Dyslexia Is Not Gastrointestinal
The stochastic parrot of OE can be easily led down the garden path with words
The recent predictable and tragicomic finding that researchers were able to plant a fake disease on a preprint server and find it later confidently presented in various LLMs only goes to show that these systems don’t have any knowledge of reality informing their guardrails and are just silly text extrusion systems.
To test OpenEvidence (OE), I entered the nonsense disease entity (“bixonimania”), which OE did not recognize, to its credit I suppose, although it may be just a matter of time (bears later confirmation). However, I then added another term to my fake patient’s presentation (“colexia”), whereupon I got this wonderful response:
This post is for paying subscribers only
Already have an account? Sign in.