AiMEN, the Pontifical Artificial Intelligence

Screenshot of the promotional website AiMEN for the mini-series The Young Pope

From Paolo Sorrentino, director of the Academy Award®-winning film ‘The Great Beauty,’ ‘The Young Pope’ tells the controversial story of the beginning of Pius XIII’s pontificate. Born Lenny Belardo, he is a complex and conflicted character, so conservative in his choices as to border on obscurantism, yet full of compassion towards the weak and poor. The first American pope, Pius XIII is a man of great power who is stubbornly resistant to the Vatican courtiers, unconcerned with the implications to his authority.

What if in the XXIst century a Pope decided to use an Artificial Intelligence to preach God’s word to Internet users?

AiMEN is the new pontifical Artificial intelligence that scans all messages posted in real time on the social networks to preach God’s word to Internet users. AiMEN was born from a collaboration between makemepulseBETCCanal + and IBM Watson.

Internet and more particularly social networks is a place full of hatred and negative feelings. The main goal of AiMEN is to spread a positive speech to all these users. AiMEN analyses the messages posted on the four main platforms (Twitter, Facebook, Youtube and Dailymotion) and answers with a Bible verse chosen according to the various subjects.

A serverless AI

As always, here at makemepulse, we love to use cutting edge technologies. The application architecture design was created to prevent any scalability issue. Based on the powerful serverless framework we have designed a scalable crawling application.

Graphic of the AWS infrastructure.

For all input sources we have a specific crawling method (streaming on twitter, calling REST api on Youtube, …), but the processing of all messages is the same. For each new message, we get recursively all the associated messages (comments and comments thread). This crawling is made using an SQS queue that triggers AWS lambda. As soon as a message is recovered we store some information in a DynamoDB table. Then the message is put in a new queue that triggers new lambda for the message analysis using IBM Watson. An analysed message will not be analysed a second time, it’s a one time operation.

Machine learning with IBM Watson

We have used IBM Watson with their powerful Natural language classifier to analyse all the messages from the database. The IBM Watson™ Natural Language Classifier service uses machine learning algorithms to return the top matching predefined classes for short text inputs. The Natural Language Classifier service can help your application understand the language of short texts and make predictions about how to handle them. A classifier learns from our example data and then can return information for texts that it is not trained on.

The service employs a new set of technologies known as “deep learning” to get the best performance possible. Deep learning is a relatively recent set of approaches with similarities to the way the human brain works. Deep learning algorithms offer state of the art approaches in speech recognition, and the Natural Language Classifier now applies the technologies to text classification.

Results

During the operation (a little more than 2 weeks), we analysed almost 4.5 million messages, 10 giga of text data, invoked more than 15 million lambda, answered to 900 000 users and exposed 2.6 million people to AiMEN. The average analysis time of a message was under 2 seconds.
Well, that was a lot of data and a great success !

GIF of Jude Law incarning the Pope in The Young Pope, saying "It's all I think about. I even dream about it".

You can view a data visualization of AiMEN on aimen.makemepulse.com

Discover more on AdWeek.

We hope you enjoyed the reading.