Jonathan Fishbein
@jfishbein.bsky.social
Great. Another f’ing social media account.
Can’t it be both?
The responses range from, LLMs just use fancy math to predict the next token, to LLMs are literally magic and no one knows how they work.
I'd love to be proven wrong about LLMs. Can someone explain how LLMs answers a question correctly without an approximate answer in its training data or doing an external lookup to an existing knowledge base.
March 26, 2025 at 1:55 AM
Can’t it be both?