🎙️ EP 53: AI Just Learned to Forget You, Literally. Not block it. Forget it.

Imagine telling AI to forget your voice and it actually does. This isn’t science fiction. Researchers just taught Meta’s VoiceBox model how to unlearn specific people, making them unclonable even when directly prompted.

In this episode, we unpack one of the most important shifts in AI safety and control—alongside a stack of breaking stories that show just how weird and wild the AI world is getting.

We’ll talk about:

• How “machine unlearning” lets speech models forget your voice on purpose (and why that kills deepfakes)
• The bizarre moment ChatGPT triggered a 1990s COM port popup—and Reddit lost its mind
• Why Meta still hasn’t released VoiceBox, and what they're possibly hiding
• Mistral’s new Voxtral model that just blew the doors off Whisper and ElevenLabs—for half the co ...