In May 2025, Jackie gave a talk at The One About AI Agents and Imposters, a community event at Lorong AI (WeWork, 22 Cross St) in Singapore. It was pitched at a 200–300 technical level: enough to explain how the technology works without losing the room.
The talk had four parts:
- What is a deepfake? — synthetic media generated with machine learning, across audio, video, images, and text. Deepfakes aren't new, but it has become far easier to make one.
- Live demo — he cloned a voice in the room in real time, twice.
- So what? — voice fraud, social engineering, and impersonation attacks.
- What's holding back mass fraud for now — real-time speech-to-text and good text-to-speech both exist, but real-time speech-to-speech with matching cadence and latency is still brittle. That gap is a temporary reprieve, and it's closing.
It ended on what people and organisations can do about it — verification challenges, call-back protocols, and out-of-band confirmation — rather than just on how scary it is.
The full write-up is at Real-time Voice Cloning in the Age of GenAI, and the slides are here.