If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

Main content

How chatbots and large language models work

Large Language Models like ChatGPT have remarkable abilities to generate content based on training data but do they have actual intelligence? Find out more about how LLM's and Chatbots work as we explore this question.

Featuring:
Cristóbal Valenzuela the creator of Runway
Mira Murati the CTO of OpenAI

Presented by: Code.org, ETS, ISTE, Khan Academy

Start learning at code.org today!

Stay in touch with us on social media:
• Twitter: https://twitter.com/codeorg
• Facebook: https://www.facebook.com/Code.org
• Instagram: https://instagram.com/codeorg
• TikTok: https://tiktok.com/@code.org
• LinkedIn: https://www.linkedin.com/company/code-org
• Medium: https://medium.com/@codeorg

Help make our work possible with a donation at http://code.org/donate!
Created by Code.org.

Want to join the conversation?

  • blobby green style avatar for user alisher.alihon
    It's really interesting to know about how AI works and how it is able to analyze all these tasks.
    (9 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user pshirt
    how does a AI understand the prompt given bye a human?
    (4 votes)
    Default Khan Academy avatar avatar for user
    • leafers ultimate style avatar for user JustClassa
      AIs are trained on "Large Language Models" are trained on readings, texts, wikipedias, webpages, human interactions. Typically billions of interactions. So... when the human prompts the bot, the bot takes the prompts and generates conversational connections, like neurons in a brain firing, that seem to, statistically, relate to your question. This is done through a process called Natural Language Processing which is a collection of highly advanced techniques to create connections between words from conversations such as the one that I am having with you now.

      One of the pitfalls of this, is that nothing that the AI generates is original. Everything in its database has happened before within its field of study. So you have to be careful when using the information that it provides you in a professional setting.
      (3 votes)
  • leafers sapling style avatar for user THE GIRL KNOW IT ALL01
    ok i know a lot but not this
    (4 votes)
    Default Khan Academy avatar avatar for user
  • male robot hal style avatar for user KEVIN
    link removed

    My original post- ca. early Oct 2023

    What makes us - collectively- think that AI can actually recreate or create from scratch, prose such as Shakespeare wrote, or by extension, symphonies such as Mozart composed, or theories, such as Faraday developed, or see into the subconscious, such as Freud postulated, or write poems, such as those by Maya Angelou? Just because the program, AI, can mimic the procedures that these people used to create their works, that doesn't mean that what AI spits out will have any value. It will always and only be a copy, a derivative - in the non-mathematical sense - and an imitation that appropriates the labor and hard wrought creativity of others.

    Can AI create by candlelight, or with bombs going off close by, or under religious persecution, or without food or water or shelter? (Do you know what? It probably can. Those very human experiences/conditions don't shape how AI gets from point A to point B. That is why, from the purely, let's say, Shakespeare prose vantage point, what AI presents to us reads as hyper real, which determines that it is hyper fake/false.)

    I appreciate that KA is putting this in front of us, but at the moment, this is snake oil. Let's see.
    (4 votes)
    Default Khan Academy avatar avatar for user
  • blobby green style avatar for user Donald Vonderheide
    How does the ai understand the prompt given by a person?
    (3 votes)
    Default Khan Academy avatar avatar for user
  • duskpin sapling style avatar for user juliannagonzalez425
    how does a AI understand the prompt given bye a human?
    (2 votes)
    Default Khan Academy avatar avatar for user
    • duskpin tree style avatar for user ✿ Tati ✿ stan skz & newjeans
      When an AI receives a prompt from a human, it uses a combination of natural language processing (NLP) techniques and machine learning algorithms to understand and process the input. NLP helps the AI analyze the structure, context, and meaning of the text. However, if there are spelling mistakes or missing punctuation, it can sometimes affect the AI's understanding and response. The AI may still attempt to provide a relevant answer, but it's important to ensure clear and accurate communication to get the best results. Hope this helps! :)
      (4 votes)
  • blobby green style avatar for user ibad.2551
    what is AI and its full form
    (2 votes)
    Default Khan Academy avatar avatar for user
  • aqualine sapling style avatar for user sanvi.krishna25
    how complex will ai get
    (2 votes)
    Default Khan Academy avatar avatar for user
  • leafers sapling style avatar for user THE GIRL KNOW IT ALL01
    this is so nice😀
    (1 vote)
    Default Khan Academy avatar avatar for user
  • piceratops tree style avatar for user Cory Agnew
    I have an idea for a form of Trauma therapy that would allow a victim to confront a virtual chatGPT AI able to safely allow healing to occur that up until now has been completely ill advised due to further trauma occurring, the responses can be limited to what allows for the person to heal through healthy communication
    (1 vote)
    Default Khan Academy avatar avatar for user

Video transcript