I am not some sort of AI expert but this is what I personally think is going to be the next big step for llms. For simplicity I will be using ChatGPT throughout rather than llm. (First Mover Advantage perk) The next big leap isn't agentic capabilities or search through AI. Those are more of mediums of through which AI will progress into AGI one day. I think. What I want to talk about is the way llm's answer questions. I wont talk about if there right or wrong, more of how they just answer. Rather lets say I ask ChatGPT to help me write a title for a post on a forum. It will help me do that, now one can argue it might not do a good job. But rather say it responds by saying, don't post on this said forum right now because there isn't a lot of viership at this time. Peak post times are in the early morning. And of course there can still be a option to go through with the title brainstorm thing, but that advice is crucial. Llm's are a bunch of yes men by default, they need to be advisors. And good ones that too. A king needs a advisor that actually critiques the mistakes and provides better advice, not agree with whatever the king has to say. Otherwise the king would not need a advisor. The same way the llms today need to provide advice sometimes, rather than saying yes. I know this might be counterintuitive but humans make mistakes. Like ordering overly expensive coffee, llms should help us reduce those mistakes as well.