Imagine your friend is telling you a story, but sometimes they make up parts that didn’t happen, that’s like what LLM hallucinations are. Now, Marc Andreessen and Lex Fridman are two people who talked about this on a fun kind of show.
What Is an LLM?
An LLM, or large language model, is like a super-smart robot that can write stories, answer questions, and even chat with you, just like your friend. But sometimes it gets confused and makes up things that aren’t real.
Why Does It Make Up Things?
Think of the LLM as a kid who’s trying to remember a long list of numbers. If they get distracted or tired, they might say something that wasn’t on the list, that's like hallucination. Marc Andreessen and Lex Fridman talked about how these models can be amazing but also have moments when they “lie” by adding things that aren't true.
So, just like your friend sometimes makes up parts of a story, an LLM might too, but it’s all part of the fun!
Examples
- A teacher thinks a student cheated on a test, but the student just guessed the right answer.
Ask a question
See also
- How Does Exception vs Errors | Chris Lattner and Lex Fridman Work?
- How Does AI Expert Lex Fridman Weighs in on Simulation Theory Work?
- Why LLMs hallucinate | Yann LeCun and Lex Fridman?
- How AI will change video game development | Tim Sweeney and Lex Fridman?
- 5 cm to inches?