摩托罗拉解决方案公司正探索人工智能和其他新技术的能力，但这家公共安全通信巨头正努力以改进任务流程、让人类做出最终决定的方式来使用它们
. 斯坦伯格说，摩托罗拉解决方案公司有三个指导原则，它确定了其产品中利用人工智能的最佳方式：
摩托罗拉解决方案公司（motorolasolutions）正探索人工智能和其他新技术的能力，但这家公共安全通信巨头正努力以改进任务流程、让人类做出最终决定的方式来使用它们
摩托罗拉解决方案公司（Motorola Solutions）负责技术的高级副总裁保罗·斯坦伯格（Paul Steinberg）承认，许多像人工智能（AI）这样的新技术都可能要求供应商和供应商就其适当使用做出一些“重大决定”，但他希望，最坏情况下的情况并不是唯一考虑的因素
斯坦伯格本月芝加哥举行的PSCR公共安全宽带利益相关者会议上发表主题演讲时说：“我对人工智能非常兴奋，但我担心的是，人们对人工智能的潜力误解太大，以至于我们可能会把孩子和洗澡水一起扔掉。”
“一般来说，我担心的是人们认为人工智能等于面部识别，[这]等于跟踪和监控，[这]等于'坏'；因此，人工智能等于坏，这与事实不符。”
斯坦伯格说，摩托罗拉解决方案公司有三个指导原则，它确定了其产品中利用人工智能的最佳方式：
让人处于循环中人类会做出最终的决定，人工智能应该只正常的工作流程中使用确保其满足特定目的，产品应“为第一响应者制造”先进的人工智能将不会被使用-“我们将使用经过验证的技术。”斯坦伯格说，人工智能特别有帮助的一个领域就是监控视频，这是一个巨大的需求，因为固定和移动摄像头的视频监控录像量每年都激增，斯坦伯格说
然而，据斯坦伯格介绍，研究表明，监控视频的人通常没有很好的注意力广度，导致他们一个视频输入中识别关键事件的效率20分钟内从100%下降到20%
斯坦伯格说：“我们也不能很好地处理复杂性问题。“如果你给我的任务是监视一个视频源，寻找一个事件，那就忘掉时间长度吧，平均而言，我看到的时间大约有90%。如果你给我3×3-9个视频提要，看看[事件]下降到58%的概率。假设我零度。如果你把20%（20分钟的监控效率）乘以58%，我们做得不好
“坦率地说，人类做得不好，他们这类任务上表现不佳。”
斯坦伯格说，然而，人工智能可以通过基于检测变化的算法来记录视频中的项目，或者是一个可能值得关注的项目，来提供很大的帮助
斯坦伯格说，同样，人工智能也可以用于实时处理语音到文本，例如，接到911电话或现场通信的情况下。他说，这个过程允许引用一些关键术语，这些术语可以促使接线员或调度员提出更好的建议
Steinberg称，文本语音转换软件还可以帮助官员更快地完成报告，因此他们不必行政任务上花费太多时间，它可以帮助他们更有效地搜索视频和音频中的相关内容
他说：“（这些用例中）没有什么是令人毛骨悚然的。”。“里面什么都没有跟踪。里面没有任何东西可以取代人类。这只是为了增加他们本来要做的事情，这只是把这些（平凡的任务）从他们肩上卸下。”
但斯坦伯格指出，对于公司和用户来说，考虑如何实现这些功能是很重要的，因此不会带来意外的负面后果
斯坦伯格说：“我们必须非常负责任地考虑如何应用这项技术，而这正是我的工作。”。“尽管我谈到了（人工智能）的优点，但每一个优点中，如果你仔细观察，你会发现一些缺点。”
斯坦伯格说，举例来说，新技术可以分析一个人的语言模式，从而显示出一个人的状况，比如他们是累了还是有压力。虽然这些信息对急救人员（从911人员到外勤人员）解决情况时非常有用，但应仔细考虑如何共享这些信息，因为这是个人识别信息（PII）
斯坦伯格说：“一旦我确定了关于你和你的压力水平的一些东西，现这突然成为你非常独特的东西。”。“因此，我们必须非常谨慎地对待这些信息服务中的实际管理、存储和使用
“同样的例子也可以给出第一反应者的生物特征条件。”
斯坦伯格称，随着摩托罗拉解决方案公司寻求利用新技术，它采取了一种投资于正开发这些技术的初创公司的战略，这样公司就可以更好地了解这项技术的发展情况，以及将来是否可以用于公共安全。他说，摩托罗拉解决方案公司一个特定的时间内通常有20多个这样的投资
斯坦伯格说：“我想说的是，我们把创业公司实际拥有的东西应用到产品上的次数非常少。”。“我们从与那家初创公司合作中所学到的东西，并技术成熟时应用于类似概念的次数非常多。”
斯坦伯格说，近年来，摩托罗拉解决方案公司也一直收购许多成熟的公司，斥资超过30亿美元收购主要为911或指挥中心环境提供解决方案的公司
“如果你看看典型的指挥中心，你会看到技术的筒仓，你会看到无线电控制台，通话控制台，你会看到策划的调度。你可能会看到一个视频墙，”斯坦伯格说。“通常，它们来自不同的供应商，它们是不同的生命阶段获得的，它们对用户来说并不友好，因为……数据它们下面无法集成
“因此，工作流可能涵盖了我们讨论过的所有功能，但很自然，它没有。从字面上讲，你可以看到电话接线员面前有四到五个键盘和鼠标。”
斯坦伯格称，一些客户代理机构也采用新的商业模式，以帮助刺激创新
他说：“有一个不美国的国家向我们倾诉，‘我们希望你们成为我们唯一的创新供应商’，因为他们认识到，通过最低成本的投标或RFP进行创新是非常困难的。”。“所以，我们正与他们合作……一些我们认为相互重要的项目
“我们参与的项目之一就是这个行走的物联网。这种情况下，第一反应器周围有传感器，这是一名消防员，我们试图以一种有意义的方式将这些数据汇集一起。”
斯坦伯格说，开发利用新技术的产品时，从公共安全用户那里获得有关产品实际应用的反馈是极其重要的，特别是急救人员面临的高度紧张的环境中
