On my last work day of my trip to Singapore, I caught up with a boss and mentor of many years, the Head of Civil Service. One of the questions he asked me was where I thought AI has the greatest potential. With 3 seconds to think, I said: where we have the biggest problems, and where we have data (or the potential to collect it quickly, at scale).

I shared the areas I am passionate about:

1.    Skills and Education. In an era of rapid job displacement, how can we constantly re-skill and place people in emerging jobs, at scale, without additional manpower resources? This is traditionally done through centralized planning, but the speed of change will render this approach ineffective. To achieve scale we must empower individuals, employers and training providers through better information, better matching, customized motivation and pathways – these can all be supported by AI, but there are other essential pieces to the puzzle. For example, we need a common Skills Framework that combines top-down skills trees with bottom up self-reporting of skills. We need to help everyone use the same language in describing skills: educators, students, workers, employers. Linkedin and Google are already working on parts of this story. It is an area ripe for public-private partnerships.

2.    Transportation. How can we optimize the flow of people, goods and services at the country level, while helping individuals feel that they benefit? Autonomous vehicles solve many problems, but one important objective is enabling all our vehicles to be optimized as a system, rather than as individual units. AI will help us optimize, but ultimately we need to deal with a very human issue: individual commuters who feel that they are sacrificing something for system-level optimization. The crux is shifting user preferences – incentives and policy complements will be just as important as the technology.

3.   Healthcare. One of the biggest problems in healthcare is enhancing quality while containing cost. Enabled by AI, how can clinical and policy interventions for a population be more upstream, targeted and outcomes-based? Imagine the benefits if we can prevent the onset of disease, manage disease before it reaches the extremely costly stage, and administer interventions based on personal – rather than general – outcomes.

But the Devil is really in the details

With the proliferation of data and advanced AI techniques, the use cases for prediction, optimization and customization are infinite.

However, if we want AI to truly deliver impact, the devil is in the details of implementation and organizational change. My current job gives me a wide view of all the different technology activities going on in Singapore. On my trip, I touched base with a wide range of folks working in technology domains – health, energy, transport, digital government and ports. I also had in-depth conversations with people who were building capabilities in the technology sector – creating a critical mass of industry capabilities, establishing a data sharing governance framework, enhancing talent development. We talked about the challenges of their work.

I was reminded that lofty goals inspire, but equally important are the small steps that will enable AI (and other technologies, for that matter) to be deployed for the maximum public good. This includes:

• People and processes. Our workforce has to trust and use new technologies. This will not come easy, if technology is perceived to be threatening or complex. I saw first-hand how a pharmacy in one of our hospitals introduced robots to sort and pack medicines. Not a single pharmacist lost their job in the process – they were retrained to invest more time in customer-facing roles.
• Governance and organizational changes. Decisions around technology investments need to be made by domain-area experts in a far more rapid and iterative way, rather than by traditional hierarchies in long sales-cycles – this is not how Governments are typically structured and we must change; There is a natural tension between delivery and experimentation in a Government’s technology agenda, since resources are finite. There is also a chicken-and-egg issue which lies in having ready use cases before collecting data, and collecting data
• Societal changes. Huge behavioral changes are needed to achieve any vision. How can we make the technology-enabled option so attractive that people prefer it over the options they are used to? We often underestimate the power of inertia. Tech companies have shown that clever user design, incentives and achieving a network effect at scale can help. This is a gold standard. Governments have a lot to learn
  

  On the need for strong partnership between the Tech and Gov communities

Back to the three areas where I believe we have the biggest problems (and hence AI can make the biggest impact). Health, education and skills, and transportation are areas where tech and government cannot afford to work without each other:

–      We will get the best outcomes if we make these shifts at scale – if say, an entire city or country is on-board. Companies have the technological expertise, countries (at least some of us) have the mandate to bring different stakeholders together.

–      Without the right policy complements, technology won’t achieve the desired impact: autonomous vehicles may lead to more congestion, better information on personal health may lead to over-consumption and cost inflation. I write more about it here.

My visit home strengthened my conviction that the government and technology communities need to work much closer together to deploy technology for public good: as we introduce new, tech-driven ways to solve big societal problems, tech companies and Governments should co-design the surrounding policy and regulatory environment and put in place incentives, nudges and public education. In addition to clever AI techniques, all these pieces have to be in place to achieve true impact.

==

I’m interested: if you were asked the same question (where does AI have the greatest potential), how would you respond?