19949 shaares
Given an audio waveform, researchers can now produce a virtually identical version that makes speech-recognition software transcribe something else entirely.
Given an audio waveform, researchers can now produce a virtually identical version that makes speech-recognition software transcribe something else entirely.
