Let's Kill the New Age Tech Buzzwords

As an observer of the recent tech trend, including "cloud-computing," "big data," "fintech," "deep learning" as well as "IoT," I find simple concepts seem to have been mystified as more people throw the above terms around, and big firms, including IBM's "iot" platform, have been quite accommodating to these buzzwords.

The more opaque these buzzwords seem, the higher the fee tech companies can charge clients. It almost resembles the structured products in finance before 2008: the less you understand something, the more miraculous it seems, and the more dangerous it becomes. Granted, the sheer admirable efforts of computer scientists and enthusiasts (me included) have been contributing and adding these endeavors, but the buzzword branding places unrealistic expectation on the performances of these methodologies. But in reality, these oracular-sounding words are...

Cloud-computing: a way to store and access information on servers that are not located within and operated by users. Recently, this concept has been gradually used instead for processing and analyzing information on computers that users do not own. Tech companies such as Amazon Web Services or Google has existing and enormous budgets for hardware, and it is much simpler to leave the maintenance burden to these firms. So, what kind of information that is so enormous that you have to leverage the mighty Amazon to analyze? This leads us to...

Big Data: Big data doesn't have a precise definition, but if the amount of information is sufficient to take your lunch money and crash your computer (aka: larger than your local memory), then you can call it data big. Big data can be traffic pattern, tweeter updates, Facebook photo uploads, Amazon's consumer buying behaviors and whatnot. Yes, big data is simply "a lot of stuff accumulated across millions of people." Big data is nothing new as information has existed for decades. What is different today is the cheapening of information storage cost. It allows firms like Google and Facebook to store as much information as they want. But without analytic, information is simply junk. That itself is being dealt with by something called "Deep Learning." Deep learning sounds deep, but it is essentially...

Deep Learning: It is based on a methodology called neural network, which has been around since the 1960s. Neural network sounds ass-kicking, but it is in essence a more sophisticated version of regression. There are many different ways you can go about learning something about some data.  But neural network is simply linear regression's Yale-educated, tailored-suit-wearing but emotionally unstable big cousin. In other words, it is a different way to find patterns in data. You can ask questions such as "Do taller people weigh more?" "Do republicans begin to find wig stores due to the ascendance of Donald Trump?" "Do the condition of oil tanks as reflected in the satellite images predict the economic growth of countries or world economy in general? (this one is real)". 

So, if neural network has been around since the 60s, why is it suddenly so hot? The answer is simple:

Big data + Big computer muscles = Uber-sexy Deep Learning.

Neural network suffered much from "overfitting," which means if you teach a computer to recognize your face and tell it it's what "humans look like." The computer will only see you as human and your cousin Jacky as non-human. Scholars has to teach computers to see "two ears" or "one nose." But neural network also suffered from a gruesomely slow training time, so it slowly faded into the woodwork. 

But in the past decade, scholars (yes, your average ivory tower eggheads invented almost everything you use today including Google) found out that, with sufficient computing power, hit it with huge amounts of data (aka: big data), neural network produces far superior results compared to many other artificial intelligence methodologies. It has since been used to solve a lot of real world problems, and it is the foundation of IBM Watson, which fiendishly smart that it is easy to quote him (it) on an impending Robot Apocalypse. But now, it can write new recipes and answer Jeopardy questions, which are pretty awesome by themselves.

Fintech means "a faster way to pay for your morning espresso or phone bills or help your cousin Tommy on his rent." Of course, these days, faster simply means "doing shit on your mobile phone." So, 90% of fintech firms are about using your mobile phone to a) make payments b) transfer money or c) save and investment money. 

IoT is often used in the same sentence with "smart" stuff. Smart fridge. Smart phone. Smart lighting. Smart house. Smart building. By "smart," it means "whatever you can control or see on your phone, hopefully automatically." How do we control it? Well, we use the "Internet-of-Things." Thus, IoT is simply a way to integrate stuff (think a house and appliances) over a network (think "cloud computing") to achieve some kind of purpose, again, hopefully automatically.

What kind of purpose? Say a city wants to reduce the load of power plants. People tend to turn air conditioning or heaters at roughly around the same time and this places enormous burden on the grid. The city collects all the usage data on how millions of households get home and turn on air conditioning during the summer over the span of several years. After enough "big data" is accumulated, analysts can use "deep learning" methodology to figure out the best way to save the electricity grid load. Through "clouds" and by connecting each household's appliance ("IoT") to a central command, the central computer can decide whose AC to turn on first and whose later to reduce the load. Viola, we just used all the buzzwords in two sentences.

I am of course oversimplifying things. Each of the above item requires years of training, millions of dollars of investments and teamwork. By demystifying these buzzwords, I hope people can see these new technologies less as god-given miracles but as fruits of strenuous efforts these professionals and scholars built with blood and sweat.