To me, the bullishness stems from the observation that current gen of LLM an plausibly lead to a route to human level intelligence / agi, because they have this causal behavior: (memory/context + inputs) --> outputs. Vaguely similar to humans.