I Believe in Reader
06 Sep 2019If you’ve gone over my projects so far, you can see that I have a mild to severe interest in artificial intelligence. Like with most interests, my motives to study this area are rooted in pop culture and sensationalized headlines like “Could Artificial Intelligence lead to Human Extinction?” Most of the time, when my interest stems from sensation instead of real world examples, I will go through a sort of cycle of diminishing interest. This cycle usually goes from fiery passion to realization of reality to being intimidated by the actual work needed to chase this interest to ultimately losing interest and giving up. I’m happy to say, that with artificial intelligence, I’ve yet to reach the last stage.
After taking two semesters of machine learning courses, reading AI Superpowers by Kai-Fu Lee, and completing an internship in Data Science, I can safely say that my interest in machine learning or artificial intelligence has held strong. Sure, I went through a realization of reality stage while pursuing machine learning, except this time, the reality of the situation did not deter me in the slightest. When my expectations of murder bots was met with the training of neural nets, my ideas changed, but my interest only grew. Soon, I started to grasp the capabilities of neural nets and artificial intelligence which only helped to establish my motivation.
After reading AI Superpowers, I felt like I really knew how impactful this seemingly new technology will be in the coming decades and I decided to push further. My sustained interest has lead me to actually designing my own AI, one that can identify healthy cells within an image. It’s not perfect, not by a long shot, but I’m just happy it exists. I’m happy I built something that I can train, test, and run with moderate accuracy. I’m happy that I have something that can tell me what mistakes I’ve made so that my next AI can be better.
As I finish up my education, the time comes for me to finish a senior project. On display in my Projects folder is my page for ‘Reader.’ A stock trend forecasting algorithm that will apply neural nets to making predictions. The reason why I wanted to do this project instead of something that handled image-based datasets was because I wanted to take a step away from visual data and try to apply machine learning to other areas of life. I have goals of one days applying machine learning to cyber security or software development itself and I thought this was a step in the right direction.
I’ve just started this project, made some steps, and already put together a model, but I’m still somewhat intimidated by the scale of what I’m trying to accomplish. Even though I’ve already built an AI, I have lacking knowledge when it comes to finance and even less when it comes to using it to predict stocks. A lot of what I’ve been trying to do has mostly been encountering new terms and matching them to their definitions. In doing this, I’ve been able to gain a description for every factor fed into the model and have set out to make factors of my own, but I still can’t shake the feeling that I’m currently treading through uncharted territory.
In the end, we’ll see what shape my project happens to take. I’m hoping for something with more than 80% accuracy, but I’m confident I can design something worth competing.
Checking back in. The date is 9/16/2019 and I’ve come up with the idea to input the GDPs of relevant countries as a factor for predicting stock trends. The idea is to input countries closely tied to the American economy, China, Japan, Korea, Saudi Arabia, etc, then check it’s impact on predicting American stock trends.
After observing their effect, the next step will be to input less relevant countries and see if there is any hidden trends there.