and so on..
There are too many unknowns for me at the moment.
Nice! Thank you for the comment. I was thinking about this topic with a friend as well. Companies have 1000s of examples of support related cases that can be used as training material.
My guess is that a solid level of pre-processing and example selection will give you the best results. It depends on how easy your requests are.
Ex: Reset my password vs "provision me a new server with XYZ specs"
well thats ur main problem right there @ruben u r using the wrong method to achieve wht u want - for wht u r trying to do FINETUNING is NOT at all the method u shod be using. instead u shod be using SE
@@thebicycleman8062 SE?
@@thebicycleman8062 What is SE?
Bad bet
I didn't think OpenAI would integrate with 3rd parties. However it was the reverse that came true, 3rd parties opened up their APIs for all LLMs to use.
Either way I thought OpenAI would want to stay on the intelligence level rather than going after actions but their solution is elegant
Nice share👍
Thanks for visiting
Thanks. Lots of potential here.
Nice glad to hear it - what're some of the ideas you have?
@@DataIndependent my company has a huge confluence library. It would be nice to have ChatGPT learn it.
3:50 ah, we have aged! Haven't we?
It was a different time! However, they didn't build direct integrations into those tools...The tools open sources LLM-friendly plugins for any LLM to use them.
LangChain can integrate with them today
@@DataIndependent yea langchain is still amazing. Talking about amazing - you are amazing too my man. I am watching your entire playlist right now.
1080p plz
Upgrading the set up shortly
Good work. Thank you.
Nice, thanks for the kind comments
Thanks Greg , Amazing Video
Nice, thanks Mike
Thank you
You're welcome
Super excited!
Awesome!
Your course has benefited me a lot. Would it be possible for you to share the PPTs used in your lectures? Thank you very much.
great video but for a 6 minute video, the first 2.30 minutes didnt add any value. just wanted to add that.
First off, I loved the video! Thanks a lot for making it.
Second, all the comments now are "ooh, plugins plugins, LangChain is obsolete." But it seems there's a waitlist? Are plugins usable immediately? It seems LangChain is ready to go...
What I'm really going after -- the ability to train ChatGPT with peer-reviewed studies and have it answer questions based on what it just learned. What is the most efficient way today ... is there already a plugin for that or should I move ahead with LangChain? (Father of a newborn, I have ZERO time and need to be super-frugal with where I spend my "free" time...)
Thanks a lot.
> Waitlist
Yep there is
> Available immediately?
Not yet, more need to be built
Check out my latest video on the 7 core concepts of langchain to see why it's still important
Nice use case, check out the latest LangChain webinar (link likely on twitter) which covered this exact use case. You could also use a tool like chatbase which has this built out of the box
@@DataIndependent awesome! Thanks a lot, I appreciate your reply (and expertise).
didn't open ai reveal they will allow plugins? wouldn't that make this obsolete ?
perhaps, but that only came out today! While LangChain has been around for over a month, that's a loooong time in the AI world
@@davidconsumerofmath agreed, and if i am not mistaken open ai allows plugins to chatgpt where LangChain can allow you to integrate into your environment easier? or am i wrong?
You're right. Because Plugins are open, other LLMs (and LangChain) can work with them as well.
LangChain takes care of a lot of functionality behind the scenes that gives you more control on how your programs actually run.
What do you think about plugins system that has just been released by OpenAI, it enable ChatGPT to communicate with other service like Wolfram, will it replace LangChain?
Check out my latest video. I go over the 7 core concepts of Langchain and 6 of them deal with things other than plugins