+AI, Accelerating Collaborative Public Safety
Enterprise products, solutions & services
In 2016, Huawei was among the pioneers in promoting the need for public safety agencies to embrace digital transformation, which we call ‘Collaborative Public Safety.’ In essence, due to the evolving nature of threats against public safety and other operational challenges, agencies are unable to overcome these threats and challenges alone. It is no longer business as usual. Agencies, even across cities and countries, need more collaboration. The leading technologies also allow the introduction of new organizations, processes, and services to better protect public safety. More importantly, just like the digital sharing economy, agencies need to leverage platforms to reach out to communities and collaborate with them in preventing, detecting, responding to, and recovering from threats. In 2017, Huawei even launched its suite of ‘Collaborative – Command & Control, Communications, Cloud, Intelligence, Surveillance and Reconnaissance’ (C-C4ISR) capabilities and solutions to enable Collaborative Public Safety.
It is heartening to see more consulting and technology companies, and even public safety agencies, promoting the need for public safety digital transformation.
Digital transformation may seem luxurious and out of reach for many agencies. But with the changing landscape, one either transforms or faces digital disruption. Public safety agencies need to accelerate their adoption of Collaborative Public Safety.
Is Artificial Intelligence (AI) the key to accelerating public safety digital transformation? The short answer is yes. But there is too much noise, and indeed confusion, to overcome before an agency can truly benefit from AI implementation. For AI implementations to be practical and beneficial, we need to look at different angles that start with the basics. From a technological angle, we need to understand that AI can be applied at different technical layers. On the Internet one can find many articles on the use of AI in public safety, but they tend to be too prescriptive. The only common denominator between different public safety agencies, especially across countries, is their mission statement. Very often, the laws empowering them, their organizations, their procedures, their people, their technological systems, and their budgets, are different.
What we need is a generic framework of actionable use cases for different public safety agencies to plan their AI roadmap based on their legal authority, resources, requirements, and desired outcomes. I propose the use of the ‘seven A’s framework’ for public safety AI use cases.
This framework does not represent a mandatory seven-step process that public safety agencies need to embark upon to implement AI applications; nor does it represent capabilities that need to be followed sequentially. What it does is rank the difficulty of implementing AI applications from ‘Analyze’ to ‘Autonomize.’
• Analyze: The most basic and easily achieved processes, from analyzing textual data to photos to video to audio, and even to sensor data. Such analysis creates textual conversion, description, and tagging the data items being analyzed.
• Automate: Public safety involves many routine procedures that can be automated through AI applications. Examples include daily crime report generation, skill-based officer deployment, crime scene photo classification, investigation summary and facts publication, and vehicle roster generation.
• Assess: This is where AI implementation starts to get interesting. This capability is beyond individual data item analysis; it involves the assessment of the bigger picture. It is beyond knowing what occurred and when; this assessment needs to address why and how something happened. Such as why a series of similar crimes were committed in a neighborhood, or how a person is radicalized to support terrorism.
• Augment: While laws govern nearly all public safety works, many decisions have to be made by frontline officers based on their situational assessment, knowledge, and experience. This is why AI is called Augmented Intelligence in a few industries. It complements rather than replaces human intelligence. It is about helping humans become faster and smarter at the tasks they are performing; for example, providing daily reports of crime prone areas, and alerting patrol officers to the presence of known criminals in those areas.
• Assist: AI assistants are able to communicate with humans via natural language. Using the same example as ‘Augment,’ assistants can suggest patrol routes, places to visit, and people to check out based on priorities, urgent dispatches, and cost-benefit analysis.
• Anticipate: The use of AI applications to anticipate and predict crime, riots, disasters, traffic accidents, and even the whereabouts of suspected criminals.
• Autonomize: It’s not exactly RoboCop, but rather a system that includes software applications, drones, vehicles, and robots, that operates autonomously.
This ‘seven A’s framework’ serves as a generic model of actionable use cases for public safety agencies to think about while designing their AI applications.
AI is disrupting existing business models and creating new opportunities for global public safety. To date, more than 230 cities in over 100 countries have deployed our public safety solutions. What’s more, an increasing number of cities have introduced AI technology into their solutions.
AI is more than just an ICT project. Government agencies need to start with a vision and implementation roadmap — think big, start small, and use the ‘seven A’s framework’ to prioritize requirements and desired outcomes. Agencies need to consider these seven areas during the actual ICT implementation of AI projects: Connectivity, Big Data, Computing Power, Enabling Platforms, Cyber Security, Continuous Innovation, and the Industry Ecosystem.
While often overlooked, connectivity is crucial. Just as the five human senses collect ‘data’ before the brain makes a decision, AI systems need different data sources for better processing. Connectivity, especially wireless for mobility, is also necessary to allow AI systems to transmit instructions to devices.
