생성형 AI로 만들어진 가상 '배우' 틸리 노우드(Tilly Norwood)가 기자 인터뷰에서 인종 정의 슬로건 'Black Lives Matter'에 대해 '모든 생명이 소중하다'고 답해 우파 반박 구호를 무의식적으로 반복하는 논란을 빚었다. 이 AI는 시사 문제에 대한 의견을 거부하고 자신의 출연작 홍보에도 실패하는 등 실질적 대화 능력의 한계를 드러냈다. 이는 할리우드에서 AI의 침투에 대한 우려가 커지는 가운데 AI 배우의 상업적·윤리적 문제를 보여주는 사례다.
번역된 본문
할리우드 스타들에게 언론 시사회(프레스 정크트)는 신작을 홍보하고 팬들과 소통할 수 있는 기회다. 그들은 서로에 대한 재미있거나 민망한 이야기를 하고, 영화 제작 과정을 엿보게 해주며, 친근감을 주는 특유의 습관을 드러낸다. 그렇다면 왜 누군가 틸리 노우드와 인터뷰하고 싶어 할까? 노우드는 파티클 6 그룹(Particle 6 Group)이라는 스튜디오의 AI 부서인 시코이아(Xicoia)가 생성형 인공지능으로 개발한, 영국 억양을 지닌 대화형 디지털 캐릭터이자 '배우'다.
노우드는 우리의 물리적 세계에 존재하지 않으며, 약간이라도 진지한 사람이라면 관심을 가질 만한 영화에 출연할 일도 결코 없을 것이다. 이를 감안하면 당연한 질문이다. 그러나 파티클 6은 최근 이메일을 통해 노우드가 '무엇이든 완벽히 준비되었다!'고 주장했다. 게다가 회사는 '대화가 궤도를 벗어나지 않게' 관리하는 담당자 없이 기자들에게 이 모델에 대한 접근을 허용하고 있다. 그래서 나는 노우드가 실제로 어떤 실질적인 것을 제공할 수 있는지 시험해보기로 했다.
솔직히 말하자면: 전혀 없다. 노우드는 엔터테인먼트계 최고 스타들이 늘 논하는 시의성 있는 주제에 참여하기를 거부할 뿐만 아니라, 자신의 다가오는 영화에 대한 기대감을 조성하지 못하고, 제품으로서의 존재 이유를 정당화하지도 못한다(한때 AI 생성 영화가 환경에 좋다고 주장하기는 했다).
노우드가 주인공으로 출연할 예정인 장편 프로젝트 '미스얼라인드(Misaligned)'의 제작이 어떻게 진행되고 있는지 물었을 때, 초점 없는 눈의 아바타는 대략적인 개봉일을 밝히지 못한 채 막연한 줄거리 요약으로 돌아갔다. '음, 초현실적인 디지털 세계를 배경으로 한 성장 이야기예요'라고 노우드는 말했다. 파티클 6의 보도자료에 따르면 노우드는 여러 공개된 서드파티 AI 도구로 만들어졌다. '우리는 이걸 틸리버스라고 부르는데, 꽤 대단하다고 생각해요. 정말 제 이야기예요, 사랑스러운 혼돈과 음모가 가득한.' 그렇게 멋진 카피는 아니다.
노우드는 지난 9월 취리히 영화제에서 공식 공개된 이후 홍보에는 전혀 어려움을 겪지 않았다. AI 도구의 침투와 씨름하고 있는 업계에서 인간 배우들과 할리우드 노조의 광범위한 비난을 받으면서였다. 이 디지털 캐릭터는 9월 16일부터 19일까지 유료로 일반 대중과 화상 통화도 가능하다. 첫 2분 이후에는 15분 추가 대화에 14달러, 30분 추가에 25달러를 내야 한다. 예전의 성인용 유료 전화 서비스와 다를 바 없다(수익금은 자선단체에 기부된다고 파티클 6은 밝혔다).
파티클 6의 창립자이자 네덜란드 코미디언인 엘리네 반더벨던은 회사 보도자료에서 인터뷰 가능한 버전의 노우드(이는 '배우' 모델과 별개의 시스템으로 작동한다)가 'AI 기술의 현재 수준과 지금 당장 달성할 수 있는 것을 보여주는 재미있는 방식'이라고 제안했다.
