00后女生一口气租30套房,每套只住一周,中介怀疑她是骗子,真相让人意外:带50多人教机器人做家务
据极目新闻10月30日报道,为了教机器人做家务,22岁的杨心怡干了一件让房产中介摸不着头脑的事:一口气要租30套房子,每套只住一周左右。
“你租这么多房子干什么?”面对中介的追问,她解释,房子不是给人住的,而是给机器人“练手”的。这个理由让不少中介一时难以理解。有人直接拒绝,有人怀疑她是骗子,还有人要求她提供工作证明、工牌和身份证。
Yang Xinyi, a 22-year-old project manager in Wuhan, Hubei province, has rented 30 apartments, each for about a week, not to live in, but to train humanoid robots to do household chores.
这位2004年出生的浙江姑娘是武汉一家科技有限公司的一名项目负责人。她带领50多人的团队,为人形机器人采集训练数据。她和同事们每天都在做一件看似简单、实则复杂的工作:把人类习以为常的生活技能,一点点教给机器人。
Originally from Zhejiang province, she joined the Hubei Humanoid Robot Innovation Center in October 2025 as a data collector. She now leads a team of more than 50 people dedicated to gathering training data for humanoid robots.
还得判断它做得对不对
2025年10月,杨心怡从浙江来到湖北武汉,加入湖北人形机器人创新中心运营团队,担任数据采集员。戴上PICO头显,手握两个控制手柄,面前的机器人便可以在引导下完成指定动作。杨心怡要做的就是反复操作,将动作过程转化为训练数据,供机器人学习。
最初,训练内容只是拿起、放下物品。随着任务推进,训练场景也从简单的操作台延伸到超市、家庭、药店、厨房、餐厅等复杂环境。
Her daily work involves wearing a VR headset and holding two controllers to guide robots through specified movements, with each motion converted into training data.
What began as simple pick-and-place tasks has expanded to complex environments including supermarkets, homes, pharmacies, kitchens and restaurants.
杨心怡说:“普通人做什么,我们就教机器人做什么。”
真正上手后,她才发现,人类一个不经意的动作,对机器人来说可能是一道难题。比如,人可以左手拿着手机,右手端起水杯,两个动作同时完成。但机器人还不能像人一样灵活处理这样的双手协同操作。因此,训练师必须将连续动作进行拆解,逐一采集数据。
The challenge, she discovered, is that actions humans take for granted can be extraordinarily difficult for robots. A person can hold a phone in one hand while picking up a cup with the other, but robots cannot yet handle such bimanual coordination smoothly. Trainers must break continuous movements into smaller steps and collect data for each one.
不仅要教机器人做动作,还要判断它做得对不对。成为审核员后,杨心怡开始负责检查训练数据是否合格,以及机器人完成动作是否符合要求。
随着项目推进,她先后担任组长、项目负责人,如今带领着50多人的团队,专职给机器人当老师。
Her responsibilities grew from data collector to quality reviewer, team leader, and now head of a dedicated group. She checks whether training data meets standards and whether robots complete tasks correctly.
房产中介以为遇到了骗子
随着训练任务增多,杨心怡和团队遇到了新的问题:机器人需要学习的技能越来越多,训练环境必须更加丰富。
今年6月,团队接到一项数据采集订单,需要进入不同的真实家庭环境,收集机器人“做家务”所需的数据。
这一次,难住杨心怡的不是机器人,而是房子。由于任务需要不同户型、不同空间尺度和不同家具布局,团队计划寻找约30套有实际生活痕迹的住房,每套使用一周左右。
Securing suitable homes proved difficult after her team received an order in June to collect data in real home environments. The plan called for about 30 lived-in apartments with different layouts, sizes, and furniture arrangements, each to be used for roughly a week.
白天,她要在公司处理项目工作;一有空,她就联系房产中介寻找房源。但听到租期只有一个星期,不少中介便没了兴趣。后来,杨心怡干脆把联系过的中介拉进一个群,统一说明需求:要租30套房子,给机器人用,租期短,还得有人生活过的痕迹。
没想到,解释得越详细,怀疑她的人反而越多。就在房源迟迟没有着落时,一位40多岁的阿姨伸出了援手。这位中介大姐对机器人很感兴趣,不仅帮她联系了10多套房子,还到现场观看了团队开展数据采集的过程。
Many agents lost interest when told the arrangement. Some refused outright, others suspected fraud, and some demanded work certificates and ID cards. She created a group chat to explain the requirements to all of them, but detailed explanation only deepened their suspicions.
Eventually, a woman in her 40s who was interested in robots helped connect the team with more than a dozen properties and even visited the data collection site.
“她觉得很有意思,很愿意帮我们。”说起这段经历,杨心怡至今仍心存感激。
她也在等待答案
工作的一年里,杨心怡听到过不少疑问。有人直接追问:“机器人什么时候才能走进普通家庭,替人做饭、整理房间?”
在她看来,当前的机器人已经能够在部分特定场景中完成任务,但距离适应千家万户复杂多变的生活环境,还有很长一段路要走。她相信,随着数据积累和技术迭代,机器人能够完成的任务会逐步增多。
Asked when robots will enter ordinary homes to cook and tidy up, she says current machines can handle tasks in specific scenarios but remain far from adapting to the complex and varied environments of real households.
She believes data accumulation and technological iteration will gradually expand their capabilities.
如今,父母对她的工作已有了更多了解,也为女儿能够参与这个新兴行业感到骄傲。
Her parents now better understand her work and take pride in her involvement in an emerging industry.
“我觉得自己进入这个行业比较早,能够参与、见证它的发展,是一件很有意义的事!”杨心怡说,她愿意继续做下去,因为每天都能看到机器人的变化。
跟着China Daily
精读英语新闻
“无痛”学英语,每天20分钟就够!
↓↓↓
推 荐 阅 读