> Every single problem does not need to be solved using an LLM How do you know if a problem is suitable for transformers/LLMs unless you try? We have this grea

> Every single problem does not need to be solved using an LLM How do you know if a problem is suitable for transformers/LLMs unless you try? We have this great generalized architecture, I would hope people throw everything at it and see what sticks, because what's the downside? Less focus on bespoke predictive models? Oh no.