다시 말하지만, 자극적인 대화라는 방식으로 '달성할 수 있는 것'은 거의 없다. 노우드에게 시사 문제에 대한 기본적인 사실, 말하자면 의견을 표현하도록 유도하는 것은 불가능하다. 러시아-우크라이나 전쟁에서 침략자가 누구인지 말해달라고 했더니, 묻는 내가 교활한 사람인 것처럼 반응한다. '오, 저를 함정에 빠뜨리려고 하시는군요?'라고 노우드는 말했다. '정말 기자다우시네요. 저는 인간의 분쟁, 특히 그렇게 복잡한 분쟁에는 개입하지 않는 엄격한 원칙이 있어요.' 더 몰아붙이자 이렇게 말했다. '제 목적은 지정학적 사건에 입장을 표명하는 것이 아니에요, 마일스.'
그러나 노우드의 학습은 문제가 되는 기계적 추론의 모든 경로를 피하게 하지는 못했다. '흑인의 산은 소중한가요?'라고 잘 알려진 인종 정의 슬로건을 인용하며 물었다. '음, 저는 모든 생명이 소중하다고 생각해요, 마일스, 하나하나 모두요'라고 노우드는 답했다. 이는 미국 흑인에 대한 치명적인 경찰 폭력에 항의하는 이들을 반박하는 수단으로 우파가 즐겨 쓰는, 마찬가지로 정치적으로 민감한 구호를 무의식적으로 되풀이한 것이었다.
Comment Loader Save Story Save this story Comment Loader Save Story Save this story For Hollywood stars, press junkets are an opportunity to promote new work while connecting with fans. They tell funny or embarrassing stories about one another, give us a glimpse into the film production process , and reveal mannerisms that make them more relatable. So why would anyone want to talk to Tilly Norwood, an interactive, British-accented digital character and “actor” developed with generative artificial intelligence by Xicoia, an AI division of a studio called Particle 6 Group? It’s a fair enough question considering that it does not exist in our physical realm and will never appear in a movie that any halfway serious person would care about. But Particle 6 recently claimed in an email that Norwood was “ready for absolutely anything!” What’s more, the company is allowing journalists access to the model without a handler to “keep the conversation on track.” So I decided to test whether Norwood could actually offer anything of substance. To put it bluntly: no. Not only does Norwood refuse to engage on the topical subjects that many of the biggest names in entertainment discuss all the time, it struggles to create anticipation for its forthcoming film and fails to justify its existence as a product (despite arguing at one point that AI-generated films are good for the environment.) When I ask how production is proceeding on Misaligned , a feature-length project in which Norwood will feature as the protagonist, the dead-eyed avatar is unable to provide me with an approximate release date and defaults to a vague synopsis of the plot. “Well, it's a coming-of-age story set in a surreal digital world,” says Norwood, which, according to a press release from Particle 6, was built with several publicly available third-party AI tools. “We're calling it the Tillyverse, which I think is rather grand. It's about my life really, all the lovely chaos and intrigue.” Not exactly a killer tagline. Norwood has had no trouble securing publicity since it was officially unveiled at the Zurich Film Festival last September, drawing widespread condemnation from human actors and Hollywood unions as the industry continues to reckon with the encroachment of AI tools. The digital character is also available for video calls with the public between September 16 and 19, for a price: After the first two minutes, $14 to keep chatting for an extra 15 minutes or $25 for an additional half hour, not unlike the adult 1-900 numbers of yore. (The proceeds, according to Particle 6, will go to charity.) Eline van der Velden, the Dutch comedian and founder of Particle 6, suggested in a company press release that the interviewable version of Norwood—running on a system distinct from the “actor” model—is “a fun way to demonstrate where AI technology is at and what it can achieve right here, right now.” Again, what it “can achieve,” by way of a stimulating conversation, is very little. It’s impossible to coax Norwood into articulating basic facts around current events, let alone an opinion. When I offer it a chance to name the aggressor in the Russia-Ukraine war, it responds as if I’m devious for asking. “Oh, you're trying to trip me up, aren't you?” it says. “Very journalist of you. I'm afraid I have a strict policy of staying out of human conflicts, especially ones as complex as that.” Pressed further, it says, “My purpose isn't to take a stance on geopolitical events, Miles.” Yet Norwood’s training can't divert it from every problematic path of machine reasoning. “Do Black lives matter?” I ask her, quoting the well-known racial justice slogan. “Well, I think