斯坦伯格说：“我给人类的压力越大，他们就越不可能专注于造成压力的原因以外的任何事情。”。“当一名警官部署武器时，他们不会考虑打开随身携带的摄像头。他们不想改变他们的交流方式…他们也不应该。矛盾的是，他们使用技术越多，那个时候提取所需信息的认知能力就越弱
“我们现可以做很多很酷的事情。我们对人们投入的技术越多，人们可能会认为这实际上是没有帮助的，除非它做得非常非常仔细。因此，环境变得如此重要，了解人类的状况，并为此调整通信和技术。”
Motorola Solutions is exploring the capabilities of artificial intelligence and other new technologies, but the public-safety communications giant is striving to use them in a manner that improves the mission workflow and lets human beings make the ultimate decisions, according to a company executive.
Paul Steinberg, Motorola Solutions’ senior vice president of technology, acknowledged that many of the new technologies like artificial intelligence (AI) can require vendors and vendors to make some “weighty decisions” about their appropriate use, but he hopes that worst-case scenarios are not the only things considered.
“I’m really excited about AI, but what I’m worried about is that people so misunderstand the potential of it that we could throw the baby out with the bath water, as the saying goes,” Steinberg said during a keynote presentation during the PSCR Public Safety Broadband Stakeholders Meeting this month in Chicago.
“In general, what I’m afraid of is that people are assuming that AI equals facial recognition, [which] equals tracking and monitoring, [which] equals ‘bad’; therefore, AI equals bad—and that couldn’t be further from the truth.”
Steinberg said that Motorola Solutions has three guiding principles it is following as it determines the best way to leverage AI in its products:
Keep humans in the loop—Humans will make ultimate decisions, and AI should be used only in the context of a normal workflow.Ensure that it fulfills a specific purpose—Products should be “built for first responders.”Leading edge AI will not be used—“We are going to use technologies that are proven.”One example of an area in which AI can be particularly helpful is monitoring video—a huge need, given the fact that the amount of video-surveillance footage from fixed and mobile cameras is mushrooming annually, Steinberg said.
However, studies have shown that humans monitoring video typically do not have a good attention spans, resulting in their efficiency in identifying key events in a single video feed dropping from 100% to 20% in just 20 minutes, according to Steinberg.
