In our earlier blog post, “How are consumers adopting artificial intelligence?,” we discussed different ways in which artificial intelligence (AI) is rapidly integrating into consumers’ daily lives. Businesses have taken notice, and according to IDC, worldwide spending on cognitive and AI systems will grow 54% in 2018 to $19 billion. Further, 40% of digital transformations are expected to use AI by 2019, and 75% of enterprise applications will use AI by 2021.
The AI juggernaut is not limited to the private sector; several federal and state governments have already made AI integration a key priority. Earlier this year, the Chinese government announced a roadmap aimed at making China the global leader in AI by 2030. As part of the public-private AI partnership to accelerate AI development and commercialization, the Chinese government named four official national champions: Baidu, for autonomous driving; Alibaba, for improving urban life and smart transportation; Tencent, for computer vision for medical diagnosis; and iFlytek, for voice recognition and intelligence.
Closer to home, the U.S. federal Investment in R&D for AI and related technologies has grown by 40% since 2015, with both general AI research and applied research in autonomous and unmanned systems designated as administration R&D priorities for 2019. At the U.S. state level, 58% of state CIOs named AI and machine learning the most impactful area for them in the next three to five years. In this blogpost, we will examine four ways in which federal, state, and city governments are adopting AI: infrastructure delivery, social and welfare services, national security, and law enforcement.
Let’s start with infrastructure delivery. Federal agencies are aggressively using AI to forecast and recover from natural disasters. The Federal Emergency Management Agency (FEMA) used computer vision analysis of satellites images during the Kīlauea eruption in Hawaii this summer to determine which buildings had been damaged and whether people were evacuating ahead of the lava’s approach. The National Center for Atmospheric Research has recently partnered with IBM Watson to forecast severe weather events more accurately, thus enabling warnings further in advance of events.
While federal agencies have taken the lead, states and cities are not far behind. Maryland has upgraded traffic signals to respond to traffic conditions in real time, and predictions based on machine learning have reduced commute times by 10–15%. Kansas City can now predict where potholes will form based on past trends, current weather data, and existing road conditions. Armed with this information, they can repave streets preemptively instead of waiting for potholes to occur. Additionally, Chattanooga is using smart grid technology to better manage power needs and mitigate outages during extreme weather.
In social and welfare services, the most significant impact of AI is in education services, with AI investment in U.S. education expected to grow 48% by 2021. Georgia Tech used an AI chatbot to respond to student questions about a required online course (see also an earlier blogpost: “How do digital voice assistants work?”). Tacoma public schools recently partnered with Microsoft on a predictive analytics project to identify students with a high likelihood of dropping out, and they have seen improvement in graduation rates from 55% to 83%. In China, schools are piloting an “intelligent classroom behavior management system” that can monitor students’ engagement and activity levels.
State and city bodies are also using AI to address homelessness and deliver child welfare services. The Center for Innovation through Data Intelligence (CIDI) in the New York City mayor's office uses machine learning to cluster high-risk and formerly homeless individuals into various groups (e.g., those with frequent jail stays, consistent supportive housing, consistent subsidized housing, earlier homeless experience, later homeless experience, and minimal service use). Through this approach, the city can better understand the needs of each cluster and identify resource requirements accordingly. In Finland, the city of Espoo analyzes 280 different factors to predict when families will need child welfare services so that support can be offered before serious issues arise.
Federal governments are using AI across a wide range of overt and covert national security objectives and in support of both public and private enterprises. Military drones and robot soldiers are already being used as the first line of attack, both in search and rescue as well as in assisting the military during combat. Intelligence agencies are reportedly supplementing spies with AI to gather information virtually, thus reducing risk to agents. Federal cybersecurity teams are using AI for intelligence gathering and active threat hunting to “deal with the new wave of sophisticated adversaries.”
To support critical infrastructure, such as financial institutions and utilities, the federal government has outlined a new cybersecurity policy and set up infrastructure to respond to virtual attacks. Given multiple instances of users’ data breaches (see also: “Why do data breaches happen?”), the IRS is using AI solutions to help protect consumer information.
AI is also being used in law enforcement. In the past, police would send patrols to areas where crimes had previously occurred. Using AI tools like PredPol and CrimeScan, they can now predict where crimes will happen and send patrols accordingly. Some cities have found AI is two times better at predicting crime than a human analyst. At crime scenes, police are increasingly using computer vision–enabled identification technology (see also: “Why is facial recognition the new face of innovation?”) tied in with body cameras to identify suspects in real time.
With regard to mass social profiling, no other government has publicly used AI as extensively as China. The Chinese government has instituted a “social credit” system that will be fully implemented by 2020 to dynamically rank all citizens. The score will predict bad behavior, and people with low scores may be unable to purchase train or plane tickets, have their internet speeds throttled, be barred from schools or employment, or be publicly called out as bad citizens.
For all the direct and indirect benefits, AI also comes with challenges. As in the case of private enterprises, governments must address two key issues in using AI—preventing existing bias in historical data from negatively impacting citizens (especially in military and criminal use cases) and ensuring that users’ privacy is not violated while tracking, collecting, and analyzing their data.
The good news is that governments already have the basic rules and guidelines in place for protections like privacy, cybersecurity, and health and safety practices. The federal government should use these as a starting point to shape the future of AI, hopefully through a “legislation not regulation” approach. When it comes to protecting citizens’ rights, expecting private sectors and local governments to self-regulate is not an approach that can work.
We would like to sincerely thank Colleen Monaco, Director of eCommerce @Disney and a guest lecturer at USC Marshall School of Business, for generously donating her time and insights to this blogpost.