I mean, the problem is LLMs. If I were to replace “school” with “biomedicine” or “protein folding,” then that would be clearly wrong. However, the AI used in those fields are machine learning models, not LLMs
Even LLMs have use cases were they are a good tool - fixing grammar, low risk translations, etc. Unconstrained chatbots with models that are way past the point of diminishing returns is just not one of them.
I mean, the problem is LLMs. If I were to replace “school” with “biomedicine” or “protein folding,” then that would be clearly wrong. However, the AI used in those fields are machine learning models, not LLMs
Even LLMs have use cases were they are a good tool - fixing grammar, low risk translations, etc. Unconstrained chatbots with models that are way past the point of diminishing returns is just not one of them.
Yes, absolutely correct. I suppose I should say that dependence on LLMs is the problem, more than anything.