Hype rating
Balanced/Informed
AI usage
Generative
Who benefits?
Companies, for Chatbots, automatic synthesized interpreting…
Voice-impaired individuals
What concerns does this bring up?
Voice cloning, deepfakes for profit and all kinds of nefarious purposes
Intellectual property rights over own voice
Privacy
Additional comments
Super interesting article with a whole list of ways to tell AI voice synthesizer from a true human voice and know if you’re interacting with a real human (I couldn’t tell the difference in the sample clips provided):
To ID speech-synthesized voice:
Listen for:
- Breathing: fake or real?
- Mistakes being made? Too bright, too perfect?
- Unnatural speech patterns such as irregular pauses and awkward phrasing when playing audio
- Strange artefacts like distortions and mismatched background noise
- limited variations in volume, cadence and tone because voices that are cloned from just a few seconds of audio may not have the full range of a human voice
- natural inflection and breathing
- Sentence-level prosody:
- Accentuation (emphasis words are given in a sentence)
- Intonation (change in pitch of the voice across a sentence) depending on whether it’s a statement, a question, etc.
- Phrasing (the way a sentence is broken up). Example “when danger threatens, children call the police” has a very different meaning from “when danger threatens children, call the police”
- Listen for “prosodic deaccenting” (where there would be human accentuation based on unspoken real-world knowledge). Example: Q: Has John read Hard Times yet? A: John doesn’t LIKE Dickens.
To know if you’re interacting with the real human you think you are:
- At home, decide family passwords
- At work, don’t take action just because you get a voice message from a CEO
- Ask personal questions, such as favorite song.
- Best: “I’ll call you back.” Call on the number you have for them.
- Ask it to shout or scream.
Note: LLM will get better and better even at this. ChatGPT’s voice function can already laugh, whisper, be interrupted and then continue what it was saying. It can also remember everything you ever told it.