Tap to unmute

Building the Future with LLMs, LangChain, & Pinecone

  • Published on May 29, 2023 veröffentlicht
  • Join Harrison Chase, the creator of the breakout library LangChain, and James Briggs, developer advocate at Pinecone, as we explore the new age of AI and it's biggest enabler - LangChain.
    You’ve seen endless examples of developers and engineers building AI-powered apps in record time using LLMs, LangChain, and Pinecone. We’re in a special moment in history where rule-based processes get replaced by powerful AI systems. At the center of this paradigm shift are Large Language Models (LLMs), vector databa\ses like Pinecone, and the LangChain library. In this online workshop, we’ll explore the tools behind this new wave of technology and how to wield them to build mind-blowing applications.
    🔗 Link to slides:
    🔗 Link to notebook:
  • Science & TechnologyScience & Technology

Comments • 51

  • Memes of Production
    Memes of Production 2 months ago +28

    You guys are living legends. Bleeding edge shit, and Harrison, your enthusiasm is infectious. Langchain and Pinecone are going to be 2 pillars of the open source AI renaissance that is already starting. Keep up the good work! I am consuming everything you're putting out there and want more.

    • blackenedblue
      blackenedblue 2 months ago

      But the cheapest hosting plan on pinecone is 70 bucks a nonth

    • Memes of Production
      Memes of Production 2 months ago

      @blackenedblue Weird, I'm doing pay as you go, but also interested in others. Any you recommend? Especially ones that have extra features for multi-modal data

    • Kyle Shrader
      Kyle Shrader Month ago

      @blackenedblue that sounds cheap for any business

    • Chris
      Chris 6 days ago

      They will be acquired or AWS will have serverless acorn

  • Bomullskaka
    Bomullskaka Month ago +4

    Awesome video and content! Some few tips since this is so good content. I have hearing problems so it would be really awesome if you could use studio microphones and also record in higher quality to give a better video and audio output. Anyhow, thank you for taking your time and provide all these information 🤗

  • Limowreck
    Limowreck Month ago +7

    Harrison I recently saw an interesting statement “the fastest growing programming language is English”. Thank you so much for creating Langchain and James thank you for your incredible tutorials and insights. ❤

    • Eyemazed
      Eyemazed Month ago

      It is unlikely that programming languages can be efficiently translated into English or any other natural language. Programming languages are designed to be precise and unambiguous, and they rely on a specific syntax and grammar to express their meaning. Natural languages like English, on the other hand, are much more flexible and often rely on context and interpretation to convey meaning.
      While there may be some efforts to make programming languages more accessible by translating certain keywords and commands into natural language equivalents, such as using "if" instead of "if statement," this approach may introduce additional ambiguity and can make code harder to read and understand for experienced programmers.
      Ultimately, the most effective way to learn and work with programming languages is to become familiar with their specific syntax and grammar. This allows programmers to write code that is precise, efficient, and easy to read and understand by others who are familiar with the same programming language.

    • Limowreck
      Limowreck Month ago

      @Eyemazed i think the use case here needs to be clarified. As we can see from openai cataclysm project, GPTforall and lanchain these LLMs employ agents which write perfect utility code on the fly in order to perform a task. The humans speak english to the LLM who then write whatever code is necessary to complete the task. There is abstraction layer in this use case which eliminates the need to translate speech to code. The speech triggers openai actions which writes the code it needs

  • Elaheh Salehi Rizi
    Elaheh Salehi Rizi Month ago

    Hi I have a question. I am using GPT-3 for the task of few-shot text classification, I have tried directly GPT-3 and also via LangChain (using the same few-shot examples for both, the same format of the prompt (the one from LangChain formatter) and the same features of GPT-3). I had better results when I used LangChain. Is there any explanation why LangChain performed better than directly using GPT-3? thanks

  • johnnybloem1
    johnnybloem1 2 months ago +2

    Thank you for educating those of us new to the LLM field how to use your applications. I did not get the powerpoint. May get a link to the repo please?

