AI speech generator 'reaches human parity' – but it's too dangerous to release

Microsoft's new AI, VALL-E 2, can reproduce human voices from a few seconds of audio, reaching human parity in speech generation. Despite its capabilities, Microsoft is not releasing VALL-E 2 to the public due to potential misuse, such as voice spoofing or impersonation.

• VALL-E 2 achieves human speech parity.
• The AI can mimic voices quickly.
• Audio quality affects VALL-E's output.
• It uses LibriSpeech and VCTK datasets.
• Concerns over misuse limit its release.