
迪士尼打造了三个《星球大战》电子动画角色,它们能对游客做出反应,也能彼此互动,全程无人操控。亚马逊云科技(AWS)提供人工智能支持。华特迪士尼幻想工程在周三的一篇帖子中将这三个安泽兰人命名为维洛·维尔、卡多·布雷亚和吉布·布雷亚。
Disney has built three Star Wars animatronics that react to guests and to each other with no one controlling them. Amazon Web Services supplies the AI. Walt Disney Imagineering named the three Anzellans Vilo Vele, Kado Bre’a and Jibu Bre’a in a post on Wednesday.
自1963年迪士尼乐园的“魔法提基屋”开放以来,迪士尼的电子动画角色一直按照固定的脚本循环运行。迪士尼表示,安泽兰人则不同,它们会留意周围发生的事情,自行决定说什么、怎么动。
Disney’s Audio-Animatronics have run on fixed, scripted loops since the Enchanted Tiki Room opened at Disneyland in 1963. The Anzellans instead notice what is happening around them, Disney said. They decide for themselves what to say and how to move.
“当我们在re:Invent大会上推出安泽兰人时,这将是真正意义上第一次根据游客的身份来个性化这一体验,”领导幻想工程研发的凯尔·劳克林对《好莱坞报道》表示。
“When we launch with the Anzellans at re:Invent, it’ll be really the first time in which we will personalize that experience based on who that guest is,” Kyle Laughlin, who leads Imagineering R&D, told The Hollywood Reporter.
游客可以首先在拉斯维加斯举行的亚马逊re:Invent大会上见到它们,大会时间为11月30日至12月4日。迪士尼尚未说明它们何时会进入旗下乐园。
Guests can meet them first at Amazon’s re:Invent conference in Las Vegas, from 30 November to 4 December. Disney has not said when they will reach its parks.
AI选台词,幻想工程师写台词 每个角色都配有摄像头和麦克风,能判断面前站着多少人以及大致年龄。劳克林对CNET表示,它们还能识别生日徽章或《星球大战》T恤。
AI picks the line, Imagineers write it Each character has cameras and microphones. They can tell how many people stand in front of them and their rough ages. They also spot a birthday button or a Star Wars T-shirt, Laughlin told CNET.
AI并不撰写它们的对白。据CNET报道,幻想工程师录制了数十万条台词,混合使用英语和安泽兰语,由系统决定播放哪一条。这些角色有固定的人设:维洛是老板,卡多是渴望关注的机械师,吉布是卡多的双胞胎,性格害羞。
The AI does not write their dialogue. Imagineers recorded hundreds of thousands of lines in a mix of English and Anzellan, according to CNET. The system chooses which to play. The characters have set personalities: Vilo is the boss, Kado the attention-seeking mechanic, and Jibu, Kado’s twin, the shy one.
系统运行在亚马逊Bedrock上,幻想工程师可以借此在不同公司的AI模型之间切换。AWS员工使用亚马逊的AI编程助手Kiro,把幻想工程的创意简报转化为软件,再由人工审核。劳克林说,这个项目只用了“短短几个月”。
The system runs on Amazon Bedrock, which lets Imagineers switch between AI models from different companies. AWS staff used Kiro, Amazon’s AI coding agent, to turn Imagineering’s creative briefs into software, which humans then reviewed. The project took “mere months”, Laughlin said.
“这些安泽兰人现在是我们全球最先进的电子动画角色,”劳克林对TechRadar表示。
“These Anzellans now are our most advanced Audio-Animatronics in the world,” Laughlin told TechRadar.
迪士尼如何使用摄像头数据?劳克林对TechRadar表示,系统只将一次互动中的数据用于当下,不会把数据保存在云端重复使用,也不会用这些数据训练AI模型。如果系统出现故障,角色会切换到另一种互动体验。
What happens to the camera data Disney uses the data from an encounter only for that moment, Laughlin told TechRadar. It does not keep the data in the cloud to reuse, and it does not train AI models on it. If the system fails, the characters fall back to another experience.
据TechRadar报道,安泽兰人运行在与迪士尼自由行走的雪宝及其BDX机器人相同的机器人平台上。劳克林说,游客会先在室内、在商店里与它们见面。
The Anzellans run on the same robotic platform as Disney’s free-roaming Olaf and its BDX droids, according to TechRadar. Laughlin said guests will meet them indoors first, in shops.
迪士尼削减成本据CNBC报道,此次发布前一天,迪士尼表示目标是到2029年每年节省30亿美元。首席财务官休·约翰斯顿对CNBC说:“我们确实预计会有更多裁员。”迪士尼此前已计划裁员数千人。
Disney’s cost cuts The launch comes a day after Disney said it is targeting $3bn in yearly savings by 2029, CNBC reported. Finance chief Hugh Johnston told CNBC: “We do expect more layoffs.”Disney had already planned thousands of layoffs.
与此同时,迪士尼正在技术方面投入。它从Character.AI聘请了首位首席技术官,并同意支付Character.AI员工的费用。本月早些时候,它与Adobe签署了AI联盟协议。
At the same time, Disney is spending on technology. It hired its first chief technology officer from Character.AI and agreed to pay for Character.AI staff. Earlier this month it signed an AI alliance with Adobe.