Do not bet against the sheer expressivity of the transformer LLM's design. They fill graveyards with those kinds of bets.
closest analogy I can see is a RISC chip -- it can't do fancy math directly, but it can get there by doing simple math
LLMs can't do real AGI but can get pretty close
alt Hacker News
closest analogy I can see is a RISC chip -- it can't do fancy math directly, but it can get there by doing simple math
LLMs can't do real AGI but can get pretty close