ChatGPT fills gaps like a confabulator, not a hallucinating mind
A Perspective article argues that large language model errors more closely resemble confabulations, while speech recognition errors, such as unsupported Whisper transcriptions, are functionally closer