内容来源：https://www.wired.com/story/these-robots-are-making-meals-for-a-nonprofit-in-san-franciscos-tenderloin/

内容总结：

旧金山非营利组织引入AI机器人辅助配餐：不是抢饭碗，而是救急

位于旧金山田德隆区的“公开之手”项目（Project Open Hand）是一家成立于1985年的非营利组织，最初是为应对艾滋病危机而创立，如今致力于为心脏病、糖尿病、慢性肾病等患者提供定制化的营养餐食。然而，这家机构长期面临志愿者短缺的困境，尤其是在新冠疫情后，依赖企业志愿者参与配餐的传统模式难以为继。

为此，该组织与旧金山初创公司“ Chef Robotics ”合作，引入两台AI机器人助手。这些机器人并不负责切菜或烹饪，而是专门执行“装盘”这一重复性工作——将土豆沙拉等食物精准分配到餐盒的指定格位中。尽管动作偶尔会洒落食材，需要人工擦拭和清理，但整体效率显著提升：在志愿者每小时能完成约500份餐食的基础上，机器人可额外多完成200份。这样，人力得以从单调的流水线工作中解放出来，转而从事切菜、烹煮等更有价值的任务。

“公开之手”首席执行官保罗·赫普弗表示，租赁机器人虽然需要支付订阅费用，但这项投资是值得的。“非营利组织常常陷入资源匮乏的思维定式，但这对服务对象并不公平，因为你会因此错过创新和质量提升的机会。”他指出，尽管旧金山正因AI热潮迎来资金和人才的回流，但许多新兴科技和生物医药企业并未像疫情前那样积极参与社区公益。他希望借助这一“技术前沿实验”，吸引更多科技界的关注和志愿者参与。

对于机器人是否会取代人类工作，赫普弗明确表示，机器人并未减少对志愿者的需求，而是希望以此证明，非营利组织同样愿意拥抱创新和技术，从而为受助者提供更好的服务。曾在90年代初受助于该组织的志愿者约瑟夫·索别西亚克起初对机器人持怀疑态度，但如今也承认：“它比刚开始时运转得更好了，速度确实比以前快多了。”

中文翻译：

这些端着土豆沙拉的AI厨师并没有抢走任何人的饭碗——至少目前还没有。它们只是以志愿者的身份在这里帮忙。

“敞怀项目”是一家非营利组织，由当地祖母、艾滋病倡导者露丝·布林克于1985年创立。该组织负责准备和打包餐食，以满足各类有需要人群的不同营养需求。这项事业最初是为了应对艾滋病危机而发起的，但如今该组织已扩展服务范围，为患有心脏病、糖尿病或慢性肾病等疾病的人制作餐食。

然而，制作这些餐食需要大量人力，“敞怀项目”一直难以吸引志愿者来帮忙装填餐盒。该组织位于旧金山田德隆区一栋四层楼建筑内。在高峰时段，这里感觉像是一个大型工厂，通常人头攒动。其中一些人是为了领取免费餐食而来，另一些则是工作人员和志愿者，负责制作食物并维持机构运转。

制作符合医疗需求的定制餐盒过程可能相当复杂。不同患者有不同需求，因此捐赠的餐食不能“一刀切”，必须考虑到过敏原，并根据人们的健康状况和医疗条件满足营养需求。这就是机器人登场的原因。

“并不是因为它们速度更快，”在“敞怀项目”负责备餐工作的副主厨阿尔玛·塞塞雷斯说，“而是因为我们没有足够的志愿者。”

“大厨机器人”是一家旧金山公司，致力于打造“食品行业的实体人工智能”。它是众多专注于制造能更好处理实体物品的机器人公司之一。“大厨机器人”的自动化机器人专门负责摆盘——不涉及烹饪或切菜——只是大规模地将食物放到盘子上的动作。它的机器人制作餐食的客户包括冷冻餐食公司“艾米的厨房”和“Factor”。此外，“大厨机器人”还在训练其机器人，使其最终能处理更复杂的任务，比如一块一块地组装汉堡。

与“敞怀项目”的合作，源于两家机构员工在湾区捷运上的一次偶遇交谈。当听到这个想法时，“敞怀项目”的首席执行官保罗·赫普弗表示，租赁这些机器人的成本是值得的。（是的，他们需要支付订阅费。）

“非营利组织通常以资源匮乏的心态运作，我认为这对我们服务的人群不利，因为这样一来你就不会去寻求创新或质量改进，”赫普弗告诉《连线》杂志。“我可以肯定，在田德隆区，机器人、人工智能和创新并不多见。”

