daily reminder llms aren't that complicated

If there are good teachers like 3b1b and Elliot then everything is easy

ikr

Hi 👋 We look a little different. A bold new brand for a bold new era — but not only that. More on what we’ve been up to…

But unbelievable inefficient

yes, but this part didn't cover parallel processing and kv cache

we just taught math to gossip

its just matrix (llm experts plz dont attack me)

But Elliot, we don’t understand it, it’s sentient! The “top” researchers say… 🤣😂🤣

exactly

what confused me at the start was thinking that modern llm's use the same seq2seq architecture as the one from Attention is All You Need, when in reality, the auto-regressive transformer is much simpler

Never deleting this app.

<3