Tech Briefs is reporting from CES 2019. Send us your questions and comments below.
LAS VEGAS, NV – In 2017, the international community faced a devastating crisis.
More than 20 million people across northeast Nigeria, Somalia, South Sudan, and Yemen were at risk of famine – a dire condition of food uncertainty.
The World Bank, along with international partners, responded by providing a $1.8 billion relief package, potentially avoiding catastrophic loss of life.
“But in some ways, we, as an international community, failed,” said Ed Hsu, a World Bank Group senior advisor, at a panel at this year’s Consumer Electronics Showcase conference in Las Vegas.
Livelihoods were destroyed, people impacted by famine slid further into poverty, and many lives were still lost, said Hsu.
The UN and the World Bank sought a more efficient way to find the areas of the world vulnerable to famine, and to provide a kind of “bridge” between humanitarian efforts and the communities needing them most.
“We have to make sure that people living in conflict and in affected situations also have access to the help they need,” said Hsu.
To provide better predictions of high-need areas, the World Bank is turning to artificial intelligence (AI) and machine learning.
By partnering with a team of global technology firms, including AI heavyweights Microsoft, Google, and Amazon Web Services, the World Bank hopes to build on existing famine early-warning systems, using machine-learning and AI to better forecast the most at-risk areas.
Starting a “FAM”: The Famine Action Mechanism
With famine, there is a clear cost to a delayed response in funding and action.
In July 2011, a famine was officially declared in Somalia. Over the course of the prior year, however, 78 bulletins and 58 briefings had been issued by famine early-warning systems.
“This delay and lack of early action contributed to the tragic loss of a quarter of a million people,” said Hsu.
The Famine Action Mechanism – an effort developed by the World Bank, United Nations, ICRC, and other global partners – uses artificial intelligence and machine learning to formalize links between early warnings, financing, and implementation arrangements.
In other words, financing is needed much earlier in the relief process, as soon as the risk of famine is detected, according to Hsu.
To ensure funds are released before a crisis emerges, the FAM will seek to make financing more predictable and strategic by, for the first time, linking famine early warnings with pre-arranged financing.
“We have to start thinking like an insurance company,” said Hsu. “When there’s a disaster, we’re covered.”
Currently, the international community, including the Food and Agriculture Association and the World Food Program, conducts comprehensive assessments of food security and other famine-related variables like poverty rates, drought, and conflict. Those surveys, however, take a significant amount of time and resources to collect the data and update the assessments.
The FAM aims at filing in the gaps between these major reporting cycles.
The initial work is promising, says Hsu, and the team is now working on providing models that can provide real-time forecasts six months out – all the way down to the district level.
“Embedding these forecasts into the work of the World Bank and the UN will really shift our paradigm to one focused on prevention,” said Hsu.
The FAM will be rolled out initially in five countries that demonstrate the most critical and ongoing food security needs, and ultimately will be expanded to provide greater global coverage.
Bringing in Competitors – To Cooperate
The Amazon Web Services team developed a machine learning “pipeline,” pro bono. The machine-learning system was created by sourcing data that could then be used to make famine-related inferences.
The artificial-intelligence system pulled from five important source-data points: Market Prices; Vegetation; Weather; Conflict Information (parsed out via text analytics); and Prior Famine Designations. The data gathering offers the kind of real-time information that cannot always be found in the traditional six-month assessments.
“We believe that machine learning is going to be an important way to be able to predict where famine will occur in the future, to help us get help to these very needy people,” said Larry Pizette, Senior Manager at Amazon’s Machine Learning Solutions Lab, at the CES presentation.
Microsoft, Amazon, and Google – three top AI companies in the world – are assisting the World Bank and the UN in the development of the AI model. Such an important task requires a unique kind of cooperation between three companies that are often highly competitive with each other.
“By convincing them to collaborate together, we could come up with a better product than if we worked with any one of them,” said Hsu.
Machine learning requires learning, after all. The World Bank has the famine expertise, and companies like Microsoft, Amazon, and Google have their own unique machine-learning know-how.
The efforts must be addressed as a team, and the more information that a machine-learning system has, the better determinations it can make, according to one member of the Amazon team.
“Having committed partners, having shared goals, and working together on them as a mission is really what drives success,” said Pizette.
Today, 124 million people across 51 countries face crisis-levels of food insecurity requiring immediate assistance. Famine is the kind of challenge that requires the best experts in the room, even if they’re competitors.
This is what the World Bank Group is about now, says Hsu: Creating partnerships between the private sector, the public sector, and governments to come up with solutions that were impossible before.
“We believe famine is our collective failure as an international community,” said Hsu. “Preventing it requires a collective solution.”
The presentation at this year’s CES was part of the conference’s first Resilience/“Smart Cities” program, which emphasizes how technology can be used to help areas prepare for disasters.
Share your questions and comments below.