Why can't the LLM just be on my laptop?

As you learned in the last lesson, LLMs have to do an enormous number of mathematical calculations at the same time to generate every single response.

That requires so much computing power that a regular laptop simply wouldn't be able to cope.

For example, GPT-4 has an estimated 1 trillion parameters. A single response requires billions of mathematical operations.

Your laptop CPU would take minutes to generate one reply. Data centres can do it in milliseconds!