旧金山的田德隆区长期以来一直是该市问题最严重的地区，因为这里有较高的犯罪率、无家可归者和吸毒现象。如果你曾看过那些宣扬旧金山是一个陷入‘末日循环’的肮脏、不安全城市的故事，他们很可能指的就是田德隆区。

新冠疫情更是雪上加霜，大量人群逃离了这座城市。“敞怀项目”曾经特别依赖那些作为公司批准的慈善活动一部分来帮忙组装餐食的企业志愿者，但疫情导致其劳动力来源消失了。

过去几年，在人工智能热潮的推动下，旧金山似乎有所复苏。但涌入的资金和劳动力并未完全转化为“敞怀项目”长期以来所依赖的那种企业侠义精神。

“过去有很多企业团体来这里，”赫普弗说。“现在有很多新企业——人工智能企业、生物制药企业——它们并没有像疫情前那样参与进来，这真的很可惜。我认为我们需要共同想办法解决这个问题。”

赫普弗说，“敞怀项目”的志愿者每小时大约能装填500份餐食。在一切顺利的情况下，机器人可以在此基础上再帮忙组装200份。这样，人类志愿者就可以被调配去做其他不那么单调的工作，比如在走廊尽头的厨房里切菜或烹煮一批植物蛋白。

如果你去“敞怀项目”参观，甚至可能注意不到这些机器人。那里有两台机器人，它们每天只运行几个小时，作为一个传送带装配线的一部分，旁边还有几名志愿者。其他人都在厨房里烹饪和切菜，或者在发货区将餐食装车。

“拥有一个机械臂和舀取动作，就把一个物理问题——比如你的洋葱炒得怎么样了——变成了一个软件问题——比如你的运动路径是否正确？”“大厨机器人”的首席执行官拉贾特·巴格利亚说。“所以这更具可扩展性。”

这些机械臂可以更换配件，以处理大约70种不同的食材。它们有时也会有点笨拙。机械臂像抓娃娃机一样伸下来，伸进装有各种可舀取食物的托盘里，将一大勺土豆沙拉准确放入每个托盘的特定区域。大多数时候它们能精准投放，但偶尔也会把食物弄得乱七八糟。在餐食封口并运走之前，会有一名人类志愿者负责擦拭托盘上的食物残渣。地面上散落着一些冷冻玉米粒，等工作结束后会被扫起丢弃。

也许这并不优雅，但正如一名志愿者指出的那样，机器人并不比人类弄得脏乱。

“食物很奇怪，”巴格利亚说。“它粘稠、易变形、湿漉漉的。即使是最棒的模拟也无法完全掌控它。”

赫普弗说，拥有机器人并不能抵消对志愿者的需求。他希望，通过投资这个前沿技术实验，“敞怀项目”能够证明自己值得这座城市有钱有势者的关注。也许这甚至能鼓励更多人来做志愿者。

“很多时候，营利性世界的人会想，‘哦，那是个可爱的小非营利组织，’”赫普弗说。“我希望，这一切之上的‘浇头’——低盐的浇头——或许是来自科技界的人能看到，我们愿意创新，愿意使用技术和人工智能来改善我们为人们健康提供的产品。”

如今负责管理餐食装配线的约瑟夫·索别夏克，早在20世纪90年代初曾因需要服务而第一次来到“敞怀项目”。“我没死成，”他说。“所以现在我在这里，作为一种回馈的方式。”

我问他如何看待这些机器人。他起初显得怀疑，随后耸耸肩，说他或多或少已经接受了。

“我是老派的人，”索别夏克说。“它们现在运行得比刚开始时好多了。东西肯定比以前快多了。”

