THERE ARE FEW THINGS THAT MATTER TO ME, BUT THEY DO IT DEEPLY.
I started programming in 2001 on the pursuit of teaching computers how to understand human languages. It frustrates me how stupid and annoying Siri, Cortana and Alexa can be sometimes. It doesn't have to be like that though.
THERE'S AN INSANE POTENTIAL ON ENABLING COMPUTERS TO UNDERSTAND THE WORLD.
The steam engine made manufacturing cheaper and transportation quicker. Electricity allowed humans to use the hours when there is no sun. Later, it let them preserve food for a longer time and heat it in a couple minutes. It took small time before it entered the entertainment industry, in the form of recorded music, movies and pictures. Personal computers allowed to store and process great amounts of information precisely and fastly. The internet connected people in remote places and created a whole new globalized economy. Smartphones brought the power of computers to their users' pockets.
ARTIFICIAL INTELLIGENCE IS NOT ABOUT REPLACING PEOPLE, BUT GIVING THEM SUPERPOWERS.
Imagine you're a musician. There's a tune in your head. You hum it to your computer and it makes scores of it for you. You can ask her to give it a guitar timbre. You tap on your table and she records the beats and turn them into the drum section.
You're a photographer. Your computer selects the best shots you took among thousands of them in a few seconds. You ask her to color correct it. She tells you it can make the sky bluer and delete the guy photobombing your perfect take over the bridge.
You're a doctor. You tell the symptoms and your observations of a patient to your computer. She suggests a few possible diagnosis. You have a strong lead over two of them, but you're not an expert on them. She summarizes the latest papers on both and emails them to you. Also, she contacts a leading researcher on the topic on your behalf.
WELCOME TO THE ERA OF DEEP LEARNING
If we were in the 1700s and I told you you could watch a concert happening on the other side of the world through a metal box, you might think I am a wizard or a lunatic. However, watching TV is totally natural for you now. Artificial neural networks have been around since the 1950s, but it wasn't until now that we have enough computing power, datasets and math developments to make it work.
Our universe has 4 perceptible dimensions for our senses: 3 spatial (width, length, depth) and time. Convolutional neural networks are capable of finding patterns in spatial data with the use of labeled samples. Recurrent neural networks can do so with temporal data and other kinds of sequences. There are also generative adversarial networks that compete for generating new data that feels real. Also, reinforcement learning which is responsible for self-driving cars and computers that can play video games.
All of them are an intersection of statistics, linear algebra and calculus. For production environments, also lots of patience, hacking datasets, coffee and Python code. So it's definitely not magic, but math is nonetheless magical.