They’re really not. They seem like it. We have a collective GPT-Mann amnesia where we read something summarised that we understand, recognise it for the bullshit that it is, and then promptly trust the next prompt when it summarises something outside our expertise.
LLMs seem to do better when summarising summaries, e.g. of classical works. Someone else has whittled down the meaning. The LLM puts it into simpler terms.
They’re really not. They seem like it. We have a collective GPT-Mann amnesia where we read something summarised that we understand, recognise it for the bullshit that it is, and then promptly trust the next prompt when it summarises something outside our expertise.
LLMs seem to do better when summarising summaries, e.g. of classical works. Someone else has whittled down the meaning. The LLM puts it into simpler terms.