Despite my positive experience between building systems based on intent recognition and how much better LLMs are than “1000 monkeys”, it seems like the two exa

Despite my positive experience between building systems based on intent recognition and how much better LLMs are than “1000 monkeys”, it seems like the two examples we have of LLM backed assistants - Google and Amazon - that it made them worse from reports. I don’t know why that is from a technical level.