1 paper · 1 filter
Biel Tura Vecino, Subhadeep Maji, Aravind Varier +11
The use of audio recordings of human speech to train LLMs poses privacy concerns due to these models' potential to generate outputs that closely resemble artifacts in the training…