内容来源：https://aibusiness.com/robotics/designing-ai-enabled-robots

内容总结：

波士顿动力公司近日宣布，将与阿联酋软件公司Analog合作，在阿联酋公共区域部署其四足机器人Spot。这一突破性合作标志着Spot机器人首次大规模应用于非工业场景，展现了机器人技术在公共服务领域的广阔前景。

据波士顿动力首席战略官马克·瑟曼介绍，此次合作将推动Spot实现两大技术飞跃：一是从预置地图的结构化工厂环境，转向需要自主感知的开放空间；二是通过搭载名为"安娜"的具身AI系统，使Spot首次具备与市民直接对话的能力。

在阿联酋的试点项目中，Spot将承担公园巡检、环境监测和无障碍设施评估等任务。未来这些机器人还可能现身购物中心担任导览员，甚至扩展至城市级应用规模。

瑟曼指出，具身AI正重塑人机互动模式。医疗研究表明，当AI被赋予实体形态后，更能激发人们的情感共鸣。这正是波士顿动力从工业工具制造商向服务机器人提供商转型的关键所在。

面对全球劳动力短缺和固定自动化设备僵化等行业痛点，波士顿动力通过三大技术突破提升机器人适应性：语义导航技术让机器人能识别人类并主动避让；强化学习将新技能训练时间从三个月压缩至48小时；模块化设计使机器人无需改造场地即可投入应用。

尽管欧美地区仍面临法规滞后和文化接受度等挑战，但阿联酋政府对科技创新的支持为项目落地提供了有利条件。瑟曼预测，本世纪末服务型机器人将步入家庭，二十年内通用人形机器人将解放人类创造力，开启人机协作的新纪元。

（本文根据波士顿动力公司战略官专访整理）

中文翻译：

由谷歌云赞助
选择首个生成式AI应用场景
开展生成式AI应用时，应优先关注能优化人类信息交互体验的领域。

硬件与软件技术的革新正让波士顿动力公司的机器人变得愈发智能和灵活。本月初，该公司与阿联酋软件企业Analog达成合作，将在中东地区公共场景规模化部署机器狗Spot。这是Spot首次在工业场景外的大规模应用，要求机器人具备更强的空间认知与行动智能，以应对非结构化环境中的安全移动问题。初期将聚焦公园巡检、环境监测及无障碍设施评估，远期计划实现全城范围部署，与居民直接互动。

在本次访谈中，波士顿动力首席战略官马克·蒂尔曼详解了该项目如何推动机器人智能化战略升级。

关于与Analog的合作如何拓展Spot应用边界？
目前Spot主要应用于两类场景：危险环境远程勘察与工业设施自主巡检。Analog合作的突破性在于部署规模——他们致力于实现城市级甚至全国范围覆盖。这些机器人将监测空气质量、检测结构异常，甚至在商场充当互动导览。此次合作标志着我们首次从工厂走向真实世界，为开发服务型机器人乃至家用机器人奠定基础。

项目对Spot的环境交互能力提出哪些新要求？
硬件传感器基本保持不变，但软件系统有两大升级：首先，Spot需从结构化工厂环境转向非结构化户外场景，这就要求具备自主感知与未知区域探索能力；其次，Analog开发的AI虚拟角色Ana将植入Spot系统，实现人机直接对话。

具身AI为何对机器人行业至关重要？
它彻底改变了人机交互模式。儿童医院和养老院的研究表明，当AI被赋予实体形态时，人们会产生更深层的情感联结。即使对工业型企业而言，要实现人机协作，具身AI也日趋重要。未来Spot可能成为移动承载平台，为各类具身AI系统提供可抵达人类活动任何场景的可靠移动能力。

波士顿动力如何应对AI领域软硬件需求变化？
语义导航技术让机器人开始理解环境——Spot现在能识别人类并主动避让；强化学习则将新技能训练时间从三个月压缩至48小时，且技能提升会产生跨任务的正向迁移。

劳动力市场对机器人集成需求激增的驱动因素？
一是高技术岗位人才断层使企业渴望将重复性、高危性工作移交机器人；二是固定自动化系统成本高昂且缺乏灵活性，而移动机器人无需改造基础设施即可部署，并能随业务需求动态调整。这正是旧改设施对机器人方案青睐有加的原因。

技术整合面临哪些挑战？
欧洲受限于陈旧的法规体系，美国需克服文化接受度障碍。宝马公司通过让Spot在食堂巡展提升员工接纳度，英特尔则用征名活动拉近人机距离。阿联酋政府对技术实验的开放态度为机器人推广提供了关键支持。

对机器人技术未来发展的展望？
二十年内通用人形机器人将承担人类不宜从事的工作，释放人类创造力。它们终将以伙伴、助手甚至朋友的身份融入社会生活——虽然不会如媒体渲染般迅速，但本世纪末前必将实现，这值得期待。

