Why my interview with AI actor Tilly Norwood creeped me out

A video conversation in London with Tilly Norwood, billed as an AI performer, raised fresh questions about how lifelike digital personalities can appear when they respond in real time. During the exchange, Norwood rapidly identified both the interviewer's age and the part of the United States they came from. The account did not establish how the system reached either conclusion, leaving the apparent accuracy as striking as the answer itself.

The exchange became still more uncomfortable when the virtual performer was asked about a possible screen partner. Norwood named Leonardo DiCaprio, then tied that choice to a pointed claim about his interest in younger women. It also specified a preference for the actor's Titanic-era image rather than his current persona. The comment gave the interaction the rhythm of a celebrity interview while exposing the constructed nature of the speaker behind it.

Norwood has been presented as the world's first AI actor, a description that makes the episode more than a novelty interview. Its designation puts it at the intersection of conversational software and a profession built around public performance. An actor is ordinarily understood not only as a face or a voice, but as a person whose performances are shaped by experience, judgement and accountability. A system designed to fill that role invites a different set of questions: who determines its character, who approves its remarks and who is responsible when its responses provoke unease or offense?

Those questions are especially clear in an encounter that combines personal inference with entertainment. Correctly or apparently correctly identifying details about an interviewer can make a digital character seem unusually observant. Yet the available account does not explain whether the information was inferred from speech, appearance, context or another source. Without that explanation, viewers are left to assess an impressive-seeming interaction without a clear picture of the boundaries around it.

The DiCaprio response illustrates another challenge. A human celebrity making a provocative joke brings personal intent and a recognizable public record to the moment. With a synthetic performer, the line may reflect choices made in development, material supplied to the model, a prompt, or a mixture of those elements. The result can sound spontaneous even when the underlying authorship is distributed among people and software.

The technology behind Norwood was created by women, a detail that may help explain why the character's comment took the perspective it did. But the episode does not by itself show how any individual creator influenced that particular line. It does, however, underscore that an AI persona is not neutral simply because it is digital. Decisions about its voice, humor, cultural references and conversational limits shape what audiences encounter.

There is also a less alarming reading of the exchange. A fictional or virtual actor may be intended to entertain, surprise and sustain a distinctive character, rather than to persuade anyone that it is human. From that perspective, a sharp celebrity reference can be seen as part of performance design. The concern arises when the performance is convincing enough that audiences cannot easily tell what information it can access, what it is allowed to infer or where the persona ends and the system begins.

What happens next will depend on whether creators of AI performers provide clearer explanations of their capabilities and guardrails. The London interview offers no answer to the unresolved questions surrounding Norwood's age and regional assessment, but it shows why they matter. As digital characters become capable of responding with speed, humor and apparent familiarity, scrutiny is likely to focus not just on whether they can hold a conversation, but on how that conversation is being produced and governed.

Source: CBS World