Predictive AI- I saw it coming.
As a child, I’ve always dreamed of a digital version of me. I really did. I was a quiet, introverted child, always interested in books and learning and experiments. Very few kids at that age found this as interesting as I did. To be fair, I didn’t miss having friends at all. I found their games boring and tedious and I infinitely preferred my books to their games. I even remember telling my mother that I wanted to build a robot version of myself, someone who read the same books as I did, who could work on experiments with me and with whom I could discuss all the stuff that I had learnt. In other words, a digital copy of myself.
Later on, when I made my peace with some human friends over the years, they would always joke that I was immensely predictable, that they could tell exactly when I would eat, sleep or perform any of my regular activities. But if you think about it, on a long enough timeline, everyone and everything becomes predictable. My friends are I have been hanging out for years. Over time, they have unwittingly gathered enough information about me to see me as predictable when, actually, it’s very normal, almost scientific. There are patterns everywhere and predictions at every corner. This is the wonderful thing about prediction. It is what encourages me the most when I face challenges, the answer is so simple and it is all around, I just need to identify it.
Getting into AI happened fairly quickly. Initially hired as an iOS developer intern at Nine Connections, I found myself working with one of our predictive technology products — one that recommends the types and timings of articles that should be shared by news publishers on their social media accounts. I had always thought that working with iOS would be the best starting point to get into AI. However, in the second week of my internship, during a casual talk with one of the founders on how to improve the recommendations, I casually proposed the idea of using artificial intelligence. Pat came the response: “Do you wanna do it?” Suddenly I was the person in charge of developing our AI arm, something I truly found amazing.
Back before the term AI was fashionable, Nine Connections had implemented predictive algorithms, data mining principles and pattern recognitions into its technology. While being surrounded by all of this technology is super-interesting, I knew I had to start at the beginning. The beginning was, predictably, rocky. Back in college, I had taken many courses on algorithms but when it comes to developing AI, I knew that I had to go back to the beginning.So that’s what I did. My start was rather basic, by googling things like ‘results from given dataset’ and ‘patterns using algorithms’. I studied predictive models, researched the basic ideas governing AI and did a ton of research. This is another reason why I think I’m made for AI. Although there are some courses floating around, a lot of it is still about teaching yourself how to do things. I love the research and the academics that go into it. A self-learner by nature, I had to read a ton of stuff before I actually made any headway.
Building AI isn’t easy. Finding reliable sources is more difficult than you think. There’s a tremendous amount of data collection and analyzing initially before we start to build a prototype. There is a lot of information out there and I found it quite hard to tie the useful stuff together. I spent a lot of time going back to look at several models and algorithms before coming up with a model that could possibly be applied to our case and more importantly, would be successful. I think the most challenging part in what I do is to clean the large sets of data and make them comprehensible to our network. I often refer back to the academic articles when I struggle at some points. Our weekly team discussions also make myself easier by bringing some other point of view to the problem we are trying to solve. Although I tend to work alone, it feels like I have help when I need it.
Almost a year in, I am quite happy with the shape of what we’ve created. One powerful aspect of our product is that the tech behind it is easily adaptable. I see our product as becoming a more generic prediction engine that can predict practically everything as we improve it. Whether it is simply predicting the stock market prices, or robots that empathize with a person’s feelings and predict human behavior; the power of self-learning entities will make it achievable. I am aware that it is an ambitious plan, but I also have the feeling that we are on the right track as a team. And I personally am proud to be associated with it.
As for my digital companion, that’s still in the works. As I move forward in my education and learn more technical concepts, I’m becoming aware that modeling such a thing will be possible in the near future. Now that I am actually working on developing AI, it won’t be long before there is a digital copy of me walking around, nose buried in a book, thinking of new experiments and exciting futuristic things to build. I sincerely hope to meet him someday.
Fehim Hathipoglu
Fehim does artificial intelligence and machine learning at Nine Connections. This post was initially published on Linkedin. His original post can be found here.
