How do large language models such as ChatGPT work?  Code.org has a
recent video on this topic in which they use a Markov chain to
illustrate the concept of statistical modeling of language.  You
can watch the video here: https://youtu.be/X-AWdfSFCHQ

Aditya Dewan and I have created an online Markov Chain Demo that allows
you to experiment with this idea in your browser.  You can input text,
generate a bigram, trigram, or tetragram dictionary, and then generate
new text using that dictionary.  Try it out here:
https://www.cs.cmu.edu/~dst/MarkovChainDemo

Click on the "About/Help" button for a brief tutorial and an explanation
of how this relates to large language models.

We're still adding features to the demo.  Feedback, bug reports, and
suggestions for improvement would all be appreciated.

-- Dave Touretzky