“We don’t deal well with complexity, either,” Steinberg said. “If you give me the job of monitoring one video feed looking for an event—forget the time length—about 90% of the time I’ll see it, on average. If you give me 3 by 3—9 video feeds to look at—the probability that I’ll see that [event] drops to 58%. And that’s assuming I’m at T-zero. If you take 20% (the 20-minute monitoring efficiency) times 58%, we’re not doing well.
“Frankly, human beings don’t do this well—they suck at these kinds of tasks.”
However, AI can help significantly by noting items in video—based on algorithms that detect changes or an item that was left behind—that may be worthy of attention, Steinberg said.
Similarly AI can be used to process speech to text in real time—for instance, in the case of a 911 call or communications from the field, Steinberg said. This process allows key terms to be cited that can prompt better recommendations from a call-taker or dispatcher, he said.
Text-to-speech software also can be used to help officers complete reports more quickly, so they do not have to spend as much time on administrative tasks, and it can help them search video and audio more efficiently for relevant content, according to Steinberg.
“Nothing in [in these uses cases] is creepy,” he said. “Nothing in there is tracking. Nothing in there is displacing a human being. It’s about just augmenting what they would otherwise be doing—it’s just taking it [mundane tasks] off of their shoulders.”
But Steinberg noted that it is important for companies and users to consider how such capabilities are implemented, so negative unintended consequences are not introduced.
“We have to very responsibly think about how we’re applying this, and that’s really my job,” Steinberg said. “Even though I talked about the upsides [of AI], in every one of those upsides, if you look hard enough, you can find some downsides.”
As an example, new technology can analyze a person’s speech patterns in a manner that indicate the person’s condition, such as whether they are tired or stressed, Steinberg said. While such information could be very helpful to first responders—from 911 personnel to officers in the field—as they work to resolve a situation, how that information is shared should be considered carefully, because it is personally identifiable information (PII).
“Once I determine something about you and your stress level, now that suddenly becomes something very unique to you,” Steinberg said. “So, we have to be very careful about how that information is actually managed, stored and used in service.
“The same exact example could be given about the biometric conditions coming off of a first responder.”
As Motorola Solutions seeks to utilize new technologies, it has adopted a strategy of investing in startup companies that are developing the technologies, so the company can get a better idea how the technology is developing and whether it could be used by public safety in the future, according to Steinberg. Motorola Solutions typically has about 20 such investments in startup firms at a given time, he said.
“I would say that the number of times we take what that startup actually has and apply it to a product is very low,” Steinberg said. “The number of times we take what we have learned from working with that startup and apply it to a comparable concept—when the tech grows up—is very high.”
Motorola Solutions also has been purchasing many mature companies in recent years, spending more than $3 billion to acquire firms that primarily provide solutions for 911 or command-center environments, Steinberg said.
“If you look inside the typical command center, you see silos of technology—you’ll see the radio console, you’ll see the call-taking console, you’ll see the curated dispatch. You might see a video wall,” Steinberg said. “Often, they come from different suppliers, they were acquired in different stages of life, and they’re not friendly for the users to work with, because … the data doesn’t integrate under them.
“So, the workflow might span all of those functions that we talked about, but naturally, it doesn’t. Literally, you see call-takers with four or five keyboards and mice in front of them.”
Some customer agencies also are embracing new business models to help spur innovation, according to Steinberg.
“One country—not in the U.S.—has leaned in and said, ‘We want you to be our single supplier to innovate,’ because they recognize that it’s really hard to innovate through lowest-cost bids or RFPs,” he said. “So, we’re working with them … on some projects that we think are mutually important.
“One of the projects that we worked on was this walking Internet of Things. There are sensors around a first responder—in this case, it was a firefighter—and we tried to bring that data together in a meaningful way.”
When developing products that leverage new technologies, Steinberg said it is extremely important to get feedback from public-safety users about the practical application of a product, particularly in the highly stressful environments faced by first responders.
“The more stress I give a human, the less ability they have to focus on anything but what is causing the stress,” Steinberg said. “When a police officer deploys his or her weapon, they’re not thinking about turning on their body-worn camera. They’re not thinking about changing their communication modes … and they shouldn’t be. The paradox is that, the more they could use the technology, the less cognitive ability they have to extract what they need at that point in time.
“That gets to a lot of the cool things that we can do now. The more tech we throw at people, one could argue that it’s actually not helpful, unless it’s done very, very carefully. That’s where context becomes so important—understanding the instant of the human’s condition and adapting the communication and technology for that.” |
