Artificial Intelligence Can Counter Hurricanes
Guest post by Amélie Buc, Stanford AI4ALL ’17 (formerly SAILORS)
AI4ALL Editor’s note: Meet Amélie Buc, a 2017 Stanford AI4ALL (formerly SAILORS) alumna. Below, Amélie explores the ways AI can be used to facilitate natural disaster relief.
This summer, I had the incredible opportunity of attending AI4ALL’s Stanford AI4ALL program at Stanford University. The experience was extremely fulfilling; by learning from some of the most brilliant minds in AI about everything from robotic assistants to the use of AI in flight systems and medical research, I found in AI a field in which all of my interests could coalesce.
The chance to explore AI from such a deeply immersive and multifaceted standpoint was singular. I was curious about AI, but before Stanford AI4ALL never realized that the field is in full swing and impacting everything from the way we drive to how doctors perform surgery. I was also surprised to discover at Stanford AI4ALL that AI can help humans achieve deeply humanitarian goals. One of these that I find particularly interesting is the mitigation of the impact of natural disasters.
In 2017, the United States has not been estranged from the consequences of natural disasters. On August 25, Hurricane Harvey became the first Category 4 storm to hit the US since Hurricane Charley in 2004. Texas was flooded, making transportation of resources difficult and identification of those who were in greatest need difficult. The storm killed 39 people, and thousands more did not have the resources to rebuild their lives. On September 5, Hurricane Irma became a Category 5 storm as it hit the northeastern Caribbean and the Florida Keys. By October 10, the storm had killed over 160 people.
One of the greatest impediments for rescue workers and volunteers is that prioritizing strategies for aid is extremely time consuming, as identifying those most affected takes considerable coordination and consideration of certain security factors.
In recent years, those looking to stop extreme weather’s destruction of communities have found a new source of aid: artificial intelligence.
We explored one means of using AI for this purpose at Stanford AI4ALL. A group of us had the chance to program a natural language processor — a type of AI algorithm — that streamlined rescue workers’ process of turning to Twitter to see what people are needing and offering. We used our algorithm as if we were working during the aftermath of Hurricane Sandy. First, we accessed a set of hundreds of tweets pulled from Twitter during Hurricane Sandy; these tweets were already pre-grouped by people into categories based on topic, such as “Food,” “Water” or “Miscellaneous,” and whether they were “Offering” something or “Needing” something. This data, called a training set, was fed into the algorithm, which was written in Python. The algorithm used word probabilities to learn from the training set which words were most likely to appear in which categories, and then was able to quite accurately categorize new, unlabelled tweets that were fed into the algorithm. If one was using this algorithm in real time for relief work, one could then pair “Needing Food” tweets with “Offering Food” tweets, for example. The algorithm proved an efficient way to possibly help others with readily available technology.
Using AI to facilitate an exchange of resources is one option. The Netherlands Red Cross is exploring another in their new initiative, 510. Intended to improve aid response in the Philippines, often hit by typhoons, 510 analyzes data such as wind speeds, rainfall in affected areas, and patterns from past disasters to within 24 hours create a “Priority Index.” The index extremely accurately predicts which communities have been most severely affected by a typhoon. 510 has already managed to speed up distribution of resources to those areas highest on the Priority Index.
The rise of social media and open sources of live data during natural disasters, such as Snapchat Stories, Instagram, and Twitter, are providing ripe opportunity for the growth of AI in the disaster-relief sector.
With natural disaster relief, “you can’t afford to lose time,” says Rafael Lemaitre, the former national director of public affairs of the Federal Emergency Management Agency (FEMA). Technology “[connects] people and the resources they need in a much faster way.”
About Amélie
Amélie Buc is a sophomore at the Trinity School in New York. She is a Stanford AI4ALL ’17 alum and an active member of organizations that promote diversity in STEM, including Girls Who Code. In her work with New York’s YouthAction organization, she has also approached numerous local congressmen and other government officials about improving after school and summer job employment opportunities for underprivileged teens, as well as increasing allocation of funding towards technology in schools and public libraries to mitigate educational disadvantages caused by lack of equal access to internet services and information. At school, she has founded the UN Women and International Politics Clubs and is a member of the Model Congress, Model UN and debate teams as well as the Girls Learn International and Girl Up clubs.