  • real23lions
    real23lions 11 days ago

    Incredibly informative. Thanks 🙏

  • tso
    tso Month ago +1

    Sick! Going to build some stuff on this asap

  • Er Alok
    Er Alok 23 days ago +1

    great work , i asked chatgpt what is Langchain, it doesnt give any helpful information as data limit is Sep 2021,
    your documentation helped me understand this wonderful tool and concept.

  • David Comfort
    David Comfort Month ago +1

    I’m trying to grasp how Azure OpenAI compares to this approach? Is it doing something similar behind the scenes?

  • Christian Magnus Ingul

    Love this content!

  • Sacred Grove
    Sacred Grove 10 days ago

    agents + Pinecone+ Zapier = true ai assistance.

  • AV Adams
    AV Adams 28 days ago

    Maybe use semantic drift as a chunking metric/cutoff.

  • togo
    togo 2 months ago +3

    Really great from Harrison 🎉

  • TheSavviestTechDude

    Do vector databases also apply to computer code and not just natural language?

    • TheSavviestTechDude
      TheSavviestTechDude Month ago

      and how does it give "meaning" to a chunk of text and is it the job of the embeddings mechanism to do that and then store it to a vectorDB ? can you combine traditional DB's ( mysql and postgres ) with vectorDB's to improve search?

    • Armond Willingham
      Armond Willingham Month ago +1

      Yes, they do. Determining how you break your files up before embedding seems important; you don’t want to cut functions in the middle, or remove relevant context

    • Harrison Bond
      Harrison Bond Month ago +1

      ​@TheSavviestTechDude it's the embedding mechanism that does the semantic conditioning. The vector DB just lets you efficiently search/cluster the embedded space. You have to use the same embedding process on your input to the database, since the database is basically just doing distance calculations.

  • Shivam Bhardwaj
    Shivam Bhardwaj 2 months ago

    Could you please provide me with this presentation, I need to educate others with this!

  • Frater ZAAZAA
    Frater ZAAZAA 2 months ago +16

    Always start your videos with a cord.

  • savire.ergheiz
    savire.ergheiz 2 months ago

    Do the Alpaca too not just the OpenAI since why called it LLMs if there is only OpenAI supported

  • Ed Davis
    Ed Davis 2 months ago

    Can you link to the Colab example please?

  • Atiyeh Jamshidpour
    Atiyeh Jamshidpour 21 day ago

    Thanks, it is amazing

  • Chris M
    Chris M 2 months ago

    Very good, informative #RAG

  • Aman kumar
    Aman kumar 13 days ago

    I want to contribute on this project

  • Davis Rogers
    Davis Rogers Month ago

    Streaming AI Paradigm Shift 2023
    So Exciting &
    So Scary
    I mean we really need to Drop Approximations when it comes to Math, and Get the Exact Solution in Front or you can create a Monster on the Molecular level.
    Remember This and call me when realize what I am WARNING you about please.

  • Ahmed Mohammed
    Ahmed Mohammed Month ago

    wow thank you

  • nurkle blurker
    nurkle blurker 2 months ago +1

    Does langchain work with weaviate?

  • purpleAiPEy
    purpleAiPEy 21 day ago

    Damn… type safety on the output.. what the heck is coding now? What does this even imply what theeee

  • The transfer Account


  • Omar Elmady
    Omar Elmady 4 days ago


  • Helpmathbit
    Helpmathbit 20 days ago

    Amomg all the presentations in Clip-Share this one seems the most likely done by crooks trying to take your money. It's just ridiculously bad quality. F the Clip-Share algorithm.

  • Helpmathbit
    Helpmathbit 20 days ago

    Amomg all the presentations in Clip-Share this one seems the most likely done by crooks trying to take your money. It's just ridiculously bad quality. F the Clip-Share algorithm.