What happens when you ask an AI to come up with charity names?
Through a light-hearted experiment, CAF’S Rhodri Davies shows that even the smallest charities can play with AI bots without needing specialist knowledge or breaking the bank.
This article was written by Rhodri Davies, who leads Giving Thought, CAF’s in-house think tank focussing on current and future issues affecting philanthropy and civil society. Rhodri has spent nearly a decade specialising in public policy around philanthropy and the work of charities, and has researched, written and presented on a wide range of topics. He has a first-class degree in Mathematics and Philosophy from the University of Oxford.
What do all of the following charities have in common?
- Lewisham Young Farmers Club
- Much Society
- Four Winds Relief
Firstly, they’re all made up. And secondly, they were created by my AI bot.
Artificial Intelligence (AI) used to be the preserve of science fiction. But now AI is a reality in our daily lives, and this connection between humans and AI looks set to become deeper and more inextricable in the future. Amazon’s Alexa, Apple’s Siri, Netflix; they all use artificial intelligence to assist us every single day. Some people are nervous about the potential for AI to be abused; and it’s easy to understand why.
Just a few weeks back, the non-profit company OpenAI – backed by Elon Musk amongst others – revealed that their new AI fake text system, which can generate fictionalised news stories, may be too dangerous to release. Afraid of possible misuse, for now the research behind the GPT2 system will remain under wraps.
However, AI-backed text generator systems aren’t some mystical, murky creation. To get a better sense of what this actually means in practice, I put together my own homemade AI system. As Head of Policy at an international charity, I wanted to give it a charity focus, partly because I was curious to see how hard it is (answer: not very), and partly because artificial intelligence is already revolutionising the way charities work. And not that many people are aware of it.
For example, since 2016 The Children’s Society has used AI-powered live translation tools when speaking with young refugees and migrants who have limited English language skills. Using the software, conversations are translated in real time using a mobile phone or Skype. Even in its relative infancy, artificial intelligence is helping to undermine the idea of things being “lost in translation”.
So I decided to build an AI system that creates suggestions for charity names, using machine learning. The results are based on feeding the names of all registered charities in England and Wales into a text-based neural network, which detects patterns and learns as it goes.
The results were often amusing, occasionally creepy, and also very insightful. For instance, it becomes clear very quickly from the list of names that religious and educational charities are widespread and that the UK charity sector is largely made of up very small organisations.
And to get a sense of the quirkier side of this experiment here, in no particular order, are some of the most memorable fake charities spewed out by my AI bot:
- The Friends of Table
- Youth and Friends of Surgery Castle
- Joseph Syndrome Trust
- People for a Team
- We a Little Blind
- Radio Fire Charity
- Rupert Studies
- Clubs International of Funeral Teachers Imagine and Nursery School
- Cross Child Contact Centre
- Friends of Hilda with Physical School
- Man First Foundation
- Cubbington Cat Club
After I’d got over the fact that I’d built an AI bot that named something “We a Little Blind”, I hit on a couple of more serious points about the nature of artificial intelligence itself.
First off, experimenting with things like AI doesn’t have to be that hard or expensive. Yes, my example is quite silly, but I did it using entirely free and open source tools with publicly available data from the Charity Commission. I was able to muddle through it by myself, and I’m definitely not a data scientist or developer. AI doesn’t have to be super-exclusive or impenetrable to the outsider. It’s actually quite the opposite.
Furthermore, tinkering with this bot gave me a sense of what AI actually is, and the fact that it is much more boring than often portrayed. Most of the time it isn’t about Minority Report style headsets, and much more about fiddling with file types and spreadsheets. (Which isn’t to say that isn’t potentially very powerful!)
Last but not least, the diversity of UK charities is amazing, and most of them are small, local ones. Only 9% of British charities have paid staff, with 91% entirely reliant on unpaid volunteers. Seeing the names that the AI bot generated really brings that point home: the vast majority weren’t as ridiculous as the amusing ones I cherry-picked above and just sounded like tiny little organisations run for local communities.
With a little research and some publically available data sets and platform tools, you too can experiment with AI to your heart’s content. For anyone looking to find out more, you can tweet me @Rhodri_H_Davies and I’ll be happy to steer you in the right direction.
So if you’re reading this and you’re even a little interested by how AI works, and wouldn’t mind knowing more, I encourage you to just have a go. You’re highly unlikely to create a creepy killer robot. And you’ll also have the chance to experience first-hand the technology that’s changing how we as human beings interact with the world.