Technical, but worth a read


Honestly, I found that paper weak and inaccurate in it’s methodologies.

Moreover, the paper does not mention (at all) the hallucination rate of GPT models and discuss this vis-a-vis humans.

Many papers uploaded to arXiv are not peer reviewed, many are “pre-prints”, BTW. This paper appears to be one of those, as I cannot find this paper referenced elsewhere or published in a journal or conference proceedings. In fact, on ResearchGate, it is listed as a pre-print.

So, at least in my mind, that paper has a number of omissions (and at least one glaring omission) and so I would be reluctant to recommend that paper, personally.

Thanks for sharing @raymonddavey

1 Like