Comment by thaumasiotes
10 days ago
As I just mentioned in the comment you're responding to, the way you convert an LLM into an equivalent Markov chain is by doing nothing, since it already is one.
> You may as well tell that LLM and a hash table is the same thing.
No. You may as well say that a hash table and a function are the same thing. And this is in fact a common thing to say, because they are the same thing.
An LLM is a significantly more restricted object than a function is.
> LLM is a lookup table which tells you "if the last N tokens were s_1, s_2, ..., s_N, then the following token will be t_1 with probability p_1, t_2 with probability p_2, etc...".
No, LLM is not a lookup table for all possible inputs.