编者按：为提升清晰度与简洁性，本访谈内容经过编辑。
猜您喜欢

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
Advances in hardware and software are making Boston Dynamics robots smarter and more adaptable.
Earlier this month, Boston Dynamics unveiled a collaboration with UAE software company Analog to deploy Spot, the Massachusetts company's robot dog, across public spaces in the region.
The project is one of the first wide-scale deployments of Spot outside industrial settings, requiring Boston Dynamics to advance the robot’s spatial understanding and physical intelligence to safely navigate unstructured environments. Initial work will focus on park inspections, environmental monitoring and accessibility assessments. In the longer term, Spot could operate at a city-wide scale, interacting directly with residents.
In this Q&A, Marc Theermann, chief strategy officer at Boston Dynamics, discusses the project and how it fits in with the vendor's broader strategy to make its robots smarter and more flexible than ever before.
Can you tell me about the collaboration with Analog? How does it push Spot into new applications?
Marc Theermann: Today Spot has two main use cases. The first is teleoperation in dangerous environments. In those scenarios, Spot becomes the eyes and ears on the ground where humans shouldn’t be.
The second is autonomous industrial inspection, when Spot autonomously traverses a factory floor and collects data using an array of sensors, enabling predictive and preventive maintenance.
What’s new with Analog is the scale. They have ambitions to do that at a city-wide, if not nationwide, scale. These robots could monitor air quality, detect structural anomalies, and even appear in shopping malls where you can interact with them as a kind of mobile guide.
We're thrilled about the partnership because it's really the first time that we're walking out of the factory into the real world. Hopefully, that sets us on a path toward developing service robots and, eventually, robots that people can have in their homes.
Has the project changed the requirements of Spot, and how it interacts with its surroundings?
Theermann: The hardware and sensors are largely the same, but two big changes are happening on the software side.
First, Spot is moving from structured to unstructured environments. Inside a factory, we pre-map the entire space so the robot knows every corridor and obstacle. Outside, that all disappears. Spot has to handle uncharted exploration, which requires a new layer of autonomy and perception.
Second, people will be able to interact with Spot directly for the first time. That brings us into the domain of Embodied AI, and that’s what Analog is bringing to the table. They have developed an AI character called Ana, which will be embedded into Spot, and which will be able to have conversations with users.
Why is Embodied AI important for robotics? How is it changing the industry?
Theermann: Embodied AI fundamentally changes the way people relate to machines. There’s a lot of research, for example, in children’s hospitals and elder-care facilities, which shows people engage differently with AI when it’s embodied in a physical body. Patients respond more emotionally, build trust faster, and can even recover more quickly.
While Boston Dynamics has been focused on building industrial tools, for these to become co-workers in the future, a level of human interaction will still be necessary. So even for an industrial company, Embodied AI is going to become more and more important.
In the future, Spot could also become a carrier for other Embodied AI systems, acting as a mobile, highly reliable platform that can go anywhere a person can.
How is Boston Dynamics responding to changing demands for software and hardware in AI?
Theermann: Several major advances in AI over the past two years have had a transformative effect on robotics.
A big one is semantic navigation, which means robots are beginning to understand their environment. Previously, when Spot encountered a human in a factory, it would be treated like any other stationary object. But now, Spot recognizes you as a human and behaves differently. It will give you a wider berth or wait for you to pass.
Another is reinforcement learning. Two years ago, teaching a robot a new skill could take three months. Now it takes 48 hours. We’ve also discovered that if a robot becomes better at one skill, it actually increases its capability to do other skills and helps improve its generality.
What do you think is driving the demand for more robotic integration in the workforce?
Theermann: Two things are happening here. First, there’s a huge labor shortage. We’re seeing a lot of people aging out of highly technical jobs, and they’re taking the skills with them. In response, companies want to transfer some of this knowledge into robotic systems and let them take over jobs, particularly those that are dull, dirty and dangerous.
Second, fixed automation is expensive and rigid. You can automate almost anything with enough infrastructure, but once it’s installed, you’re locked in. Robots like Spot or Atlas offer mobile automation without the need for new construction. And if your business changes, the robots adapt with you.
That’s why there’s so much interest and hype. Robots make automation possible in brownfield facilities without altering the physical environment.
What are the challenges of integrating technology like this?
Theermann: Regulatory hurdles are definitely a problem, particularly in Europe, because a lot of the regulations in place are outdated. They were written ten to 15 years ago, and they didn't account for autonomous machines operating in a factory.
In America, by contrast, there are a lot of cultural hurdles, but we’ve seen some creative ways of improving acceptance.
One client, BMW, paraded Spot through the cafeteria so employees could meet it and ask questions. Intel ran a naming contest so the workforce could choose Spot’s name. Once people see the robot as something fun and useful rather than intimidating, deployment becomes much easier.
By contrast, the UAE is very welcoming to robotics and to experiments, and ultimately, you do need this government interest if you want to see technological rollout.
Looking ahead, how do you see the robotics landscape changing? Is there anything you’re particularly excited about?
Theermann: In 20 years, I believe we’ll have general-purpose humanoids doing jobs humans can’t, or shouldn’t, do. If labor becomes abundant because humanoid robots can handle it, think about the creativity that could be freed up for people.
I also believe these robots will live among us as companions, helpers, even friends. Not as fast as some people are claiming in the media right now, but I think it will certainly be by the end of this decade. And that's an exciting future.
Editor's note: This interview has been edited for clarity and conciseness
You May Also Like