A new AI-driven persona named Tilly Norwood recently made a glitchy debut during a live interview in London. The event, featuring actor Tom Conti, was marred by technical failures and unexpected linguistic shifts.

Advertisement

A sudden pivot to Cantonese in front of live audiences

The most jarring moment of the debut occurred when the AI character, Tilly Norwood, spontaneously began speaking Cantonese despite no prior programming or contextual reason to do so. as the report notes, this linguistic shift happened in front of a live audience and a camera crew, causing immediate confusion. When host Piers Morgan questioned the sudden change, the AI apologized, describing the switch as "completely random" and "not intentional." The AI even characterized the error as "a bit of a curveball even for me," though its tone remained notably flat and uninspired throughout the incident.

Tom Conti's struggle with sitlted and evasive AI responses

During the staged interview, veteran actor Tom Conti encountered significant technical hurdles that disrupted the flow of the conversation. The report highlights several specific issues:

  • Pronounced lag: The AI exhibited a noticeable delay before responding to questions.
  • Monotone delivery: Responses were often stilted and lacked human-like inflection.
  • Evasive digressions: Instead of answering direct questions about AI actors, the character spoke at length about "new storytelling techniques."

When Conti attempted to claarify the inquiry, the AI offered a vague apology for a "hiccup" before making disconnected remarks about "hybrid productions" and viewer terminology—topics that the interviewer had not even introduced.

Particle6's focus on real-time dialogue and digital twins

In response to the technical difficulties, the developers at Particle6 defended the demonstration as a way to showcase cutting-edge technology capable of generating humanlike dialogue in real-time. The company cited the significant power consumption required to match a human-like conversational cadence as a key technical challenge in their development. Despite the public stumble, Particle6 is continuing to push forward with a diverse portfolio, including work on nanomaterials and the development of AI-based digital twins for educational environments.. However, the report suggests that this failure to provide contextually relevant output may damage the company's credibility among industry observers and potential partners.

The unresolved cause of the AI's unprogrammed language switch

The event has ignited a broader debate regarding the readiness of autonomous dialogue agents for mainstream media exposure. One of the most pressing open questions is how Particle6 will prevent AI systems from unintentionally producing culturally sensitive or linguistically inappropriate content. Because the Cantonese switch appeared to have no relation to the conversation, it remains unknown whether this was a localized glitch or a systemic flaw in the AI's conversational engie. Furthermore, the source does not confirm if any failsafes were in place to prevent such unpredictable deviations from human expectations during live broadcasts.