> Anthropic's more technical users inherently understand how LLMs work.

good (if superficial) post in general, but on this point specifically, emphatically: no, they do not -- no shade, nobody does, at least not in any meaningful sense

▲

omnicognate 4 days ago | parent | next [-]

Understanding how they work in the sense that permits people to invent and implement them, that provides the exact steps to compute every weight and output, is not "meaningful"?

There is a lot left to learn about the behaviour of LLMs, higher-level conceptual models to be formed to help us predict specific outcomes and design improved systems, but this meme that "nobody knows how LLMs work" is out of control.

▲

recursive 4 days ago | parent [-]

None of that is inherent, and vanishingly few of Anthropic's users invented LLMs.

▲

omnicognate 4 days ago | parent [-]

What is "inherent" supposed to mean here?

LLMs are understood to the extent that they can be built from the ground up. Literally every single aspect of their operation is understood so thoroughly that we can capture it in code.

If you achieved an understanding of how the human brain works at that level of detail, completeness and certainty, a Nobel prize wouldn't be anywhere near enough. They'd have to invent some sort of Giganobel prize and erect a giant golden statue of you in every neuroscience department in the world.

But if you feel happier treating LLMs as fairy magic, I've better things to do than argue.

	▲	recursive 4 days ago \| parent [-]
		Inherent means implicit or automatic as far as I understand it. I have an inherent understanding of my own need for oxygen and food. I don't have an inherent understanding of English, although I use it regularly. Treating LLMs as fairy magic doesn't make me feel any happier, for whatever it's worth. But I'm not interested in arguing either. I never intended to make any claims about how well the principles of LLMs can be understood. Just that none of that understanding is inherent. I don't know why they used that word, as it seems to weaken the post.

▲

lukev 4 days ago | parent | prev | next [-]

If we are going to create a binary of "understand LLMs" vs "do not understand LLMs", then one way to do it is as you describe; fully comprehending the latent space of the model so you know "why" it's giving a specific output.

This is likely (certainly?) impossible. So not a useful definition.

Meanwhile, I have observed a very clear binary among people I know who use LLMs; those who treat it like a magic AI oracle, vs those who understand the autoregressive model, the need for context engineering, the fact that outputs are somewhat random (hallucinations exist), setting the temperature correctly...

	▲	kiitos 4 days ago \| parent [-]
		> If we are going to create a binary of "understand LLMs" vs "do not understand LLMs", "we" are not, what i quoted and replied-to did! i'm not inventing strawmen to yell at, i'm responding to claims by others!

▲

shloked 4 days ago | parent | prev | next [-]

I should've been clearer, but what I meant was language models 101. Normal people don't understand even basics like LLMs are stateless by default and need to be given external information to "remember" things about you. Or, what is a system prompt.

▲

kingkawn 4 days ago | parent | prev [-]

Thanks for this generalization, but of course there is a broad range of understanding how to improve usefulness and model tweaks across the meat populace.