By Michael Castaneda
Published: February 27th, 2019
A lot has been written about Artificial Intelligence (AI) and the effects it could have on our future. Recently, the topic has received more coverage thanks to a threeday symposium at MIT last month. Discussions and recommendations on how AI can be regulated through new laws were some of the central points made during the conference. Unfortunately, there were only a few acknowledgments on how AI works and the effects it already has on our lives. Because of that, I want to concentrate on the exciting aspects of theoretical AI, which alternatively, is known as Machine Learning (ML).
Machine Learning is the process in which software’s become self-autonomous through applications of algorithms and statistics, allowing them to recognize patterns and make inferences.
Data is what software’s rely on to grow. The more data software’s ingest, the more accurate its results would be. Currently, we have so much data in the world that machines today are still too far behind to ingest it all. For example, someone can write a machine learning process and put it out into the internet and it will set itself to learn. In fact, this has been done many times. If someone wants to learn about speech, they can set an ML process to listen to talk radio. (Note: by this stage, problems can arise because there is so much hate on the internet and talk radio that ML processes have learned racial bias. Also, remember that some of these processes will be used to determining bank loans and who might get arrested. But we will return to that.)
The levels of sophistication of the training determine how the ML will learn, and there are many different approaches to it. Some have another process that acts as a reinforcement agent in the learning process, whilst some are more of a Black box effort. The general theme is to take varying levels of approximation and hone it down to something more concrete. That is called Deep Learning, where it correlates to another theme called Neural Networks. The goal of Neural Networks is that it tries to map computer processes as similar as possible to what our brains have.
But does it really achieve that? No, and it doesn’t matter. Computers don’t do math the way we do. They can’t handle the symbol manipulations of Calculus, for example. But with numerical methods, computers do math better than most people except for maybe mathematical prodigies like Dr. Arthur Benjamin. Given this, computers have become really good at recognizing shapes, which makes self-driving cars more of a reality. However, the context of it still poses a challenge, which is key.
Without context, machine learning can be a little scary. If machine learning systems can execute actions without remorse or favors the norm without context, then not only are there possibilities of danger, but also a very boring digital existence.
There have been machine learning algorithms that identify people with darker skin as a threat. This is great if you are a white criminal, but not so good if you are Black, Indian, or a dark-skinned Hispanic or Asian.
Also, it would be boring because say I go down a rabbit hole of retro Jim Carrey movies on Netflix, Deep learning would type me and never let me learn the genius of Taylor Perry. I’ll just be watching dumb 90’s comedies for the rest of my life. I admit I am exaggerating here, but you get the point.
Who Is Using It?
The two big players are, of course, the United States and China. China uses machine learning to keep track of every citizen of China and possibly everyone they can in the world. They keep track of everything you do and give you a score based on how well you are behaving as per the Communist Party’s rules. If you are not behaving as they think you should, well, we don’t hear too much from those people.
In the US, every person is being digitally strip-mined and US Companies are making billions off of it. Silicon Valley has converted our lives into data in its database. Every piece of information about us from our conversations, shopping histories, our browser histories and if you let them, your DNA is being run through massive machine learning algorithms for purposes that are unknown to us.
Technology is neither good nor bad a lot of the time, but how and what it’s being used for is what makes the difference. There are many useful ways ML has made our Google searches better, and there are many instances where I would like to see improvement. For one, when I type “New” on my iPhone, why won’t it ever suggest “York”?
Machine learning processes can also be applied to many other fields. When it processes every type of cancer known, it can shed light on a disease that might have slipped through the cracks.
Levels of gun violence can also be lowered because it turns out that people who commit these massacres have similar credit card behaviors; such as stockpiling guns and ammo over a short period of time by opening lots of new credit cards to achieve that. Credit companies are aware of that behavior but don’t report it, because they are not required to.
Therefore, sensible laws are needed.
It is easy to anticipate counterarguments – if companies and governments are to be limited in their use of machine learning, then our streets will not be safe, companies earning will be stymied, and human progress will be limited. Thus, it is a good time to get ahead of these curves before Fox News starts a 24/7 propaganda campaign against machine learning regulations.
The first thing to ask for is a cogent disclose on how your information is being used and who owns the data on your habits. Whatever rewards points you get at Duane Reade is certainly not enough to let any business entity the rights to gain power over you, your family, or your future family.