language models
also referred to as: language model
5 statements across 1 episodes · 2 bullish · 0 bearish · 1 people on the record · first statement Sep 9, 2024 by Nick Frosst · across every show →
Everything said about language models, oldest first
Sep 9, 2024 neutral
Sep 9, 2024 positive
Frosst: Fine-Tuned LLMs Generalize to Unseen Request Types
“If you train it on just those conversations, and you It's never been asked to do a summary. It will still be a, do a good job when you ask it to do a summary. So it generalizes to request is never asked, never been asked in the training data.”
Sep 9, 2024
Frosst: True intelligence requires agency and experimentation, not just language
“And you don't learn by just seeing as many examples of this as possible. You learn by interacting with the world. You learn by having agency in it. You learn by, you know, making predictions of the world and experiments in the world and validating them. And th…”
Sep 9, 2024 neutral
Frosst: ChatGPT's capabilities were already possible with earlier LLMs
“Everything that you can do with a language model right after ChatGPT, you could pretty much do with a language model before. It just sucked. Sucked in that it was, like, really hard to do. You had to spend a lot of time prompt engineering, and it was, like, re…”
Sep 9, 2024 positive
Frosst: General-purpose language models replaced task-specific models across NLP tasks
“Language models were this new thing, whereas if you wanted a language model to, you know, extract numbers from a
PDF or summarize a call transcript or doing something like that.
This was the first time where you were going to be best served by a general purpos…”