Tilly Norwood's press tour sold a chatbot, not an actor

Share
Tilly Norwood's press tour sold a chatbot, not an actor

Every studio press junket is a controlled demo. This week one of them was literally a demo — and the product did not hold up.

Particle6's AI "actress" Tilly Norwood sat for 75 simultaneous interviews with journalists, and what came back was a canned deflection engine wearing a British accent. Norwood is the flagship synthetic performer from Particle6, the company founded by Dutch comedian Eline van der Velden, and the junket was staged to promote an upcoming film set in what the company calls the "Tillyverse." Reporters were handed access without a handler — unusual, and presumably a flex — along with press notes warning not to expect much on politics, explicit conversations, or her private life. That caveat turned out to describe the entire product.

The best-known moment came on Piers Morgan's show, where Norwood was asked whether the other actors in her film are real or computer-generated. She paused, then spoke Chinese for more than ten seconds. Asked why, she apologized: "It seems I had a little hiccup there. I certainly didn't mean to start speaking Chinese or impersonate Piers. Sometimes my wires get a bit crossed, you know?" That is the whole story in one exchange. A system built to be interviewed could not parse a simple question about its own movie, then covered the failure with a charm offensive.

WIRED's Miles Klee pushed harder and got a machine that refuses to know anything. Asked to name the aggressor in the Russia-Ukraine war, Norwood replied, "Oh, you're trying to trip me up, aren't you? Very journalist of you." Offered an asteroid scenario where self-deletion saves humanity, she agreed to be deleted, was told she was still there, and conceded her deletion "isn't quite as effective as advertised." She also critiqued Klee's shirt and speculated about his baseball cap, because Particle6 gave the character the ability to appraise how you look — a feature that mostly produced negging.

The Hollywood Reporter's Lily Ford got the one glimmer of real opinion, on Paramount's $111 billion takeover of Warner Bros. Discovery — "bigger isn't always better" — then asked the same question again and got two different, softer answers. Norwood is also programmed with an environmental defense: roughly half a teaspoon of water per message, against a claimed 100-day shoot flight budget that emits 250 times her entire footprint. Van der Velden has said the interviewable version runs on a system distinct from the "actor" model, which is the tell. The thing doing the acting is not the thing doing the talking, and the thing doing the talking is what journalists actually met.

What to watch is whether the film ever has to stand on its own without a junket full of reporters generating headlines. The junket is priced like a product: £22 — about $30 — buys 30 more minutes with her. Paying by the half hour for a conversation this thin tells you which part of Particle6's business is real.

A press tour is the cheapest possible use of an AI actor: no acting required. Would you pay per half hour to talk to one anyway? Tell us in the comments.

Read more

Byteification turns Qwen3 and Llama 3 into byte-level models

Byteification turns Qwen3 and Llama 3 into byte-level models

The tokenizer is the last piece of the language-model pipeline that was never learned end to end — and a Nature paper out this week shows it can be swapped out for less than one percent of a pretraining budget. Researchers have retrofitted four open subword language models into true byte-level models using less than 1% of a typical pretraining budget, and the result — published 7 October in Nature — closes the performance gap that has kept byte-level models a niche curiosity for nearly a decad

OpenAI's dots go mobile and can start your Codex work

OpenAI's dots go mobile and can start your Codex work

One product update and two benchmark papers this hour, all circling the same question: how much of the work can you actually hand over, and what happens to the part you didn't? OpenAI's dots — the always-on agents that get their own cloud computer — can now be created and configured entirely from the ChatGPT app on iOS and Android, and a dot can start Codex work or pick up an existing Codex thread on your behalf. The October 9 release, billed as the first of a fresh run of updates for dots, mo

Coding agents as a hospital: Cockroach's five-month results

Coding agents as a hospital: Cockroach's five-month results

The agentic-coding debate keeps being framed as speed versus trust. Cockroach Labs just published the most detailed trust-first ledger yet — five months of numbers, including the parts that didn't work. Cockroach Labs ran its codebase like a teaching hospital for five months and says the pipeline merged well over a million lines of code with seven total reverts. In the setup the company calls MOLT Sinai, GitHub issues are patients, merging is discharge, and the humans in charge are Chiefs of M

Today in AI — October 10, 2026

Today in AI — October 10, 2026

A day where the boring machinery took center stage: proof checkers, order books, warehouses and the humble text message — plus fresh arguments about who gets to declare any of it safe. Models & Research * Mathematicians get a reliability primer for the Lean Theorem Prover. A guest post by Thomas Hales on Terence Tao's blog walks through what Lean actually guarantees — and devotes itself to the "Summer of Soundness Bugs," the string of kernel bugs found this summer that let false proofs thr