The second was because the benefits of "Expert Systems" were oversold, and didn't prove to be as beneficial as expected.
There will be a heavy correction in the market caps of companies that bet heavy on AI, as the promises don't live up to the hype. However, in this case, there clearly is benefit to using LLMs, the cost of training has proven to be prohibitive. Slower and more prudent approaches, with better hardware will allow progress to continue. If nothing else, it'll be a mild AI winter.