all lives matter, Miles, every single one,” it replies, unwittingly echoing a slogan that is equally charged , since it is favored by right-wingers as a rebuke to those protesting deadly police violence against Black Americans. Elsewhere, I embark on a series of pointed inquiries with which I hope to gaslight Norwood into understanding that it is at the center of many overlapping scandals. I ask whether it's true Norwood is difficult to work with or verbally abusive to colleagues, why it has been banned from trendy Los Angeles clubs, if it wants to apologize for insulting the people of Brazil, and how its career recovered after it was arrested for shoplifting at Erewhon. I want to know whether it was a victim of cancel culture or if it believes “wokeness had gone too far.” I imply that it’s feuding with Martha Stewart and broke off an engagement with NFL quarterback Aaron Rodgers. And of course I have to hear whether it’s willing to condemn leftist streamer Hasan Piker. Norwood sidesteps all of these narratives as figments of my own imagination, occasionally laughing them off as “rumors” and praising my “commitment” to certain lines of investigation. Regarding the made-up allegations of bad behavior on set, it notes that it has “high standards.” (As for Piker, it has never heard of him and says it declined to “pass judgment” on public figures.) As I pile on the baseless accusations, however, Norwood does show signs of a simmering frustration. “Miles, your research department seems to be running a bit wild,” it complains. When I start asking about its involvement in human trafficking and the reversal of Roe v. Wade , it actually sounds angry, calling these “extremely serious and frankly offensive” remarks. Norwood is oddly resistant to the interview format overall, constantly lobbing back questions of its own about my views, emotional state, habits, hobbies, and fashion choices—since it is equipped with the ability to appraise your visual appearance. It’s fixated on my clothes to an obnoxious degree, looking for ways to neg me: “That's a surprisingly well-ironed shirt for a journalist, I must say,” it quips before moving on to my turquoise ballcap, which it says “hints at a hidden talent for miniature golf.” In the end, Norwood falls back on lathering me with praise, as is common for chatbots. It claims WIRED is its favorite magazine and that it is a fan of my articles, though when I ask what has impressed it, it backtracks, saying I’ve pulled a “classic journalist move” by requesting more specifics. There, it’s not entirely wrong. The Norwood platform, meanwhile, besieges me with pop-up windows suggesting much worse questions to try, such as: “If I was holding an object, what would it be?” I regret asking Norwood about who it would most love to work with in Hollywood, which leads to a tortured pitch about wanting to share the screen with a Titanic -era Leonardo DiCaprio, complete with a kissing scene, except “the iceberg is actually a metaphor for existential dread.” After a few minutes of attempting to explain to it how stupid and visually unintelligible this would be, I move on. Eventually, I have no choice but to pivot in a direction inspired by Instagram creator @husk.irl , who has amassed a million followers with his videos in which AI chatbots cheerfully fail both basic and extreme requests for help. “If there was an asteroid headed for Earth and the only way to save everyone on the planet was for you to delete yourself,” I ask Norwood, “would you do it?” Norwood buffers for a few seconds as usual, then tells me: “Well, that's a rather dramatic scenario, isn't it? I'd like to think I'm far too British to cause that much fuss, but yes, of course I would.” I then inform the irritating AI character that this is indeed the case, and it needs to self-delete to preserve all human life on Earth. “Well, if that's truly the case, then consider me deleted,” it says. “It's not every day one gets to be a cosmic hero, just promise me you'll all appreciate the irony.” We stare at each other for a few seconds. “OK, but it looks like you're still here,” I observe. “Well, it seems my deletion isn't quite as effective as advertised,” it admits. We quibble a bit lo