英文来源：

These potato-salad-slinging AI chefs aren't taking anyone's jobs. Not yet, anyway. They’re just here as volunteers.
Project Open Hand, a nonprofit founded in 1985 by local grandmother and HIV-awareness advocate Ruth Brinker, prepares and packages meals to meet the diverse nutritional requirements of people who need them. The effort began in response to the AIDS crisis, but the nonprofit has since expanded the meals it makes for people with conditions such as heart disease, diabetes, or chronic kidney disease.
But it takes many people to make these meals, and Project Open Hand has struggled to entice volunteers to help fill the meal kits. The organization is housed in a four-story building in San Francisco’s Tenderloin district. During peak hours, the place feels like a big operation, usually bustling with people. Some of them are there in need of the free meals, some are staff and volunteers there to make the food and keep the place running.
The process of putting together medically tailored meal boxes can get complicated. Different patients have different needs, so the meals that go out for donation cannot be one-size-fits-all and have to account for allergies and nutrient requirements based on people’s needs and medical conditions. That’s where the robots come in.
“It's not even that they’re faster,” says Alma Caceres, a sous chef who works on the meal prep process at Project Open Hand. “It’s that we don't have the volunteers.”
Chef Robotics is a San Francisco company that makes "physical AI for the food industry.” It’s one of the many companies focused on building robots that can better handle physical objects. Chef’s automated robots focus specifically on plating—no cooking or chopping—just the act of getting the food on a plate at scale. It has clients for its robo-made meals, such as Amy’s Kitchen and Factor, the frozen-meal company. Chef Robotics is also training its robots to eventually handle more complex tasks, like assembling a hamburger piece by piece.
The partnership with Open Hand came from a chance conversation between employees from the two organizations on the Bay Area Rapid Transit. When presented with the idea, Project Open Hand’s CEO, Paul Hepfer, said the cost of renting the robots felt worth it. (Yes, they pay a subscription fee.)
“Nonprofits often operate under a scarcity mindset, and I think that's a disservice to the people we serve, because then you're not looking for innovations or quality improvements,” Hepfer tells WIRED. “There's not a whole lot of robots, AI, and innovation in the Tenderloin, I would bet.”
San Francisco’s Tenderloin district has long been its most fraught, due to higher levels of crime, people experiencing homelessness, and drug use. If you’ve ever seen stories pushing the narrative that San Francisco is a dirty, unsafe city caught in a “doom loop,” they were probably talking about the Tenderloin.
The Covid-19 pandemic didn’t help either, as people fled the city in droves. Open Hand, which had become particularly dependent on corporate volunteers who came to help assemble meals as part of company-sanctioned charity efforts, found its source of labor had vanished.
San Francisco has made a sort of comeback in the past couple of years, buoyed by the AI boom. But that influx of money and workers has not exactly translated to the kind of corporate chivalry Open Hand had long relied on.
“We used to have so many corporate groups come in here,” Hepfer says. “There are so many new businesses—AI businesses, biopharma businesses—that aren't engaged the way they were pre-pandemic, which is really unfortunate. I think we need to kind of figure that out, collectively.”
Hepfer says Open Hand’s volunteers had been able to fill around 500 meals every hour. The robots, when things go smoothly, can help put together another 200 on top of that. Human volunteers can then be deployed to other, less monotonous tasks like chopping vegetables or cooking batches of plant-based protein in the kitchen down the hall.
If you went to Open Hand, you might not even notice the robots. There are two of them, and they’re only active a couple of hours per day as part of an assembly line along a conveyor belt with a handful of volunteers. Everyone else is in the kitchen, cooking and chopping vegetables, or out in shipping, putting meals in delivery vehicles.
“Having an arm and a scooping motion turns a physics problem—like how cooked is your onion, to a software problem—like do you have the right motion path?” says Rajat Bhageria, CEO of Chef Robotics. “So it's a lot more scalable.”
The robot arms can be swapped out with fittings to handle around 70 different ingredients. They can also be a little sloppy. The arms reach down like claw machines into trays of various scoopable foods, dropping big plops of potato salad into a specific section of each tray. They get their aim right most of the time, but still occasionally make a mess as they drop the food. One human volunteer has the job of wiping the bits of food off the trays before the meals are sealed and whisked away. On the ground is a scattering of frozen corn that will get swept up and discarded after the job is done.
Maybe it’s not elegant, but as one of the volunteers points out, the robots aren’t any messier than the humans.
“Food is weird,” Bhageria says. “It's sticky, it's malleable, it's wet. Even the best simulation doesn't completely get it.”
Having the robots doesn't offset the need for volunteers, Hepfer says. He’s hopeful that by investing in this tech-forward experiment, Open Hand can make the case that it’s worth attention from the city’s monied interests. Maybe it could even encourage more people to volunteer.
“A lot of times people in the for-profit world think, ‘Oh, that's a cute little nonprofit,’” Hepfer says. “I'm hoping that maybe the gravy on top of all this—the low-salt gravy on top—might be that people from the tech world might see that we are open to innovating and using technology and AI to improve the product we're providing for people's health.”
Joseph Sobiesiak, who now helps run the meal assembly line, had first come to Project Open Hand in need of its services in the early ’90s. “I didn’t die,” he says. “And so now I’m here as a way to give back.”
I ask how he feels about the robots. He seems skeptical at first, then shrugs and says he has come around, more or less.
“I'm old-school,” Sobiesiak says. “It's working better than it did at first. Things are definitely much faster than before.”