510 & OUR USE OF AI

WHAT IS ARTIFICIAL INTELLIGENCE

Artificial intelligence (AIis, as its name suggests, a form of intelligence that mimics the natural, human thought processes, but is displayed by machines, particularly by computer systems. Expressions of this intelligence are, for example, learning, problem solving, & self-correction.  

Machine learning is an application of AI and refers to a computer’s ability to learn without being explicitly programmed.  

Deep learning is a class of machine learning algorithms, focusing on processing (big) data to ‘learn’ to identify complex patterns automatically. 

WHO IS IT FOR

AI has a great potential to assist humans in many tasks, both simple and  challenging ones. Its applications can, and already are serving, healthcare, transportation, & finance. At 510 we are exploring its potential in Humanitarian Aid.

WHY IS IT NEEDED

At 510, we are building up toward using AI to help with disaster preparednessearly warning early action and response. 

WHAT ARE WE DOING  

Nowadays have access to multiple digital sources and tools that can inform on what’s happening at a particular location. For example, we can use social media (Facebook, Twitter), websites, maps, satellite images and collaborative maps such as OSM to focus on a specific area.  

We have been using AI models for the following purposes: 

     1. To get a better image of the situation on the ground 

Timeliness is of utmost importance during a disaster. In the event of a disaster or crises an AI model can assist us in two ways: 

  • To identify which buildings are affected  
  • To assess physical damage of those buildings  

    2. To make better, more informed decisions, faster 

AI models can return an assessment of damage after a disaster, identifying areas with no visible damage, and areas with partial, complete or extensive damageWe share this data with humanitarian workers on the ground so that they can better organize and prioritize their response.  

    3. To predict disasters and intervene before the damage is done 

Data from previous disasters can be used to create machine learning models to predict damage during future such events. Thus, it becomes possible to prepare and reduce the impact of a disaster. We use text mining and natural language processing techniques to extract impact data from humanitarian documents. 

Within all these categories, any use of data must respect the ethical and practical standards we set in our data responsibility policy 

WHAT ELSE CAN WE DO

As part of our innovation efforts, we see further applications for AI, such as:  

  • Automated information retrieval from social media 
  • Automated damage assessment from mobile pictures 
  • Automated damage assessment from satellite images 
  • Building material classification from street view images  

WHEN & WHERE WE STARTED

We have been working with AI since 2016, when we began developing a damage assessment tool using deep learning.  

The blog posts below pertain AI applications so far: 

Damage assessment tool

Finding impact data in humanitarian documents

Can text mining optimize humanitarian response

Written by Catalina Barzescu 

Comments are closed