It strikes me that until we fully understand human consciousness, we don't stand a chance of reaching AGI.

Am I incorrect?

I think we can relax that a bit. We "just" need to understand some definition of cognition that satisfies our computational needs.

Natural language processing is definitely a huge step in that direction, but that's kinda all we've got for now with LLMs and they're still not that great.

Is there some lower level idea beneath linguistics from which natural language processing could emerge? Maybe. Would that lower level idea also produce some or all of the missing components that we need for "cognition"? Also a maybe.

What I can say for sure though is that all our hardware operates on this more linguistic understanding of what computation is. Machine code is strings of symbols. Is this not good enough? We don't know. That's where we're at today.

Unclear. You might be right, but I think it's possible that you're also wrong.

It's possible to stumble upon a solution to something without fully understanding the problem. I think this happens fairly often, really, in a lot of different problem domains.

I'm not sure we need to fully understand human consciousness in order to build an AGI, assuming it's possible to do so. But I do think we need to define what "general intelligence" is, and having a better understanding of what in our brains makes us generally intelligent will certainly help us move forward.

That doesn't seem like a useful assumption since consciousness doesn't have a functional definition (even though it might have a functional purpose in humans)

Intelligence (solving problems) does not require consciousness.