How can we help?

How do the Large Language Models powering Khanmigo work?

Updated

Khanmigo is powered by GPT-4, a Large Language Model developed by OpenAI. Large Language Models (LLMs) are a type of artificial intelligence trained to understand and generate human language. They work by analyzing massive amounts of text from books, websites, articles, and more, to learn the patterns and structures of language

The more data an LLM is trained on, the better it becomes at recognizing patterns in language. Once trained, it can generate new text by predicting the most likely next words or phrases based on the patterns it has learned.

Educational applications and limitations

LLMs have many practical applications in education. They can act like a personal tutor, helping you understand subjects by generating clear explanations, definitions, and relevant examples. If you're learning a new language, you can practice grammar and vocabulary through conversation. They can also guide you through problem-solving steps rather than simply giving you answers.

However, language models aren’t perfect and can sometimes make mistakes. This is partly because they are trained on large datasets that may include biases or inaccuracies. Additionally, language itself is complex and ever-evolving, which can make it challenging for models to fully understand nuance or context. If the context of a question or conversation is unclear, the model may misinterpret it or respond inaccurately. Sometimes, language models may even produce off-topic or inappropriate responses. That’s why it’s important to use them thoughtfully, and with an understanding of their limitations.

Privacy and safety

When using Khanmigo or any LLM,  it’s important not to include personally identifiable information (PII) in your questions or conversations. This includes full names, addresses, phone numbers, school names, student IDs, or personal details about yourself or others. While these models are designed with safety in mind and are generally considered safe to use, keeping sensitive information out of your prompts helps protect your privacy and ensures a safer online experience. LLMs can provide helpful responses and educational support without needing any personal information from you.

Getting the best results from Khanmigo

Here are some ways you can get the highest-quality responses from Khanmigo or any LLM:

  1. Be clear and specific: When asking a question or inputting a problem into a language model, be as clear and specific as possible. Provide all the necessary information and context to help the model understand what you're asking for. This will help ensure that you receive an accurate and helpful response.
  2. Check multiple sources: While LLMs can be a great resource for learning and understanding, it's always a good idea to check multiple sources to verify the information. Don't rely solely on the response from a language model, but also consult textbooks, articles, or other trusted sources.
  3. Understand the limitations: Remember that LLMs can make errors and may have knowledge gaps. It's important to understand their limitations and be aware of the potential biases or errors in the data they are trained on. Use them as learning tools to supplement your learning, not as the sole source of information.
  4. Practice critical thinking:  Don't just accept responses at face value. Practice critical thinking and try to understand the reasoning behind the response. This will help you better understand the concept and improve your problem-solving skills.

By following these guidelines, you can effectively use LLMs to enhance your learning while staying aware of their limitations. Understanding how these tools work helps you use them responsibly and get the most educational value from your interactions.