On the sidelines of an AI for Good conference in Geneva, where humans and robots mingled and discussed how machine learning could help global development, IRIN rolled up its data science sleeves and deployed a bot on an important new challenge: making up names of aid agencies and aid job titles.
It’s a new take on some of the jargon and repetitiveness that can be found in the relief and development sector, and hopefully gives a taste of the rudiments of machine learning.
First we downloaded a list of over 9,000 aid organisations from the UN’s Financial Tracking Service (FTS) database. The A-Z list goes from “A Call to Serve International” to “Zwanan Development Organization”. Then we fed them into a machine learning system running open-source code in the Python language.
The aid organisation names were processed by a recurrent neural network built on the TensorFlow system. We won’t (also, can’t) explain how the maths works in detail. However, it functions like any machine learning process – the foundation of the current explosion in “AI” across a range of sectors – it analyses the patterns of letters and words and then attempts to generate new examples using the rules it has deduced from the original data.
(Our idea to explore the jargon and nomenclature in aidspeak owes much to Janelle Shane, who feeds machine learning algorithms with unlikely lists, such as the names of craft beers or the names of pet guinea pigs).
Here are some of our favourites, generated entirely by code.
Aid agency names produced by AI
Some sound quite plausible – if you do set one up, please let us know and send the logo.
Support International Aid Foundation
Community Aid Association for Development
Association des Femmes de l'Environnement
Medicultural Community Foundation
Some a little pompous:
International Council of Society
Internationalist Fund for Rehabilitation
Others, possibly creative and/or radical:
Sure Vision for Red Cross
Africa Relief and Empower Swedish Family
People Trust of South ye Community Humanitarian Family
Water World Sudan
AI-generated NGO: People Trust of South ye Community Humanitarian Family
Sometimes our bot seemed to get a little lost and confused:
Alliance for Development and Development
Miseration Mission for Comment of Africa
Alliance Save the Development International
Gy Alliance Conflicted Now Women International
El Organisation des Futures
And, some are just plain weird (sorry not sorry):
ACF and Development Porn
Scam for National Aid
Bin Welfare Relief and Training
Next, we moved onto humanitarian job titles. We fed the bot more than 2,000 vacancies on the UN’s ReliefWeb jobs pages. And these are some of the best new jobs titles it came up with. Perhaps some of these fit your current role? Or make you wish you could apply?
AI-generated job title: Regional Regionalization of International Programme for Head of Party
Humanitarian job titles invented by AI
Chief of Party (Sex)
South Changes Manager
Chef of Finances and Grant & Manager
Regional Leader (Interant) Livelihood and Consultant
Multiple International Director
Nation demonic Manager
Consultant Africa and Charging and Programme
Regional Regionalization of International Programme for Head of Party
Communications Coordinator and Memaliature Manager
Integration Director (Support) (H/F)
International Parent (H/F) Adviser
Street Communication Laborator
Development Coordinator – Safety Senior Charge Policy Specialist
And some that sound more like entry level roles:
Engager Programmes Intern
Project (unimation volunteer)
Editor Intern Coordinator
As always, comments and reactions are welcome below - tell us your favourites!