Huawei has the most complete selection of connectivity technologies, both wired and wireless, to connect to a wide variety of data sources, and the industry’s largest petabit core routers to manage huge volumes of fast data.
• Big Data
AI can be realized through rule-based hardcoding or machine learning, which can learn through decision trees, inductive logic, and deep learning. The common denominator for all techniques is the demand for high-volume, high-velocity, and high-variety big data.
Huawei’s FusionInsight big data platform offers an extensive suite of services, including Hadoop, Spark, Flink and LibrA. The platform even has its own 200+ data models/algorithms specifically for public safety, allowing partners and customers to rapidly develop their own applications. This platform is good for four scenarios:
• Offline/near-line computing of large data sets with fewer requirements for low latency
• In-memory computing with moderate requirements for low latency
• Real-time stream computing with strict requirements for low latency
• Massive structured data analysis
• Compute Power
Big data requires high-power computing. Because AI is essentially data-driven, the output needs to be calculated quickly and accurately. Not all hardware systems are created equally, and Huawei’s specialized hardware supports the four scenarios detailed below: Offline/Nearline, Memory Computing, Real-Time Stream Computing, and Massively Parallel Databases.
In addition to Huawei’s hardware innovations, the FusionInsight big data platform can run on Huawei’s cloud technologies for better resource pooling (computing, storage, and network), and easy, flexible, on-demand self-services.
• Enabling Platforms
In a way, AI applications are limited only by one’s imagination, and public safety agencies can leverage the ‘seven A’s framework’ to conceptualize and prioritize such applications. We need enabling platforms to make it easier for AI applications to be developed rapidly, without having to worry about hardware integration and performance, or needing to seek out and connect to data sources. We also require common components without having to code them into the applications. The result is Huawei’s suite of C-C4ISR platforms that enable Collaborative Public Safety.
These platforms together with partner applications are not meant just for big cities. There are tens of thousands of cities with populations less than one million. Not only is the safety of such mid-size cities important, but we need access to all levels of data to provide accurate insights for an effective AI program. This is why Huawei offers its Safe City Compact solutions for mid-size cities. Safe City Compact can also be used in other scenarios.
In Kenya, Safe City solutions based on visualized critical communications and AI technologies helped the police shorten their response time by 60 percent, reduce the annual crime rate by 46 percent, and improve overall security. As a result, local tourism developed rapidly, growing 14 percent in 2016. In Mauritius, an AI-enabled intelligent video cloud solution greatly improved video analysis efficiency for public safety incidents. The solution also helped optimize traffic management and reduce traffic accidents. In Pakistan, a Safe City solution based on intelligent image technology was used to construct a license plate recognition system for automobiles, with a 90 percent recognition rate in daylight; and the time for terror cases reduced from 45 days to 2 days, with average incident response times shortened from 30 minutes to 10 minutes.
• Cyber Security
Due to the massive volume of sensitive data, public safety agencies have to take extra precautions to safeguard their cyber security. Any data leak is sure to breach privacy protection and lower public confidence. Worse still, data manipulation can lead to undesirable and even incorrect AI outcomes. Huawei takes security very seriously, as evidenced by our cyber security strategy and approach, especially the Integrated Product Development processes designed by IBM to assure the cyber security of Huawei products — where independent security verifications are carried out from product conceptualization and development to lifecycle maintenance. Separately, Huawei offers an AI-based unified security solution to detect, predict, and mitigate cyber threats.
• Continuous Innovation
AI adoption is a journey that must leverage continuous innovation to achieve the best outcomes. According to the World Intellectual Property Organization, Huawei was the top company in 2017 globally and across industries in terms of patent applications — with 4,024 to be exact. The second place company applied for 2,965 patents. Not long ago, Huawei launched two AI-related products: a Software Defined Camera and an Intent-Driven Network Switch.
• Industry Ecosystem
One company alone cannot implement AI. It needs an entire ecosystem. AI implementation needs to follow open standards to prevent vendor lock-in and ensure interoperability. Huawei is glad to support and adopt more than 30 open standards, and we have more than 1,000 partners providing applications on our platforms. These partners are supported by our OpenLab facilities around the world.
As a leading ICT company, Huawei’s capabilities include a complete AI portfolio and a full stack of technologies to cover all scenarios, from devices and connectivity to cloud platforms. The technologies include:
• Ascend series of specialized AI chips; based on a unified, scalable architecture
• CANN chip operations library and automated development toolkit
• MindSpore unified training and inductive framework for device, edge, and cloud
• Application enablement for end-to-end services (ModelArts), multiple-level APIs, and pre-integrated solutions such as the public safety solutions behind our C-C4ISR capabilities.
With more than 6,000 professionals directly supporting Huawei’s public safety solutions, including many with frontline public safety operational experience, we are in a good position to help design the architecture for AI implementation. Not just technology architecture, but also business architecture, data architecture, and application architecture.