In a single case from the examine cited by AP, when a speaker described “two different women and one girl,” Whisper added fictional textual content specifying that they “had been Black.” In one other, the audio mentioned, “He, the boy, was going to, I’m unsure precisely, take the umbrella.” Whisper transcribed it to, “He took a giant piece of a cross, a teeny, small piece … I’m certain he didn’t have a terror knife so he killed a variety of individuals.”
An OpenAI spokesperson advised the AP that the corporate appreciates the researchers’ findings and that it actively research how one can cut back fabrications and incorporates suggestions in updates to the mannequin.
Why Whisper confabulates
The important thing to Whisper’s unsuitability in high-risk domains comes from its propensity to typically confabulate, or plausibly make up, inaccurate outputs. The AP report says, “Researchers aren’t sure why Whisper and comparable instruments hallucinate,” however that is not true. We all know precisely why Transformer-based AI fashions like Whisper behave this fashion.
Whisper relies on know-how that’s designed to foretell the subsequent most probably token (chunk of knowledge) that ought to seem after a sequence of tokens supplied by a person. Within the case of ChatGPT, the enter tokens come within the type of a textual content immediate. Within the case of Whisper, the enter is tokenized audio knowledge.
The transcription output from Whisper is a prediction of what’s most probably, not what’s most correct. Accuracy in Transformer-based outputs is usually proportional to the presence of related correct knowledge within the coaching dataset, however it’s by no means assured. If there may be ever a case the place there is not sufficient contextual info in its neural community for Whisper to make an correct prediction about how one can transcribe a specific phase of audio, the mannequin will fall again on what it “is aware of” in regards to the relationships between sounds and phrases it has realized from its coaching knowledge.
