Last week, Google succeeded in creating an Artificial Intelligence that defeated a Go Grand Master. This was a landmark achievement for AI research. The future will be AI systems. So where do we stand when it comes to AI in Sri Lanka? To answer this question, we decided to attend the Colombo AI March Meetup. Here’s what we found.
The event opened with Dr. Srinath Perera – VP Research at WSO2 taking the stage. Dr. Srinath opens his presentation by introducing the different types of analytics. The different types of analytics are:
- Predictive analytics
- Streaming analytics
- Batch analytics
Machine learning is a subset of artificial intelligence, which is about systems learning to carry out new tasks based on data from previous tasks. An example of machine learning would be the Google self-driving cars. If you asked an amateur programmer to build such a system, they would say, “Okay! I’ll code it.” However, an experienced programmer would say, “There are too many test cases!” With machine learning, the system would be constantly learning how to tackle new test cases, based on the data it has about previous test cases.
He then went on to introduce the WSO2 Machine learner. With it, you can upload some data or select some existing data. Once you’ve uploaded this data you can then explore the data by visualizing it through different graphs such as scatter plots or bar graphs and identify the different fields. This visualization helps you compare results and better understand the data.
“The real fun in machine learning is inside the system. If you try explaining it to your mother, you’ll have a tough time.” – Dr. Srinath Perera
He then shares a few real-life examples of machine learning in action. From estimating how long it takes to get through airport security to identifying defective parts on a manufacturing line, machine learning has a wide range of applications. Our favorite was the example of WSO2’s famous super bowl prediction system, built using WSO2’s Machine Learner. This system collected data about all the teams taking part in this year’s super bowl and then predicted the possibility of one team winning over the other. At the end of the day, WSO2 beat Microsoft Cortana and predicted 7 out of 11 Superbowl games successfully.
Having shared a few use cases, Dr. Srinath concluded his presentation by sharing a few things they are working on at WSO2. One such project is regarding IoT analytics. Much of the data collected via IoT devices will be significantly different from the data we collect today. As such, machine learning will be needed to properly understand this data. With that, Dr. Srinath concluded his presentation.
The second speaker of the day was Prof. Asoka Karunananda – Dean of the Faculty of Computing and Research & Development at KDU. Asoka’s presentation was the simplest of the presentations, which gave us a broad introduction as to what AI is. One could say that this should have been the first presentation of the day.
He began with an introduction to the history of Artificial Intelligence. AI was born in 1956 at Dartmouth, USA, with John McCarthy coining the term. The goal of AI back then was to develop a machine that has natural intelligence. While the goal was clear, there was much debate amongst the community regarding John McCarthy’s term. However, the term Artificial Intelligence was established and today it’s an important subject in Computer Science and Engineering.
Prof. Asoka then highlights some issues regarding AI research today. The key issue is that people are afraid of AI because it has the power to control almost every aspect of our lives. However, Prof. Asoka went on to say that AI will be at the heart of all software solutions in the future. What we need to ensure is that we don’t build AI systems that we can’t control.
Having spoken about the key issue, Prof. Asoka then gave us a broad view of AI before introducing us to the four schools of thought regarding AI. The four schools of thought are:
- Expert Systems: The original school of thought that wanted to create machines that mimic humans
- Thinking Humanly: The second school of thought expanded the idea of the previous school and wanted to create machines that behave and act like humans.
- Thinking Rationally: The third school decided to create systems that follow logical rules. However, this might not be applicable to all situations
- Acting Rationally: This is the modern approach to AI. The goal is to create a system that focuses on doing the right thing.
Prof. Asoka then went on to talk about the changing landscape of Artificial Intelligence. According to him, there is a growing interest in modeling the human mind. The goal is to reduce the gap between man and machine. This could be done by either adding biological brains to robots or adding machines to biological brains. It may sound like science fiction today, but there are people working on making such things a reality. One such person is Steve Potter, who created a robot powered by a rat brain.
Prof. Asoka then concluded his presentation by showing us where Sri Lanka is in terms of AI research. As he mentioned at the start of his presentation, AI is an integral part of computing and engineering fields today. This is why it’s all too common to see AI being a regular research topic among local universities. In the future, local companies will be regularly working to build software solutions with AI.
Following Prof. Asoka’s presentation, we saw a panel discussion about AI research. The panelists were: Dr. Srinath, Prof. Asoka and Ramesh Maddegoda – software architect at Virtusa. The panel discussion answered many questions regarding Artificial Intelligence the audience had. A common topic was what anyone should learn to build AI systems. Prof. Asoka said it’s important to look at multi-agent technologies and have an understanding the mind. Srinath and Ramesh suggested we look at deep learning technologies.
Artificial Intelligence is the future and there’s no stopping it. We in Sri Lanka are at a point where we can catch up with the rest of the world in terms of AI research. If we do so, then eventually we will regularly be building software solutions utilizing artificial intelligence. The only thing we need to make sure is we don’t create a racist sex robot like Microsoft accidentally did.